Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
1
5.
Benefits
Analysis
5.1
Introduction
The
proposed
LT2ESWTR
will
reduce
the
occurrence
of
viable
waterborne
pathogens,
particularly
Cryptosporidium,
in
drinking
water
delivered
by
public
water
supplies
that
use
surface
water
or
ground
water
under
the
direct
influence
of
surface
water
(
GWUDI).
The
quantified
health
benefits
estimated
for
this
rule
result
from
reducing
the
incidence
of
adverse
health
effects
(
illnesses
and
possible
premature
death)
due
to
drinking
water
containing
Cryptosporidium
and
other
pathogens.

Section
5.2
describes
the
risk
assessment
used
to
estimate
the
number
of
illnesses
and
deaths
associated
with
endemic
cryptosporidiosis
that
will
be
avoided
because
of
the
LT2ESWTR.
Section
5.3
describes
the
methods
for
monetizing
these
benefits.
The
quantified
benefits
presented
in
this
chapter
are
not
the
only
ones
expected
from
the
implementation
of
this
rule.
Other
benefits,
including
those
associated
with
reductions
in
sporadic
cryptosporidiosis
outbreaks,
reductions
in
endemic
illnesses
and
outbreaks
from
other
pathogens,
and
improved
aesthetic
water
quality,
are
described
in
section
5.4,
but
are
not
quantified
in
this
analysis.
In
addition,
the
benefits
realized
from
regulating
uncovered
finished
water
reservoirs,
a
provision
of
the
proposed
rule,
are
not
quantified.
A
summary
of
uncertainties
is
included
as
section
5.5,
and
a
comparison
of
the
monetized
benefits
of
the
regulatory
alternatives
are
presented
in
section
5.6.

The
remainder
of
this
chapter
is
organized
as
follows.

5.2
Quantified
Health
Benefits
from
Reduction
in
Exposure
to
Cryptosporidium
5.2.1
Overview
of
Risk
Assessment
Methodology
5.2.2
Hazard
Identification
5.2.3
Dose­
Response
Assessment
5.2.4
Exposure
Assessment
5.2.5
Risk
Model
Structure
5.2.6
Individual
Annual
Risk
Distributions
5.2.7
General
Population
Risk
 
Number
of
Cases
Avoided
5.2.8
Reduction
in
Sensitive
Subpopulation
Risk
5.3
Monetized
Benefits
from
Reduction
in
Exposure
to
Cryptosporidium
Resulting
from
the
LT2ESWTR
5.3.1
Value
of
Reduction
in
Cryptosporidiosis
Cases
5.3.2
Monetization
of
Benefits
to
Sensitive
Subpopulations
5.4
Other
Benefits
of
LT2ESWTR
Provisions
5.4.1
Reduction
in
Outbreak
Risk
5.4.2
Costs
to
Households
to
Avert
Infection
5.4.3
Enhanced
Aesthetic
Water
Quality
5.4.4
Risk
Reduction
from
Co­
occurring
and
Emerging
Pathogens
5.4.5
Benefits
from
Other
Rule
Provisions
5.4.6
Summary
of
Nonquantified
Benefits
5.5
Summary
of
Uncertainties
5.6
Comparison
of
Regulatory
Alternatives
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
2
5.2
Quantified
Health
Benefits
from
Reduction
in
Exposure
to
Cryptosporidium
This
section
describes
the
risk
assessment
methods
and
assumptions
used
to
quantify
the
expected
health
benefits
of
the
LT2ESWTR.
It
also
provides
the
results
of
these
calculations,
expressed
in
terms
of
reduced
cases
of
illness
and
avoided
deaths.

The
quantified
health
benefits
presented
in
sections
5.2
and
5.3
are
derived
from
estimates
of
the
Pre­
LT2ESWTR
annual
levels
of
illness
and
death
caused
by
endemic
exposure
to
Cryptosporidium
in
drinking
water,
and
the
reductions
expected
as
a
result
of
the
LT2ESWTR.
Annual
endemic
cases
are
those
occurring
as
a
result
of
Cryptosporidium
present
in
drinking
water
under
normal
operating
conditions.
This
endemic
level
does
not
include
illnesses
and
deaths
attributable
to
outbreaks
of
cryptosporidiosis
 
those
that
are
associated
with
events
or
conditions
that
are
outside
of
normal
treatment
plant
operating
conditions.

Endemic
levels
of
cryptosporidiosis
cannot
be
measured
directly
because
symptoms
are
generally
underreported
(
relatively
few
seek
medical
attention),
and
because
there
are
many
potential
causes
of
gastrointestinal
illness
resembling
cryptosporidiosis.
Usually
only
in
an
outbreak
will
doctors
test
stool
samples
for
Cryptosporidium.

Even
outbreaks
are
not
always
recognized,
again
because
symptoms
are
underreported
and
not
always
recognized
as
being
due
to
cryptosporidiosis.
Data
on
occurrence
specifically
related
to
outbreaks
were
not
available
and
dependable
methods
to
model
the
future
occurrence
of
outbreaks
have
not
been
proven.
Because
of
these
difficulties,
the
incidence
of
illness
is
modeled
(
as
opposed
to
directly
measured),
and
only
for
endemic
illnesses.
Thus,
the
illnesses
estimated
quantitatively
in
this
Economic
Analysis
(
EA)
should
be
thought
of
as
representing
a
steady,
underlying
level
of
illness
unadjusted
for
outbreaks
of
the
disease.

The
risk
assessment
used
to
calculate
the
benefits
of
the
LT2ESTWR
involves
a
two­
dimensional
Monte
Carlo
simulation
model
designed
to
explicitly
consider
probability
distributions
describing
the
uncertainty
in
some
of
the
model
inputs,
and
the
inherent
variability
in
others.
The
structure
of
this
model,
and
the
basis
for
the
characterization
of
the
uncertainty
and
variability
distributions
used
in
it,
are
described
in
more
detail
in
this
chapter.
The
calculations
for
the
model
were
carried
out
in
SAS
v8.2
(
Appendix
T
provides
details
of
the
programming
code,
input
data,
and
output
results).

The
risk
assessment
is
designed
to
compare
benefit
estimates
(
reductions
in
risk)
across
key
variables:
system
type
and
size,
treatment
(
filtered
or
unfiltered),
and
occurrence
data
sets
of
Cryptosporidium
in
source
water
(
Exhibit
5.1).
Changing
these
variables
between
model
runs
tests
the
assumptions
regarding
Cryptosporidium
occurrence
in
source
water,
annual
exposure,
current
treatment
in
place,
and
options
for
treatment
in
response
to
the
LT2ESWTR.
The
model­
to­
model
comparisons
provide
an
additional
measure
of
variability,
among
types
of
systems,
in
addition
to
the
variability
within
each
type
of
system
addressed
by
the
core
model.
Model
runs
produce
total
national
benefit
estimates
as
well
as
break
outs
for
these
variables.
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
3
Exhibit
5.1
Risk
Assessment
Model
Categories
Data
Set
System
Size
(
population
served)
Filtered
Plants
Unfiltered
Plants1
CWS
NTNCWS
TNCWS2
CWS
ICR
Very
Small
(
#
500)
T
T
T
T
Small
(
501­
10,000)
T
T
T
T
Medium
(
10,001­
100,000)
T
T
T
T
Large
(>
100,000)
T
T
T
ICRSSM
Very
Small
(
#
500)
T
T
T
Small
(
501­
10,000)
T
T
T
Medium
(
10,001­
100,000)
T
T
T
Large
(>
100,000)
T
T
ICRSSL
Very
Small
(
#
500)
T
T
T
Small
(
501­
10,000)
T
T
T
Medium
(
10,001­
100,000)
T
T
T
Large
(>
100,000)
T
T
Note:
[
1]
The
ICRSSM
and
ICRSSL
have
no
unfiltered
source
water
occurrence
data.
Section
5.2.7.1
describes
how
unfiltered
ICR
data
were
adjusted
to
produce
risk
estimates
for
the
ICR,
ICRSSM,
and
ICRSSL
data
sets.
There
is
only
one
unfiltered
noncommunity
water
systems
(
NCWS);
it
was
grouped
with
community
water
systems
(
CWSs)
for
data
analysis.
[
2]
There
are
no
large
filtered
transient
noncommunity
water
systems
(
TNCWSs).

Sections
5.2.1
through
5.2.5
present
the
risk
assessment
methodology.
Model
results
for
the
baseline
and
four
regulatory
alternatives
considered
in
this
Economic
Analysis
(
EA)
are
presented
and
discussed
in
sections
5.2.6
through
5.2.8,
as
well
as
in
Appendices
C
and
O.

5.2.1
Overview
of
Risk
Assessment
Methodology
Risk
assessment
is
an
analytical
tool
that
is
used
to
characterize
the
expected
incidence
of
adverse
health
effects
associated
with
exposure
to
an
environmental
hazard,
in
this
case
Cryptosporidium.
It
is
also
used
to
estimate
the
benefits
of
actions
taken
to
reduce
exposure
to
that
hazard.

The
risk
assessment
used
to
estimate
the
potential
benefits
of
the
LT2ESWTR
comports
with
a
standard
framework
for
risk
assessment
employed
by
the
U.
S.
Environmental
Protection
Agency
(
EPA).
The
framework
is
organized
in
accordance
with
the
EPA
Policy
for
Risk
Characterization
(
USEPA
1995a),
EPA's
Guidance
for
Risk
Characterization
(
USEPA
1995b),
and
EPA's
Policy
for
Use
of
Probabilistic
Analysis
in
Risk
Assessment
(
USEPA
1997a).
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
4
This
standard
framework
requires
the
use
of
scientific
data
(
or
reasonable
assumptions
if
data
are
not
available)
to
produce
estimates
of
the
nature,
extent,
and
degree
of
a
risk.
When
the
risks
posed
are
not
the
same
for
all
persons,
that
variability
in
risk
should
be
described.
Further,
data
are
seldom
known
with
certainty,
and
therefore,
that
uncertainty
must
be
described
and
its
impact
on
the
risk
estimates
characterized.
The
risk
assessment
used
here
incorporates
both
types
of
information
 
the
variability
associated
with
the
distribution
of
risk
levels
within
the
affected
population
and
the
uncertainty
expressed
by
confidence
bounds.

According
to
the
1995
EPA
Policy
for
Risk
Characterization
(
USEPA
1995a),
health
risk
assessments
for
environmental
contaminants
generally
have
three
components
that
together
comprise
risk
characterization:

°
Hazard
Identification
addresses
the
nature
of
the
potential
adverse
health
effects
associated
with
exposure
to
the
contaminant.

°
Dose­
Response
Assessment
addresses
information
concerning
the
relationships,
quantitative
where
possible,
between
the
magnitude
of
exposure
to
the
contaminant
and
the
extent
and
severity
of
the
adverse
health
effects.

°
Exposure
Assessment
estimates
both
the
number
of
people
in
the
population
exposed
to
the
contaminant
and
the
distribution
of
levels
of
exposure
within
that
population.

Risk
characterization
combines
the
hazard
identification,
dose­
response
assessment,
and
exposure
assessment
information
to
describe
overall
risk
to
the
exposed
population,
in
terms
of
both
the
distribution
of
risk
levels
in
the
population
and
the
total
number
of
cases
of
adverse
effects
expected.

Exhibit
5.2
depicts
these
elements
of
the
risk
assessment
for
characterizing
the
endemic
risk
of
illness
and
death
from
exposure
to
Cryptosporidium
in
drinking
water
systems.

To
derive
benefit
estimates
of
the
LT2ESWTR
using
this
risk
assessment
framework,
the
analysis
calculates
the
difference
between
illness
and
death
estimates
for
the
baseline
(
Pre­
LT2ESWTR)
condition
and
illness
and
death
estimates
after
implementation
of
the
LT2ESWTR.
Benefit
estimates
are
the
number
of
illnesses
and
deaths
avoided
because
of
a
regulatory
requirement
(
i.
e.,
Pre­
LT2ESWTR
and
the
four
regulatory
alternatives
for
LT2ESWTR
described
in
section
3.3).
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
5
Hazard
Identification
°
Cryptosporidium
health
endpoints:
Illness
(
cryptosporidiosis)
and
Mortality
Risk
Characterization
°
Estimated
cases
of
illness
and
death
in
the
affected
population
°
Distribution
of
individual
risk
levels
of
illness
and
mortality
Dose­
Response
Assessments
Relationships
for
probability
of:
°
Infection
given
exposure
°
Illness
given
infection
°
Death
given
illness
Exposure
Assessment
°
Number
of
people
exposed
to
Cryptosporidium
in
finished
drinking
water
°
Distribution
of
average
daily
Cryptosporidium
ingestion
levels
across
the
exposed
population
Exhibit
5.2
Health
Risk
Assessment
Framework
for
Cryptosporidium
5.2.2
Hazard
Identification
This
section
presents
summary
information
on
the
adverse
health
effects
associated
with
Cryptosporidium
ingestion
from
drinking
water,
including
a
discussion
of
cryptosporidiosis
and
the
potential
for
illness.
It
also
discusses
the
potential
for
mortality,
particularly
among
the
immunocompromised.
For
further
information
on
health
effects
associated
with
Cryptosporidium,
see
Chapter
2
of
this
document
as
well
as
the
Occurrence
and
Exposure
Assessment
for
the
Long
Term
2
Enhanced
Surface
Water
Treatment
Rule
(
USEPA
2003c).

Ingesting
Cryptosporidium
oocysts
can
cause
cryptosporidiosis,
which
typically
is
an
acute,
selflimiting
illness
with
symptoms
that
include
diarrhea,
abdominal
cramping,
nausea,
vomiting,
and
fever
(
Juranek
1995).
There
is
no
treatment
that
can
eliminate
a
Cryptosporidium
infection,
and
only
a
few
antiparasite
or
antimicrobial
agents
have
shown
even
a
slight
ability
to
reduce
a
patient's
parasite
load
(
Guerrant
1997).
In
some
occurrences,
cryptosporidiosis
can
be
fatal,
particularly
among
subpopulations
such
as
Acquired
Immunodeficiency
Syndrome
(
AIDS)
patients,
the
elderly
with
other
underlying
illnesses,
and
other
immuno­
compromised
individuals.
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
6
Limited
information
is
available
on
the
endemic
incidence
of
cryptosporidiosis
in
the
United
States.
Mead
et
al.
(
1999)
have
estimated
that
there
are
approximately
15
million
physician
visits
annually
for
diarrhea,
and
that
approximately
2
percent
of
these,
or
300,000
cases,
are
due
to
cryptosporidiosis.
They
also
estimate
that
of
these
300,000
cases,
only
about
10
percent
are
attributable
to
food­
borne
transmission,
with
the
remainder
due
to
the
consumption
of
contaminated
water
(
from
drinking
or
recreational
exposure)
or
person­
to­
person
contact.
The
number
of
endemic
cases
estimated
by
Mead
et
al.
is
probably
low
because
not
all
who
experience
diarrhea
visit
a
physician.
Mead
et
al.
estimate
that
there
are
approximately
211
million
episodes
of
gastroenteritis
in
the
United
States
each
year,
of
which
only
about
38
million
are
attributable
to
known
pathogens.
Published
information
does
not
provide
estimates
of
cases
due
solely
to
drinking
water.

Another
potential
indicator
of
endemic
cryptosporidiosis
is
the
fraction
of
the
human
population
that
has
positive
antibodies
against
Cryptosporidium.
Studies
of
a
variety
of
populations
have
found
a
reactivity
to
C.
parvum
antigens
in
25
to
35
percent
of
adults;
this
number
is
even
higher
in
developing
countries
(
Chappell
et
al.
1999).
While
these
positive
reactions
indicate
past
exposure
to
Cryptosporidium,
the
number
of
exposures,
durations
of
illness,
and
individual
susceptibilities
to
reinfection
cannot
be
determined
by
simple
testing.
Levels
of
one
antigen,
IgG,
have
been
observed
to
drop
over
time,
so
a
high
level
of
IgG
antibodies
in
an
individual
can
be
an
indicator
of
recent
exposure
or
infection
(
van
Herck
et
al.
2000;
de
Melker
et
al.
2000),
but
since
the
rate
of
decrease
is
not
known
and
possibly
differs
for
each
individual,
it
is
difficult
to
estimate
endemic
rates
through
immunology.

Many,
probably
most,
infected
individuals
do
not
seek
medical
treatment
for
their
symptoms.
If
they
do
seek
medical
treatment,
primary
care
physicians
may
not
be
able
to
isolate
Cryptosporidium
as
the
cause
of
the
illness.
If
diagnosed,
physicians
may
not
report
the
information
to
the
Centers
for
Disease
Control
(
CDC).
These
compounded
effects
could
lead
to
gross
underreporting
and
underestimating
of
cryptosporidiosis
cases
(
Okun
et
al.
1997).
Additionally,
individuals
can
be
infected
with
Cryptosporidium
yet
exhibit
no
symptoms
of
infection.
This
is
seen
more
among
people
without
pre­
existing
immunity
to
Cryptosporidium,
and
can
further
hide
the
true
incidence
of
cryptosporidiosis
in
the
United
States.

Although
the
focus
of
the
risk
and
benefits
analysis
conducted
here
is
on
endemic
cases,
most
of
the
information
available
on
the
health
hazards
from
exposure
to
Cryptosporidium
derives
from
studies
involving
outbreaks.
The
1993
outbreak
in
Milwaukee,
Wisconsin
was
the
largest
recorded
outbreak
of
waterborne
disease
in
the
United
States.
Using
standard
epidemiological
methods,
CDC
estimated
that
of
approximately
800,000
people
served
by
the
water
system,
over
400,000
(
50
percent)
became
ill
(
Craun
et
al.
1998).
Of
those,
4,000
required
hospitalization
(
approximately
1
percent
of
those
becoming
ill),
and
there
were
54
cryptosporidiosis­
associated
deaths,
with
at
least
46
of
these
being
immunocompromised
individuals
(
as
reported
on
death
certificates)
(
Mackenzie
et
al.
1994;
Hoxie
et
al.
1997).

Several
subpopulations
are
more
sensitive
to
cryptosporidiosis,
including
the
young,
elderly
with
other
underlying
illnesses,
malnourished,
disease­
impaired
(
especially
those
with
diabetes),
and
a
broad
category
of
those
with
compromised
immune
systems
(
Rose
1997).
Subpopulations
with
compromised
immune
systems
include
AIDS
patients,
those
with
lupus
or
cystic
fibrosis,
transplant
recipients,
and
those
on
chemotherapy
(
Rose
1997).
Symptoms
in
the
immunocompromised
subpopulations
are
much
more
severe,
including
debilitating,
voluminous
diarrhea
that
may
be
accompanied
by
severe
abdominal
cramps,
weight
loss,
malaise,
and
low­
grade
fever
(
Juranek
1995).
Moreover,
mortality
is
a
substantial
threat
to
the
immunocompromised
infected
with
Cryptosporidium:
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
7
The
duration
and
severity
of
the
disease
are
significant:
whereas
1
percent
of
the
immunocompetent
population
may
be
hospitalized
with
very
little
risk
of
mortality
(<
0.001),
Cryptosporidium
infections
are
associated
with
a
high
rate
of
mortality
in
the
immunocompromised
(
50
percent).
(
Rose
1997).

Exhibit
5.3
contains
detailed
information
on
some
of
the
cryptosporidiosis
symptoms
observed
during
the
Milwaukee
outbreak.

Exhibit
5.3
Symptoms
of
205
Patients
with
Confirmed
Cases
of
Cryptosporidiosis
During
the
Milwaukee
Outbreak
Symptom
Percent
of
Patients
Mean
Range
Diarrhea
93%
~
12
days
duration
1
 
55
days
duration
Abdominal
Cramps
84%
N/
A
N/
A
Weight
Loss
75%
10
pounds
1
 
40
pounds
Fever
57%
100.9
/
F
99.0
/
 
104.9
/
F
Vomiting
48%
N/
A
N/
A
N/
A:
Not
applicable.
Source:
Mackenzie
et
al.
1994.

Although
the
Milwaukee
outbreak
represents
the
largest
number
of
cases
in
a
single
cryptosporidiosis
outbreak
in
the
United
States,
most
identified
outbreaks
have
occurred
in
small
systems
serving
fewer
than
10,000
people.
Between
1991
and
1996,
6
outbreaks
caused
by
Cryptosporidium
in
small
water
systems
resulted
in
271
reported
cases
of
cryptosporidiosis
and
3,822
estimated
cases
(
USEPA
2003c).
Three
of
the
six
outbreaks
were
in
small
surface
water
systems,
and
three
occurred
in
GWUDI
systems.
During
small­
system
outbreaks,
the
percent
of
the
exposed
population
becoming
ill
ranged
from
8
to
70
percent.

Again,
outbreak
cases
are
believed
to
represent
only
a
portion
of
the
total
incidence
of
cryptosporidiosis.
Only
large
outbreaks
cases
concentrated
in
a
specific
location
are
likely
to
be
detected
and
reported.
Endemic
cases
(
which
are
the
focus
of
this
analysis)
and
smaller
outbreaks
are
less
likely
to
be
identified.

5.2.3
Dose­
Response
Assessment
This
section
presents
information
on
the
relationship
between
ingestion
of
Cryptosporidium
oocysts
and
infection,
illness,
and
mortality.
A
dose­
response
model
incorporates
that
information
to
estimate
the
probability
of
Cryptosporidium
infection.
This
model
is
the
first
of
two
steps
of
the
risk
assessment
model.
The
specific
variables
of
the
dose­
response
model
are
described
in
this
section
and
include
the
following:

°
Infectivity
dose­
response
function
1
A
person
is
considered
ill
due
to
Cryptosporidium
if
they
have
symptoms
of
cryptosporidiosis.
A
person
is
considered
infected
if
their
stool
contains
oocysts.

Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
8
°
Morbidity
factor
used
to
compute
the
risk
of
illness,
given
that
an
infection
has
occurred
°
Mortality
factor
to
compute
the
risk
of
death,
given
that
an
illness
has
occurred
Previous
risk
assessments
by
Haas
et
al.
(
1996),
Rose
(
1997),
and
Teunis
et
al.
(
2002)
have
focused
on
assessing
the
dose­
response
relationship
and
exposure
risks
of
Cryptosporidium.
The
risk
assessment
for
waterborne
cryptosporidiosis
done
by
Haas
et
al.
(
1996)
took
existing
data
on
Cryptosporidium
infectivity
and
used
an
exponential
dose­
response
model
to
determine
the
median
infectious
dose
(
ID
50).
This
information
was
then
used
to
determine
a
dose­
morbidity
ratio,
and
the
finished
water
concentration
of
oocysts
that
would
be
acceptable
given
guidelines
for
annual
risk
of
infection
from
any
one
type
of
pathogen.
The
infectivity
data
used
were
from
healthy,
pre­
immune
subjects
and
were
extrapolated
to
the
broader
population.

Cryptosporidium
occurs
in
slightly
different
forms,
and
the
forms
of
concern
here
are
those
infectious
to
humans.
Several
Cryptosporidium
varieties
(
called
isolates)
have
been
collected
and
cultivated
for
use
in
clinical
trials.
Two
studies
by
Teunis
et
al.
(
2002)
examined
the
dose­
response
variation
between
different
isolates
of
Cryptosporidium
and
between
hosts
with
varying
immune
responses.
These
studies
concluded
that
for
Cryptosporidium,
illness
and
infectivity1
vary
among
isolates,
and
that
for
individuals
with
elevated
IgG
levels,
the
probability
of
infection
drops
sharply
and
the
probability
of
illness
(
given
infection)
decreases
slightly.
The
assessment
by
Rose
(
1997)
added
to
the
previous
work
by
documenting
sources
of
Cryptosporidium,
those
populations
that
are
more
susceptible
to
infection,
historical
cryptosporidosis
outbreaks,
and
the
occurrence
of
Cryptosporidium
in
the
environment.

Infectivity
Dose­
Response
Model
The
dose­
response
model
used
to
characterize
the
likelihood
of
infection
from
ingestion
of
water
containing
Crytposporidium
in
the
risk
assessment
and
benefits
analysis
in
this
EA
is
taken
from
the
work
of
Haas
et
al.
(
1996).
The
basic
form
of
this
dose­
response
model
is:

P
I(
d,
r)
=
1­
e­
dr
where:

P
I(
d,
r)
is
the
probability
of
an
individual
becoming
infected
following
ingestion
of
water
providing
an
expected
dose
d
(
number)
of
organisms,
each
having
the
expected
probability
r
of
surviving
to
cause
an
infection.

The
"
expected
dose"
d
of
organisms
is
the
product
the
average
concentration
of
organisms
(
oocysts)
in
the
water
being
ingested
and
the
volume
of
water
ingested.
An
"
expected
dose"
of
1.0
oocyst
means
that
for
some
portion
of
time,
none
would
be
consumed,
and
for
the
rest
of
the
time,
one
or
more
oocysts
would
be
consumed.
If
the
average
Crytposporidium
concentration
is
measured
to
be
100
oocysts
per
100
liters,
or
1.0
oocyst
per
liter,
and
an
individual
consumes
1
liter
of
water,
the
expected
dose
d
for
the
dose­
response
model
would
1.0
oocyst.
Although
the
actual
number
of
organisms
that
any
2
The
Poisson
model
has
the
general
form
P(
X=
x)
=
(
8
xe­
8
)
/
x!
indicating
that
the
probability
that
a
specific
number
of
events
X
will
equal
some
specified
value
x
is
related
to
some
average
or
expected
number
of
those
events
given
by
8
.
So
in
this
example,
the
probability
that
X
=
0
(
oocysts)
given
that
8
=
1
is:
(
10
*
e­
1)
/
(
1)!
=
(
1)(
0.368)
/
(
1)
=
0.368.
Therefore,
the
probability
of
ingesting
1
or
more
oocysts
1
 
0.368
=
0.632.

Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
9
individual
would
consume
would
be
some
discrete
whole
number
(
including
zero),
the
average
number
per
unit
volume
of
water
may
be
a
fractional
value.

The
expected
organism
survival
probability
r
is
a
measure
of
the
inherent
infectivity
of
a
particular
strain
of
Crytposporidium
organisms
in
the
water
being
consumed.
As
a
probability
measure,
the
value
of
r
must
fall
between
0
and
1.
Values
approaching
1
indicate
a
highly
infectious
strain,
implying
that
individual
organisms
in
that
strain
are
highly
likely,
if
ingested,
to
survive
long
enough
to
initiate
an
infection.
Conversely,
values
approaching
zero
imply
that
ingested
organisms
of
that
strain
are
not
likely
to
survive
to
initiate
an
infection,
and,
therefore,
the
strain
is
one
of
low
infectivity.

As
an
example
of
the
dose­
response
model,
consider
an
individual
who
on
a
given
day
consumes
1
liter
of
water
that
has
an
average
Crytposporidium
concentration
of
1
oocyst
per
liter
of
a
highly
infectious
strain
with
an
r
value
of
1.

P
I(
d,
r)
=
1­
e­(
1)(
1)
=
1
 
e­
1
=
1
 
0.368
=
0.632
The
dose­
response
model
indicates,
therefore,
that
there
is
a
63.2
percent
chance
that
this
individual
will
become
infected.

This
outcome
of
only
a
63.2
percent
chance
of
getting
infected
rather
than
a
100
percent
probability
may
seem
questionable
at
first,
given
that
the
example
describes
an
individual
ingesting
1
liter
of
water
having
an
average
concentration
of
1
oocyst
per
liter
with
an
infectivity
probability
of
1.

The
reason
that
the
probability
of
infection
in
this
case
is
less
than
100
percent
reflects
the
"
expected
dose"
aspect
of
the
model.
The
dose­
response
model
is
actually
built
from
two
components.
The
first
of
these
components
is
a
Poisson
distribution
that
describes
the
probability
of
ingesting
one
or
more
oocysts
given
the
average
concentration
in
water
and
the
amount
of
water
consumed.
The
second
component
is
a
binomial
distribution
that
describes
the
probability
that
a
sufficient
number
of
organisms
will
survive
to
cause
an
infection,
which
is
the
function
of
the
expected
number
ingested
based
on
d
from
the
Poisson
component
and
their
survival
probability
r.

So,
when
r
=
1
indicating
certainty
that
an
infection
will
occur
even
if
only
one
oocyst
is
ingested,
the
Poisson
component
of
the
model
indicates
that
there
is
a
36.8
percent
probability
that
an
individual
consuming
1
liter
of
water
known
to
have
an
average
of
1
oocyst
per
liter
will
instead
ingest
zero
oocysts.
Therefore,
there
is
a
63.2
percent
chance
of
ingesting
1
or
more
oocysts.
2
Since
the
oocysts
in
this
particular
example
have
an
r
value
of
1,
i.
e.,
certainty
of
survival
to
cause
an
infection,
the
overall
resulting
probability
of
infection
is
also
63.2
percent.

A
full
description
of
the
derivation
of
the
above
dose­
response
model
from
its
component
Poisson
and
binomial
parts
is
provided
by
Haas
et
al.
(
1996).

It
is
important
to
note
that
the
basic
dose­
response
model
described
above
is
for
a
single­
exposure
situation.
That
is,
it
describes
what
the
probability
of
infection
is
for
an
individual
with
a
single
exposure
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
10
to
Cryptosporidium
in
water
with
an
expected
dose
of
d
of
organisms
with
survival
probability
r.
It
does
not
directly
describe
the
more
pertinent
concern
of
what
the
probability
would
be
of
an
individual
becoming
infected
over
time
 
for
example,
an
entire
year
 
from
repeated
exposures
to
that
strain
of
Cryptosporidium
in
water
with
an
expected
dose
of
d
each
day.

This
information
can,
however,
be
derived
from
the
dose­
response
model
for
a
single
exposure
in
a
fairly
straightforward
manner.
While
the
model
as
shown
gives
the
probability
of
becoming
infected
from
that
single
exposure
(
say
it
is
the
first
of
350
days
of
anticipated
exposure
over
a
year
from
a
CWS),
it
can
also
easily
be
transformed
to
provide
the
probability
of
not
becoming
infected
that
day
as:

P
Ø(
d,
r)
=
1­
P
I(
d,
r)
=
1­
(
1­
e­
dr)
=
e­
dr
Assuming
independence
from
one
day
to
the
next
with
respect
to
not
becoming
infected,
the
probability
of
not
becoming
infected
after
n
days
of
exposure
would
be
the
cumulative
product
of
the
individual
daily
probabilities
so
that
the
probability
of
not
becoming
infected
after
n
days
of
exposure
would
be:

P
Ø(
d,
r,
n)
=
(
e­
dr)
n
=
e­
drn
Because
this
is
the
probability
of
never
getting
infected
after
n
days
of
exposure,
one
can
then
readily
convert
this
to
the
probability
of
"
ever"
getting
infected
after
n
days
of
exposure
as:

P
I(
d,
r,
n)
=
1­
P
Ø(
d,
r,
n)
=
1­
e­
drn
This
is
the
form
of
the
model
that
is
used
in
the
LT2ESWTR
risk
and
benefits
modeling.

The
probability
of
"
ever"
getting
infected
includes
the
probability
of
more
than
one
infection
over
the
period
of
exposure,
but
this
form
of
the
model
does
not
explicitly
allow
for
estimating
the
specific
probability
of
1
versus
2
versus
3
infections,
etc.
While
such
modeling
could
be
done
(
it
would
be
a
more
complex
model
where
the
outcome
of
1
day
would
not
be
entirely
independent
of
positive
outcomes
on
previous
days)
it
would
not
add
meaningful
information
to
what
is
provided
by
the
model
currently
being
used.
This
is
because
the
values
for
the
expected
dose
and
survivability
parameters
based
on
actual
Cryptosporidum
data
result
in
very
low
probabilities
of
an
individual
ever
getting
infected
over
a
year.
Overwhelmingly,
these
"
ever
infected"
outcomes
predicted
by
the
current
model
are
expected
to
be
only
one
infection
over
the
n
days,
with
negligible
incidence
of
multiple
infections.
Additionally,
having
one
previous
infection
provides
some
protection
against
a
second
infection.

The
derivation
and
assumptions
for
the
inputs
for
the
dose­
response
model
parameters
(
d,
r,
n)
are
described
in
more
detail
in
other
sections.

Morbidity
Rate
The
above
elements
of
dose­
response
relate
to
the
prediction
of
an
infection
occurring
given
various
exposure
circumstances.
As
noted
at
the
outset
of
this
section,
the
hazard
identification
for
Cryptosporidium
includes
not
only
the
risk
of
infection,
but
also
the
risk
of
illness
resulting
from
an
infection.
Not
all
infections
will
result
in
illness
and
observable
symptoms.
The
probability
of
becoming
ill
given
an
infection
is
called
the
morbidity
rate.
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
11
Some
studies
indicate
that
the
morbidity
rates
increase
at
higher
doses
(
DuPont
et
al.
1995).
However,
for
this
risk
assessment,
the
morbidity
rate
is
independent
of
dose.
After
examining
the
potential
impact,
EPA
determined
that
a
higher
morbidity
at
higher
dose
effect
was
not
significantly
relevant
to
this
analysis.
The
fundamental
effect
being
quantified
is
the
endemic
rate
of
illnesses
and
deaths
from
persistent
(
and
low)
levels
of
Cryptosporidium,
not
the
higher
levels
that
might
occur
in
an
outbreak.
The
underlying
dose
data,
both
as
measured
and
modeled,
reflect
at
most
a
few
oocysts
per
day
for
individuals.
In
the
risk
assessment
model,
the
portion
of
the
risk
posed
by
the
small
portion
of
the
population
ingesting
even
an
expected
two
oocysts/
L
is
negligible;
the
portion
of
the
risk
posed
by
people
ingesting
three
or
more
oocysts/
L
is
virtually
zero.
Thus,
the
results
of
the
analysis
would
not
be
affected
by
using
increased
morbidity
rates
with
significantly
higher
doses.
(
Although
not
quantified,
the
risks
of
outbreaks
are
considered
and
are
discussed
in
section
5.4.1.)

To
develop
an
estimate
of
morbidity
rate,
EPA
analyzed
available
literature
and
identified
studies
with
applicable
data.
Some
of
the
preliminary
human
ingestion
trials
were
conducted
on
healthy
individuals
with
no
evidence
of
previous
C.
parvum
infection
(
DuPont
et
al.
1995).
Other
studies
challenged
individuals
with
existing
antibodies
or
re­
challenged
those
who
had
participated
in
earlier
studies.
DuPont
et
al.
(
1995)
found
that
39
percent
of
those
infected
had
clinical
cryptosporidiosis.
Haas
et
al.
(
1996)
provided
information
based
on
the
same
data
also
suggesting
a
morbidity
rate
of
39
percent,
but
also
computed
95
percent
confidence
limits
of
19
and
62
percent.
More
recently,
a
study
found
that
after
repeated
exposure
to
C.
parvum
(
IOWA
strain),
the
morbidity
rate
was
the
same
as
for
the
initial
exposure
in
re­
infected
subjects
(
Okhuysen
et
al.
1998).
Okhuysen
et
al.
(
1998)
also
found
that
58
percent
of
their
subjects
who
received
doses
of
Cryptosporidium
developed
diarrhea,
which
is
an
underestimate
of
morbidity
since
symptoms
other
than
diarrhea
contribute
to
the
morbidity
rate.
However,
these
subjects
were
given
doses
higher
than
those
projected
in
water
supplies.
Chappell
et
al.
(
1997)
observed
that
the
rate
of
diarrheal
illness
was
higher
for
the
TAMU
or
UCP
isolates
of
C.
parvum
than
for
the
IOWA
isolate
first
studied
by
DuPont
et
al.
(
1995)
and
Haas
et
al.
(
1996).

Given
these
results
and
the
morbidity
variability
associated
with
C.
parvum
during
reported
outbreaks,
the
actual
morbidity
rate
may
vary
with
the
type
of
strain
to
which
a
population
is
exposed,
as
well
as
with
the
immune
status
of
the
exposed
population.
However,
the
prevalence
of
strains
and
the
immune
status
of
the
population
are
unknown
and
therefore
not
quantified
for
this
risk
assessment.
The
uncertainty
around
the
value
for
morbidity,
though,
is
considered
in
the
risk
assessment.
The
quality
of
available
data
does
not
support
making
more
than
a
generalized
estimate
of
the
range
and
nature
of
uncertainty.
The
underlying
data
do
support
the
use
of
a
distribution
with
a
central
tendency
and
provide
information
to
establish
reasonable
ranges.
As
a
result,
morbidity
was
modeled
as
an
uncertain
variable
having
a
triangular
distribution.

Analysis
of
the
reviewed
research
resulted
in
a
mode
of
50
percent,
lower
bound
of
30
percent,
and
upper
bound
of
70
percent
for
the
triangular
uncertainty
distribution.
The
following
limitations
in
the
research
were
identified
and
considered
in
the
derivation
of
the
above
values:
the
Okhuysen
et
al.
(
1998)
results
based
on
diarrheal
rates
are
probably
an
underestimate;
Chappell
et
al.
(
1997)
found
that
diarrheal
rates
were
higher
for
isolates
other
than
IOWA;
and
the
general
population
likely
has
a
higher
morbidity
rate
than
the
healthy
individuals
used
in
the
study
groups.

The
central
tendency
(
mode)
for
the
distribution
used
in
the
risk
assessment
model
is
50
percent.
This
is
a
bit
below
the
Okhuysen
et
al.
(
1998)
results
(
58
percent),
but
above
the
values
estimated
by
Du
Pont
et
al.
(
1995)
and
Haas
et
al.
(
1996)
(
39
percent).
These
studies
used
the
IOWA
isolate,
and
a
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
12
simple
average
of
them
results
in
a
value
of
48.5
percent.
The
mode
was
rounded
up
to
50
percent
to
account
for
the
apparent
underestimation
of
these
studies,
as
noted
above.

The
upper
bound
for
the
distribution
used
in
the
risk
assessment
model
is
70
percent.
The
upper
bound
was
set
above
the
95
percent
confidence
limit
of
62
percent
estimated
by
Haas
et
al.
(
1996).
This
reflects
that
the
absolute
limit
of
the
triangular
distribution
would
reasonably
be
above
that
95
percent
confidence
limit
and
the
apparent
underestimation
of
these
studies,
as
noted
above.
The
difference
in
the
upper
bound
(
70
percent)
and
the
Haas
et
al.
95
percent
confidence
limit
(
62
percent)
represents
only
3
percent
of
the
triangular
distribution,
indicating
that
the
upper
tail
of
the
triangular
distribution
is
comparable
to
upper
portion
of
Haas's
distribution.

The
lower
bound
for
the
distribution
used
in
the
risk
assessment
model
is
30
percent.
The
lower
bound
was
set
higher
than
the
19
percent
estimated
by
Haas
et
al.
(
1996).
While
this
bound
does
not
encompass
the
lower
95
percent
confidence
level
in
the
distribution
used
in
the
risk
assessment,
it
does
account
for
the
apparent
underestimation
in
the
studies.

Mortality
Rate
The
third
dose­
response
relationship
used
in
this
analysis
is
the
probability
of
fatality
given
that
an
illness
has
occurred.
There
are
no
general
data
on
the
rate
of
mortality
from
cryptosporidiosis.
To
derive
mortality
estimates,
data
from
the
Milwaukee
outbreak
are
used
and
adjusted
to
reflect
changes
in
rates
of
illnesses
and
advanced
treatments
that
have
lessened
mortality
among
persons
living
with
AIDS.
Further
adjustments
are
used
to
reflect
the
differences
between
the
populations
of
those
living
in
areas
served
by
filtered
and
unfiltered
systems.
Since
there
is
uncertainty
around
the
mortality
rate
used
in
the
dose­
response
model,
EPA
conducted
a
sensitivity
analysis
that
varied
the
AIDS
mortality
rate
by
+/­
50
percent.
This
analysis
and
its
results
are
described
in
Appendix
R.

The
starting
point
is
the
mortality
rates
associated
with
the
Milwaukee
Cryptosporidium
outbreak.
In
that
outbreak,
54
people
died
who
had
cryptosporidiosis
listed
on
their
death
certificate.
Of
those,
46
also
had
AIDS
listed
as
an
underlying
cause
of
death
(
Hoxie
et
al.
1997).
The
Milwaukee
outbreak
had
an
estimated
403,000
cases
of
illness
(
Kramer
et
al.
1996b).
The
unadjusted
mortality
rate
for
AIDS­
related
deaths
is
thus
46
deaths/
403,000
illnesses,
or
11.41
deaths/
100,000
illnesses.
The
rate
for
the
other,
non­
AIDS­
related
deaths
is
thus
8
deaths/
403,000
illnesses,
or
1.98
deaths/
100,000
illnesses.
(
All
calculations
in
this
section
are
rounded
for
ease
of
presentation,
but
unrounded
data
are
used
in
the
analysis.)

There
were
no
further
adjustments
made
to
the
non­
AIDS
mortality
rate.
A
review
of
available
statistics
showed
that
data
to
compare
the
incidence
of
the
other
underlying
illnesses
(
coccidiosis
(
presumably
cryptosporidiosis),
viral
hepatitis,
brain
tumor,
heart
failure,
and
alcoholic
cirrhosis
of
the
liver)
between
Milwaukee
in
1993
and
the
nation
in
1999
or
2000
were
generally
unavailable.
Even
comparison
of
proxy
data
(
death
rates
rather
than
incidence)
proved
of
little
value.
Data
for
Milwaukee
were,
in
general,
inconclusive;
too
few
cases
were
reported
to
make
statistics
meaningful.
Only
in
the
case
of
alcoholic
cirrhosis
of
the
liver
were
data
statistically
significant,
and
in
that
case,
the
rate
of
deaths
per
100,000
population
were
comparable
between
Milwaukee
(
3.36)
and
the
nation
as
a
whole
(
3.03)
(
CDC
2001a;
Hoxie
et
al
1997;
CDC
1995).

The
Milwaukee
AIDS­
related
mortality
rate
was
adjusted
to
account
for
the
decrease
in
the
mortality
rate
among
people
with
AIDS
from
the
time
of
the
Milwaukee
incident
to
1999
(
the
most
recent
3
Data
on
the
AIDS
population
and
on
the
population
served
by
the
water
system
for
the
area
directly
affected
by
the
Milwaukee
Cryptosporidium
outbreak
are
inconsistent
in
the
sources
used
and
the
area
covered,
and
individual
estimates
varied.
Data
for
the
entire
State
of
Wisconsin
are
used
as
the
best
consistent
source
of
AIDS
data
and
population
data.
The
State­
level
data
are
from
U.
S.
Census
and
CDC
sources.
These
data
are
comparable
to
other
data
used
in
this
analysis.
See
Appendix
C
for
more
details.

Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
13
year
with
comparable
data),
and
the
difference
in
the
Milwaukee
AIDS
population
in
1993
to
the
national
AIDS
population
in
1999.
These
adjustments
are
described
below;
the
adjusted
calculation
is:

Deaths/
100,000
illnesses
in
the
Milwaukee
outbreak
(
11.41)
×
factor
to
adjust
for
lessened
mortality
among
persons
with
AIDS
×
factor
to
adjust
for
AIDS
populations
=
AIDS­
related
deaths
per
100,000
cryptosporidiosis
illnesses.

The
mortality
rate
for
AIDS
has
declined
greatly
since
1993
due
to
the
use
of
combination
retro
viral
therapies
and
other
factors.
The
AIDS
mortality
rate
(
deaths
per
100,000
AIDS
population)
in
1999
was
3,187
(
10,122
deaths
in
a
population
of
317,652)
(
CDC
2000;
CDC
2001b).
In
1993,
this
rate
was
25,963
(
45,271
deaths
in
a
population
of
174,369)
(
CDC
2000).
The
ratio
of
these
rates
is
12.3
percent,
that
is,
the
rate
of
deaths
among
AIDS
patients
for
all
reasons
in
1999
was
only
12.3
percent
of
what
it
was
in
1993.

The
second
adjustment
accounts
for
difference
in
the
percent
of
the
national
population
that
was
living
with
AIDS
in
1999
and
the
percent
of
the
Milwaukee
population
that
was
living
with
AIDS
in
1993.
This
adjustment
is
calculated
separately
for
areas
served
by
unfiltered
systems
and
filtered
systems.
As
an
approximation
of
the
value
for
populations
served
by
unfiltered
systems,
the
percentage
of
the
population
living
with
AIDS,
which
is
0.179
percent
(
56,266
in
a
population
of
31,488,807)
was
used.
As
an
approximation
of
the
percentage
in
areas
served
by
filtered
systems,
national
estimates
were
used,
less
what
had
been
accounted
for
by
unfiltered
systems.
(
See
Appendix
C
for
details
on
these
data
and
calculations.)
The
rate
for
filtered
systems
is
0.105
percent
(
based
on
an
AIDS
population
of
261,386
in
a
population
base
of
249,973,099)
(
CDC
2001b;
US
Census
2000).

The
percentages
of
people
living
with
AIDS
in
1999
served
by
filtered
and
unfiltered
systems
are
used
separately
to
adjust
and
update
the
1993
incidence
rate
of
AIDS
in
Wisconsin.
The
data
on
AIDS
incidence
and
population
should
represent
the
same
location;
however,
the
areas
for
which
data
are
available
do
not
match
the
exact
geography
of
the
areas
served.
3
The
ratios
that
come
from
this
approach
are
still
useful
as
approximations,
and
their
use
is
an
improvement
over
not
including
adjustments
for
these
factors.
In
Wisconsin
in
1993,
the
percentage
of
the
population
that
had
AIDS
was
0.017
(
862
persons
with
AIDS
in
a
population
of
5,044,318).
Extrapolating
the
Wisconsin
data
to
all
populations
served
by
unfiltered
and
filtered
systems,
gives
a
factor
of
10.46
for
unfiltered
systems
(
0.179
percent/
0.017
percent)
and
a
factor
of
6.12
(
0.105
percent/
0.017
percent)
for
filtered
systems.
That
is,
the
incidence
of
people
living
with
AIDS
in
1999
in
areas
served
by
unfiltered
systems
is
10.46
times
the
incidence
in
Wisconsin
in
1993.
Similarly,
there
are
6.12
times
as
many
people
living
with
AIDS
in
1999
and
served
by
filtered
systems
as
there
were
in
Wisconsin
in
1993.

Using
the
Milwaukee
AIDS­
related
mortality
rate
and
the
adjustment
factors
described
above,
the
final
mortality
rate
for
unfiltered
systems
(
expressed
as
deaths
per
100,000
cryptosporidiosis
illnesses)
is
14.65
(
11.41
×
12.3%
×
10.46).
Similarly,
for
filtered
systems,
the
AIDS­
related
mortality
rate
is
8.57
deaths
per
100,000
cryptosporidiosis
illnesses.
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
14
The
risk
assessment
model
uses
a
combination
of
AIDS­
related
and
non­
AIDS­
related
mortality.
Thus,
adding
together
these
rates
yields
an
overall
mortality
rate
for
unfiltered
systems
of
16.6339
deaths
per
100,000
cryptosporidiosis
illnesses
(
14.65
AIDS
+
1.98
non­
AIDS).
For
filtered
systems,
this
figure
is
10.5575
deaths
per
100,000
cryptosporidiosis
illnesses
(
8.57
AIDS
+
1.98
non­
AIDS).
These
mortality
factors
are
constants
in
the
model
(
that
is,
no
uncertainty
is
attributed
to
these
parameters).

The
mortality
rate
from
the
Milwaukee
outbreak
may
not
reflect
the
overall
mortality
rates
from
low­
level
endemic
exposure.
The
estimated
levels
of
Cryptosporidium
in
the
finished
water
supplies
during
the
Milwaukee
outbreak
were
much
higher
than
the
levels
expected
in
systems
complying
with
the
SWTR,
IESWTR,
and
LT1ESWTR.
Thus,
the
higher
level
of
Cryptosporidium
in
the
water
supply
could
have
resulted
in
a
higher
mortality
rate
than
that
expected
from
endemic
exposure
if
responses
increased
more
than
proportionately
at
higher
dose
levels.

No
data
are
yet
available,
however,
to
support
this
hypothesis;
data
are
available
to
indicate
only
a
higher
probability
of
infection
resulting
from
higher
ingested
doses.
In
an
outbreak
in
Las
Vegas,
similar
mortality
rates
were
observed
in
AIDS
patients
(
52.6
percent
among
AIDS
patients
in
Las
Vegas
compared
with
68
percent
among
AIDS
patients
in
Milwaukee).
These
similar
rates
were
observed
despite
the
hypothesis
that
the
drinking
water
had
been
contaminated
over
an
extended
period
of
time
with
intermittent
low
levels
of
oocysts,
unlike
Milwaukee's
massive
contamination
(
Rose
1997).
A
recent
study
by
Hunter
et
al.
(
2001),
suggests
that
the
level
of
endemic
diarrhea
from
all
sources
was
underestimated
in
the
Milwaukee
incident,
leading
to
an
overestimation
of
the
number
of
diarrheal
illnesses
due
to
cryptosporidiosis.
A
lower
estimate
of
illness
would
consequently
raise
the
mortality
rate
per
case
of
illness
by
holding
deaths
constant
as
illnesses
decreased.
However,
there
is
currently
no
consensus
on
whether
to
accept
the
Hunter
et
al.
conclusions,
and
responses
to
their
analysis
are
being
prepared
by
other
investigators
of
the
Milwaukee
outbreak.
The
model,
therefore,
uses
the
Hoxie
et
al.
1997
illness
estimates
(
cited
previously)
for
the
Milwaukee
outbreak.

5.2.4
Exposure
Assessment
This
section
discusses
the
three
elements
needed
for
characterizing
human
exposure
to
infectious
Cryptosporidium
oocysts
in
drinking
water.

°
The
distribution
of
total
and
infectious
Cryptosporidium
in
finished
water,
reflecting
source
water
levels
and
treatment
effectiveness
(
section
5.2.4.1)

°
The
distribution
of
individual
daily
drinking
water
consumption
and
number
of
days
of
exposure
(
section
5.2.4.2)

°
The
estimated
population
served
by
systems
potentially
affected
by
the
LT2ESWTR
(
section
5.2.4.3)

5.2.4.1
Distribution
of
Infectious
Cryptosporidium
in
Finished
Water
The
distribution
of
infectious
Cryptosporidium
in
finished
water
to
which
the
affected
population
is
exposed
reflects
three
factors:

°
The
distribution
of
total
Cryptosporidium
concentrations
in
source
water
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
15
°
The
fraction
of
those
oocysts
that
are
considered
to
be
infectious
°
The
removal
and
inactivation
rates
of
the
infectious
Cryptosporidium
predicted
for
Pre­
LT2ESWTR
and
predicted
Post­
LT2ESWTR
treatment
conditions
The
National
Distribution
of
Cryptosporidium
Concentrations
in
Source
Water
Simulated
source
water
Cyrptosporidium
concentrations
are
drawn
from
the
occurrence
distributions
described
in
Chapter
4.
Section
4.2.2
gives
a
general
overview
of
the
occurrence
modeling
approach,
and
sections
4.4.3
and
4.5.3
provide
more
detail
on
the
unfiltered
and
filtered
occurrence
models,
respectively.

At
the
broadest
level,
there
are
four
basic
occurrence
models
that
provide
input
to
this
EA:
three
based
on
filtered­
system
data
from
the
ICR,
ICRSSL
and
ICRSSM,
and
one
based
on
unfiltered­
system
data
from
the
ICR.
The
separate
unfiltered
system
model
is
motivated
by
fundamental
differences
between
the
quality
of
source
water
in
filtered
and
unfiltered
systems.
Differences
among
the
three
filtered
systems
data
sets
arise
from
different
survey
sampling
plans,
lab
methods,
and
sampling
periods
(
see
section
4.2.1
for
a
detailed
description
of
these
survey
differences).

National
benefit
estimates
are
derived
from
each
of
the
three
filtered­
systems
data
sets
and
compared
with
one
another,
but
they
are
not
combined
or
weighted
in
any
way.
Each
of
the
data
sets
has
strengths,
but
none
was
judged
as
superior
in
estimating
national
levels
of
filtered
systems
occurrence.
The
fact
that
there
are
three
different
data
sets
for
filtered
systems,
leading
to
three
distinct
occurrence
distributions,
reflects
significant
uncertainty
about
the
true
national
Cryptosporidium
distribution
and
its
stability
over
time.
Using
all
three
models
serves
two
purposes:
it
captures
this
uncertainty
and
portrays
it
clearly,
while
at
the
same
time
providing
three
distinct,
independently
drawn,
plausible
representations
of
the
true
national
distribution.

As
discussed
in
section
4.2.2,
rather
than
fitting
a
single
log­
normal
occurrence
distribution
to
each
of
the
four
data
sets
described
above,
the
Cryptosporidium
modeling
approach
generates
a
collection
of
log­
normal
distributions
from
each
data
set.
So,
for
example,
1,000
mean
and
standard
deviation
pairs
are
drawn
from
the
ICR
filtered­
systems
model
to
serve
as
input
to
the
risk
assessment
model.
Each
pair
defines
a
single
log­
normal
distribution
that
could
have
plausibly
generated
the
ICR
survey
results
for
filtered
systems.
The
other
three
models,
corresponding
to
the
other
three
data
sets,
are
sampled
in
the
same
way.
The
result
is
the
four
collections
of
log­
normal
distributions
that
are
summarized
in
Exhibits
4.6,
and
4.11
through
4.13.

These
collections
of
occurrence
distributions
serve
as
inputs
to
the
Monte
Carlo
risk
simulation
(
part
of
the
risk
assessment
model)
described
in
section
5.2.5.
In
this
risk
simulation,
an
outer
loop
captures
uncertainty
about
risk
parameters,
and
an
inner
loop
models
variability
in
risk
from
water
system
to
water
system.
For
each
of
these
uncertainty
loops,
a
single
occurrence
distribution
is
drawn
from
a
given
collection
of
1,000
distributions.
Then,
within
the
variability
loop,
the
selected
log­
normal
distribution
is
used
to
simulate
variability
in
occurrence
 
both
system­
to­
system
differences
in
average
Cryptosporidium
concentration
and
sample­
to­
sample
differences
over
time.
This
process
is
repeated
until
250
uncertainty
loops
have
been
completed,
yielding
250
national
risk
curves.

Again,
the
overall
risk
model
is
described
in
more
detail
in
section
5.2.5.
The
key
point
here
is
that
the
occurrence
inputs
to
this
risk
model
are
carefully
structured
to
separate
uncertainty
about
the
true
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
16
national
distribution,
on
the
one
hand,
from
the
estimated
system­
to­
system
variability
in
Cryptosporidium
concentration
on
the
other.
This
approach
provides
a
good
match
between
the
occurrence
inputs
and
the
general
structure
of
the
broader
risk
model.

Infectious
Oocyst
Fraction
An
important
parameter
when
assessing
the
risk
associated
with
a
given
concentration
of
Cryptosporidium
in
a
drinking
water
source
is
the
percentage
of
oocysts
that
are
infectious.
The
methods
used
to
analyze
Cryptosporidium
in
the
ICR
and
ICRSSs
measured
total
oocyst
counts
without
regard
to
how
many
were
actually
infectious.
Because
oocysts
degrade
in
the
environment,
it
is
expected
that
only
a
fraction
of
the
oocysts
counted
in
these
surveys
would
be
capable
of
causing
infection
in
a
susceptible
host.
Consequently,
the
distributions
of
total
Cryptosporidium
occurrence
based
on
these
surveys
are
believed
to
overestimate
the
concentration
of
infectious
oocysts.

Further,
the
parameter
actually
of
concern
is
the
ratio
of
the
infectious
oocyst
fraction
in
the
environment
to
the
same
fraction
in
dose­
response
studies.
Even
in
the
best
controlled
laboratory
studies,
the
fraction
of
infectious
oocysts
is
less
than
100
percent
(
but
was
above
80
percent
in
the
studies
considered
here).

There
is
no
direct
way
to
assess
the
infectivity
of
oocysts
counted
with
the
ICR
Method
in
the
ICR
or
with
Methods
1622/
23
in
the
ICRSSs.
Rather,
related
information
is
gleaned
from
two
sources:
the
physical
structure
of
observed
oocysts
and
a
comparison
study
where
samples
were
analyzed
with
both
Method
1623
and
a
cell
culture
test
for
oocyst
infectivity.
From
these
two
sources,
an
estimate
was
made
of
the
most
likely
proportion
of
counted
oocysts
(
in
the
environment)
that
were
infectious
(
at
least
in
a
laboratory
setting).

As
discussed
in
section
4.5.3,
Cryptosporidium
oocysts
counted
with
the
ICR
Method
or
Methods
1622/
23
are
characterized
in
one
of
three
ways:
(
1)
those
with
internal
structures,
i.
e.,
those
having
recognizable
structure
consistent
with
Cryptosporidium;
(
2)
oocysts
with
amorphous
structures,
which
indicates
that
material
is
present
in
the
oocyst,
but
it
cannot
be
confirmed
as
Cryptosporidium;
or
(
3)
empty
oocysts.
Assignment
of
these
labels
is
dependent
upon
analyst
judgment
and
none
is
a
certain
indicator
of
whether
an
oocyst
is
truly
infectious.
Oocysts
with
internal
structures
are
generally
considered
to
have
the
highest
likelihood
of
being
infectious,
though
laboratory
studies
have
shown
oocysts
can
lose
infectivity
without
loss
of
internal
structures.
Oocysts
with
amorphous
structures
may
be
still
infectious
or,
alternatively,
may
be
some
other
microorganism
that
mimics
the
structure
and
properties
of
Cryptosporidium.
Oocysts
that
are
empty
of
internal
structures
are
assumed
to
be
non­
infectious
(
LeChevallier
et
al.
1997a).

In
the
ICR
data
set,
laboratories
characterized
23
percent
of
the
oocysts
counted
as
having
internal
structures,
39
percent
having
amorphous
structures,
and
38
percent
as
being
empty.
With
the
ICRSSs,
37
percent
of
the
oocysts
had
internal
structures,
47
percent
had
amorphous
structures,
and
16
percent
were
empty.
If
it
were
assumed
that
the
empty
oocysts
could
not
be
infectious,
then
these
data
suggest
that
the
percentage
of
counted
oocysts
that
were
infectious
were
at
most
62
percent
in
the
ICR
and
84
percent
in
the
ICRSS.

The
lower
percentage
of
empty
oocysts
in
the
ICRSS,
versus
the
ICR,
may
be
attributable
to
the
improved
sample
purification
technique
in
Methods
1622/
23.
This
technique,
immunomagnetic
separation,
prevents
many
non­
Cryptosporidium
particles
from
being
transferred
to
the
slide
for
examination;
some
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
17
of
these
non­
Cryptosporidium
particles
may
have
been
incorrectly
counted
as
empty
oocysts
in
the
ICR
(
Connell
et
al.
2000).
Moreover,
the
LT2ESWTR
would
require
use
of
Methods
1622/
23
for
assigning
systems
to
bins,
so
the
ICRSS
data
may
be
more
reflective
of
data
that
would
be
generated
under
this
rule.

A
study
by
LeChevallier
et
al.
(
2003)
provides
another
indication
of
the
percentage
of
oocysts
counted
by
Method
1623
that
are
infectious.
This
study
involved
intensive
sampling
of
six
source
waters
for
Cryptosporidium
and
other
microbiological
and
water
quality
parameters.
Each
Cryptosporidium
sample
was
analyzed
by
both
Method
1623
and
a
method
that
used
cell
culture
and
polymerase
chain
reaction
(
CC­
PCR)
to
measure
viability
and
infectivity.
Cryptosporidium
oocysts
were
detected
in
60
of
593
samples
(
10.1
percent)
by
Method
1623
and
infectious
oocysts
were
detected
in
22
of
560
samples
(
3.9
percent)
by
the
CC­
PCR
procedure.
Recovery
efficiencies
for
the
two
methods
were
similar.
According
to
the
authors,
these
results
suggest
that
approximately
37
percent
of
the
Cryptosporidium
oocysts
detected
by
Method
1623
were
viable
and
infectious.
Only
one
sample
was
positive
by
both
Method
1623
and
CC­
PCR,
though
this
result
is
consistent
statistically
with
the
low
oocyst
concentration.

When
using
the
data
sets
derived
from
the
ICRSS,
EPA
characterized
the
percent
of
oocysts
that
are
infectious
as
an
uncertain
variable
with
a
triangular
distribution
having
a
lower
bound
of
30
percent,
a
mode
of
40
percent,
and
an
upper
bound
of
50
percent.
The
mode
is
consistent
with
results
from
the
LeChevallier
et
al.
study
(
2003)
where
the
number
of
samples
with
infectious
oocysts
was
37
percent
of
the
number
with
oocysts
counted
using
EPA
Method
1623.
It
is
also
consistent
with
the
37
percent
of
oocysts
counted
during
the
ICRSS
that
had
internal
structures,
which
are
considered
the
most
likely
to
be
infectious.
The
bounds
were
set
at
+/­
25
percent
of
the
mode,
balancing
the
good
quality
of
the
LeChevallier
et
al.
(
2003)
data
with
the
uncertainty
in
applying
this
result
on
a
national
basis.
This
distribution
also
recognizes
that
an
unknown
fraction
of
the
47
percent
of
oocysts
counted
in
the
ICRSS
with
amorphous
structures
were
infectious,
which
could
lead
to
a
total
fraction
of
infectious
oocysts
greater
than
40
percent;
alternatively,
a
fraction
of
the
oocysts
with
internal
structures
were
likely
not
infectious,
which
could
lead
to
a
total
fraction
of
infectious
oocysts
less
than
40
percent.

When
using
the
data
sets
derived
from
the
ICR,
EPA
characterized
the
percent
of
oocysts
that
are
infectious
as
an
uncertain
variable
with
a
triangular
distribution
having
a
lower
bound
of
15
percent,
a
mode
of
20
percent,
and
an
upper
bound
of
25
percent.
The
lower
range
for
the
ICR
distribution
reflects
the
higher
rate
of
empty
oocysts,
which
are
considered
to
be
non­
infectious,
detected
by
the
ICR
Method
(
38
percent
in
the
ICR
vs.
16
percent
in
the
ICRSS),
and
the
lower
rate
of
oocysts
with
internal
structures
(
17
percent
in
the
ICR
vs.
39
percent
in
the
ICRSS)
data.

Pre­
LT2ESWTR
Removal/
Inactivation
of
Cryptosporidium
Filtration
is
currently
the
primary
treatment
mechanism
used
in
PWSs
to
remove
Cryptosporidium
Finished
water
Cryptosporidium
concentrations
developed
for
this
risk
characterization
reflect
predicted
filtration
improvements
to
meet
IESWTR
and
LT1ESWTR
requirements.
Chapter
4,
section
4.5.4
presents
the
methodology
used
to
estimate
Pre­
LT2ESWTR
(
Post­
IESWTR
and
Post­
LT1ESWTR)
removal
levels.
To
summarize,
Pre­
LT2ESWTR
removal
is
modeled
as
triangular
distributions
as
follows:

°
For
small
systems
(
serving
fewer
than
10,000
people):
2
to
4
log
range
of
Cryptosporidium
removal
to
capture
system­
to­
system
variability;
possible
modes
of
the
triangular
distributions
between
2.25
and
3.25
to
capture
uncertainty
in
the
"
true"
distribution
of
removal.
4
Percent
of
plants
shown
in
Exhibit
4.11,
column
C.

5
Percent
of
plants
shown
in
Exhibit
4.11,
column
F.

Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
18
°
For
large
systems
(
those
serving
at
least
10,000
people):
2
to
5
log
range
of
removal
to
capture
system­
to­
system
variability;
possible
modes
of
the
triangular
distributions
between
2.5
and
3.5
to
capture
uncertainty
in
defining
the
"
true"
distribution
of
removal.

These
distributions
are
intended
to
bound
the
uncertainty
in
the
most
likely
removal
values,
as
defined
by
the
variable
modes
of
the
triangular
distributions.
Uncertainty
and
variability
in
removal
are
modeled
independently
of
source
water
Cryptosporidium
concentration.
Also,
removal
is
modeled
independently
of
filtration
type
(
e.
g.,
conventional
or
direct).
This
is
primarily
because
the
analysis
assumes
that
few
systems
designed
their
plants
to
account
for
Cryptosporidium
concentrations.
In
addition,
reviews
of
the
ICR
data
do
not
reveal
treatment
designs
to
be
correlated
to
Cryptosporidium
levels.
Further,
few
medium
and
small
systems
have
collected
data
that
they
could
have
used
for
this
purpose.

As
noted
in
Chapter
4,
a
small
fraction
of
filtered
plants4
are
predicted
to
have
added
advanced
technologies
that
provide
5.5
logs
of
removal
or
inactivation
of
Cryptosporidium
before
the
implementation
of
LT2ESWTR.
Although
several
technologies
are
capable
of
this
level
of
performance,
the
model
only
specifically
takes
account
of
those
using
microfiltration/
ultrafiltration
(
MF/
UF).
Those
plants
that
had
MF/
UF
in
place
are
removed
from
the
baseline
for
filtered
plants
because
they
are
exempt
from
additional
monitoring
and
treatment
requirements
under
the
LT2ESWTR.
Therefore,
no
adjustments
are
made
in
the
risk
model
to
account
for
these
plants.
A
small
number
of
plants5
are
predicted
to
install
these
technologies
to
comply
with
the
Stage
1
or
Stage
2
Disinfection
Byproduct
Rule
(
Stage
1
DBPR
or
Stage
2
DBPR).
These
plants
are
also
excluded
from
the
model
for
estimating
risks,
although
first­
round
monitoring
costs
are
included.
Benefits
related
to
Cryptosporidium
reduction
for
these
plants
are
attributed
to
the
other
rules
and
thus
not
captured
in
this
EA.

Pre­
LT2ESWTR
finished
water
Cryptosporidium
concentrations
(
mean
and
median
values)
are
summarized
in
Chapter
4,
Exhibit
4.19a
and
b.
Values
are
slightly
lower
for
medium
and
large
systems
because
they
are
predicted
to
have
better
removal
performance
following
the
implementation
of
the
IESWTR
and
LT1ESWTR.

Post­
LT2ESWTR
Removal/
Inactivation
of
Cryptosporidium
The
additional
Cryptosporidium
reduction
gained
through
the
addition
of
advanced
technologies
for
the
LT2ESWTR
is
estimated
through
a
four­
step
process:

STEP
1
 
Predict
source
water
occurrence
for
each
plant
At
the
beginning
of
each
uncertainty
loop,
the
model
defines
an
occurrence
distribution
by
randomly
selecting
a
log­
normal
mean
and
standard
deviation.
At
each
iteration
(
at
the
variability
level),
the
model
then
randomly
draws
annual
plant
means
from
the
distribution
defined
in
the
associated
uncertainty
step,
to
simulate
plant­
to­
plant
variability
in
occurrence.
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
19
STEP
2
 
Predict
bin
classification
for
each
plant
Section
4.5.6
and
Appendix
B
summarize
the
predicted
bin
classification
for
the
LT2ESWTR
Preferred
Regulatory
Alternative,
as
well
as
the
two
alternative
bin
classifications
considered
in
this
EA
(
regulatory
alternatives
A2
and
A4).
In
general,
the
risk
model
uses
a
probability
function
that
takes
a
"
true"
source
water
concentration
and
adjusts
for
test
method
recovery
to
classify
a
plant
into
a
treatment
bin.
Predicted
binning
has
substantial
impacts
on
estimated
costs
and
benefits
of
the
LT2ESWTR;
therefore,
a
sensitivity
analysis
was
performed
to
evaluate
effects
of
predicted
bin
assignment
based
on
alternative
source
water
occurrence
distributions
(
see
Appendix
B).

STEP
3
 
Adjust
bin
classification
for
plants
with
treatment
credits
prior
to
the
LT2ESWTR
Some
plants
may
have
advanced
treatment
in
place
following
the
IESWTR
and
LT1ESWTR
Because
these
plants
are
meeting
compliance
requirements
prior
to
the
LT2ESWTR,
benefits
and
costs
associated
with
existing
toolbox
options
are
not
attributable
to
this
rule.
The
risk
model
takes
into
account
the
higher
level
of
treatment
achieved
by
these
systems
for
both
pre­
LT2ESWTR
finished
water
occurrence
and
determining
the
required
log
credit.
EPA
estimated
the
percentage
of
plants
that
will
achieve
the
treatment
requirements
for
combined
filter
performance,
presedimentation
basin,
and
secondary
filtration
toolbox
options,
as
described
in
Appendix
A.
Each
of
these
toolbox
options
provides
0.5
log
treatment
credit.
Additionally,
a
few
of
these
plants
may
have
more
than
one
of
these
options
in
place
and
thus
receive
1.0
log
credit.
Exhibit
5.4
shows
the
percent
of
plants
estimated
to
receive
0.5
or
1.0
log
credit.

Exhibit
5.4
Percent
of
Plants
With
Pre­
LT2ESWTR
Treatment
Credit
System
Size
(
population
served)
Percent
of
Plants
Achieving
Very
Small
and
Small
(
10,000
and
fewer)
37%

Medium
(
10,001­
100,000)
55%

Large
(
more
than
100,000)
58%

Source:
Appendix
A,
Exhibit
A.
7.

STEP
4
 
Determine
actual
log
reduction
achieved
Systems
have
various
treatment
options
available
to
meet
Cryptosporidium
reduction
requirements
for
a
given
bin.
Some
technologies
(
such
as
bag
and
cartridge
filters)
are
projected
to
be
used
only
by
small
systems,
and
some
technologies
can
only
be
used
by
larger
systems.
For
example,
chlorine
dioxide,
ozone,
and
secondary
filters
are
judged
to
be
impractical
for
systems
serving
fewer
than
500
people
(
Exhibit
6.9).
Once
those
constraints
are
accounted
for,
the
selection
of
technologies
(
detailed
in
Chapter
6
and
Appendix
F)
is
performed
using
a
"
least­
cost"
approach,
whereby
EPA
estimates
(
for
the
purpose
of
estimating
the
national
costs
attributable
to
this
rule)
that
systems
will,
for
the
most
part,
select
the
least
costly
technology
available
to
meet
treatment
requirements
for
that
bin.
In
many
cases,
the
least
costly
technology
results
in
higher
levels
of
Cryptosporidium
inactivation
or
removal
than
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
20
required
for
that
bin
(
this
is
always
the
case
when
UV
is
selected).
Therefore,
this
risk
analysis
incorporates
estimates
of
"
actual"
reduction
achieved
beyond
bin
requirements.

Exhibit
5.5a
and
5.5b
present
the
predicted
actual
Cryptosporidium
reduction
achieved
for
systems
in
five
action
levels
for
systems
without
and
with
Pre­
LT2ESWTR
credit,
respectively.
Different
reduction
estimates
are
shown
for
very
small,
small,
medium,
and
large
systems
because
the
least­
cost
decision
tree
changes
as
system
size
changes,
primarily
because
of
different
economies
of
scale
among
treatment
technologies
(
that
is,
as
system
size
increases,
some
technologies
will
become
less
costly
than
others).
Also,
different
technologies
are
available
for
different
size
systems,
usually
because
of
practical
limitations
such
as
very
small
systems
not
having
24­
hour
per
day
staffing.
The
technology
selections
for
each
bin
are
developed
independent
of
regulatory
alternative
and
source
water
occurrence
distribution.
For
example,
for
any
regulatory
alternative
that
requires
2.0
log
reduction,
it
is
estimated
that
large
systems
will
actually
select
technologies
that
achieve
a
3.0
log
reduction
90
percent
of
the
time
(
the
remaining
10
percent
achieve
2.0
and
2.5
log
reduction),
regardless
of
source
water
occurrence
distribution.
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
21
Targeted
Log
Reduction
[
2]

0.5
Log
1
Log
1.5
Log
2
Log
2.5
Log
0.5
0%
0%
0%
0%
0%
1
90%
90%
0%
0%
0%
1.5
0%
0%
0%
0%
0%
2
0%
0%
90%
90%
0%
2.5
0%
0%
0%
0%
0%
3
10%
10%
10%
10%
100%
Total
100%
100%
100%
100%
100%
0.5
1%
0%
0%
0%
0%
1
90%
91%
0%
0%
0%
1.5
0%
0%
0%
0%
0%
2
0%
0%
10%
10%
0%
2.5
0%
0%
0%
0%
10%
3
9%
9%
90%
90%
90%
Total
100%
100%
100%
100%
100%
0.5
58%
0%
0%
0%
0%
1
1%
10%
0%
0%
0%
1.5
0%
0%
6%
0%
0%
2
0%
0%
3%
7%
0%
2.5
0%
1%
1%
3%
10%
3
42%
90%
90%
90%
90%
Total
100%
100%
100%
100%
100%
0.5
56%
0%
0%
0%
0%
1
1%
17%
0%
0%
0%
1.5
0%
0%
6%
0%
0%
2
0%
0%
3%
7%
0%
2.5
0%
1%
1%
3%
10%
3
43%
82%
90%
90%
90%
Total
100%
100%
100%
100%
100%
Actual
Log
Reduction
Achieved
Very
Small
Systems
(
Serving
<
500)

Small
Systems
(
Serving
501
to
10,000
people)

Medium
Systems
(
Serving
10,001
to
100,000
people)

Large
Systems
(
Serving
more
than
100,000
people)
Exhibit
5.5a
Actual
Log
Removal
Achieved
for
Systems
without
Credits[
1]

Notes:
[
1]
Percent
of
total
number
of
systems
assumed
to
achieve
actual
log
reduction
levels
to
meet
specific
treatment
bin
requirements.
[
2]
Log
reduction
requirements
associated
with
treatment
bins
for
all
regulatory
alternatives.
Source:
Appendix
F,
Exhibits
F.
3
through
F.
6,
"
Actual
Log
Credit"
and
"
Percent
of
Plants
Selecting
Technology
by
Bin"
columns.
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
22
Targeted
Log
Reduction
[
2]

0.5
Log
1
Log
1.5
Log
2
Log
2.5
Log
0.5
0%
0%
0%
0%
0%
1
90%
90%
0%
0%
0%
1.5
0%
0%
0%
0%
0%
2
0%
0%
90%
90%
0%
2.5
0%
0%
0%
0%
0%
3
10%
10%
10%
10%
100%
Total
100%
100%
100%
100%
100%
0.5
1%
0%
0%
0%
0%
1
90%
91%
0%
0%
0%
1.5
0%
0%
0%
0%
0%
2
0%
0%
10%
10%
0%
2.5
0%
0%
0%
0%
10%
3
9%
9%
90%
90%
90%
Total
100%
100%
100%
100%
100%
0.5
32%
0%
0%
0%
0%
1
1%
9%
0%
0%
0%
1.5
0%
0%
3%
0%
0%
2
0%
0%
5%
7%
0%
2.5
1%
1%
2%
3%
10%
3
67%
90%
90%
90%
90%
Total
100%
100%
100%
100%
100%
0.5
32%
0%
0%
0%
0%
1
1%
10%
0%
0%
0%
1.5
0%
0%
3%
0%
0%
2
0%
0%
5%
7%
0%
2.5
1%
1%
2%
3%
10%
3
67%
89%
90%
90%
90%
Total
100%
100%
100%
100%
100%
Medium
Systems
(
Serving
10,001
to
100,000
people)

Large
Systems
(
Serving
more
than
100,000
people)
Actual
Log
Reduction
Achieved
Very
Small
Systems
(
Serving
<
500)

Small
Systems
(
Serving
501
to
10,000
people)
Exhibit
5.5b
Actual
Log
Removal
Achieved
for
Systems
with
Credits[
1]

Notes:
[
1]
Percent
of
total
number
of
systems
assumed
to
achieve
actual
log
reduction
levels
to
meet
specific
treatment
bin
requirements.
[
2]
Log
reduction
requirements
associated
with
treatment
bins
for
all
regulatory
alternatives.
Source:
Appendix
F,
Exhibits
F.
7
through
F.
10,
"
Actual
Log
Credit"
and
"
Percent
of
Plants
Selecting
Technology
by
Bin"
columns.

5.2.4.2
Distribution
of
Individual
Daily
Drinking
Water
Consumption
The
second
element
of
the
exposure
assessment
is
the
characterization
of
the
variability
in
drinking
water
consumption
among
different
people
in
the
exposed
population.
EPA
has
recently
completed
a
review
of
estimated
per­
capita
water
ingestion
based
on
data
collected
by
the
U.
S.
Department
of
Agriculture's
(
USDA)
1994­
96
Continuing
Survey
of
Food
Intakes
by
Individuals
(
CSFII).
There
are
two
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
23
distributions
of
consumption:
Distribution
1,
which
served
as
the
basis
for
this
risk
assessment;
and
Distribution
2,
which,
for
comparison,
is
included
alongside
Distribution
1
in
Exhibit
5.6.

Distribution
1
is
based
on
total
water
ingested
from
all
sources
and
has
a
mean
of
1.24
liters
per
day
(
L/
day)
and
a
90th
percentile
of
2.345
L/
day.
Only
the
mean
was
used
in
model
runs
that
estimate
the
national
number
of
cryptosporidiosis
cases
under
the
baseline
and
each
regulatory
alternative
(
i.
e.,
variability
in
individual
consumption
is
not
factored
into
uncertainty
bounds
around
aggregate
national
exposure
estimates).
The
full
Distribution
1
is
used,
however,
in
the
modeling
of
risk
distributions
that
pertain
to
individuals,
where
variability
in
consumption
plays
a
much
larger
role.

EPA's
Office
of
Water
is
continuing
to
evaluate
drinking
water
consumption
data
from
USDA's
1994­
1996
CSFII
study.
The
drinking
water
consumption
distribution
used
as
the
standard
condition
for
the
benefits
analysis
reflects
these
CSFII
data.
This
distribution
is
currently
considered
the
best
estimate
of
water
consumption
among
the
population
that
consumes
drinking
water
from
either
community
or
noncommunity
water
supplies.

Distribution
2,
which
is
based
on
USDA
consumption
data
for
"
community
water
supply,
all
respondents",
has
a
mean
of
0.93
L/
day.
As
a
result,
if
Distribution
2
were
used
instead
of
Distribution
1,
it
would
generate
roughly
75
percent
(
0.93/
1.24=
75
percent)
of
the
cases
predicted
by
Distribution
1
(
the
exponential
dose­
response
curve
is
close
to
linear
in
consumption
across
the
range
of
typical
exposures).

Exhibit
5.6
Distribution
of
Individual
Daily
Drinking
Water
Consumption
(
L/
person/
day)

Percentile
Distribution
1
Distribution
2
1
0.047
0
5
0.184
0
10
0.296
0.032
25
0.584
0.264
50
1.045
0.710
75
1.640
1.313
90
2.345
2.016
95
2.922
2.544
99
4.808
4.242
Source:
USDA's
1994­
96
Continuing
Survey
of
Food
Intakes
by
Individuals
(
CSFII).
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
24
System
Size
(
Population
Served)
Number
of
Systems
Total
Population
Served
Annual
National
Exposure
in
Person­
Days
A
B
C
<
500
6
1,418
496,300
501
­
10,000
43
177,699
62,194,650
10,001
­
100,000
19
560,967
196,338,450
>
100,000
7
10,999,693
3,849,892,550
Totals
75
11,739,777
4,108,921,950
Percent
of
All
PWSs
1.0%
6.8%
6.8%

<
500
1,115
224,780
78,672,850
501
­
10,000
2,384
8,189,755
2,866,414,234
10,001
­
100,000
1,303
41,366,595
14,478,308,308
>
100,000
266
108,945,071
38,130,774,749
Totals
5,068
158,726,200
55,554,170,140
Percent
of
All
PWSs
65.3%
92.1%
92.5%

<
500
511
85,770
21,442,422
501
­
10,000
186
302,991
75,747,719
10,001
­
100,000
9
301,749
75,437,146
>
100,000
1
166,735
41,683,817
Totals
707
857,244
214,311,103
Percent
of
All
PWSs
9.1%
0.5%
0.4%

<
500
1,720
171,247
30,824,460
501
­
10,000
179
393,715
70,868,631
10,001
­
100,000
13
382,696
68,885,280
>
100,000
0
0
0
Totals
1,912
947,658
170,578,371
Percent
of
All
PWSs
24.6%
0.6%
0.3%
Unfiltered
CWSs
Filtered
CWSs
Filtered
NTNCWSs
Filtered
TNCWSs
5.2.4.3
Population
Affected
by
the
LT2ESWTR
and
Exposure
The
number
of
systems
and
total
population
affected
by
the
LT2ESWTR
are
discussed
in
Chapter
4
for
both
unfiltered
(
section
4.4.2)
and
filtered
(
section
4.5.2)
plants.
Exhibit
5.7
summarizes
these
numbers.
Note
that
unfiltered
plants
are
all
within
CWSs.
Note
also
that
more
than
85
percent
of
the
population
affected
by
the
LT2ESWTR
are
served
by
medium
and
large
CWSs.

Exhibit
5.7
Number
of
Systems,
Population
Served,
and
Annual
National
Exposure
by
System
Type
Sources:
[
A]
Unfiltered
CWSs:
Exhibit
4.5,
Column
A;
Filtered
PWSs:
Exhibit
4.11,
Column
G.
[
B]
Unfiltered
CWSs:
Exhibit
4.5,
Column
D;
Filtered
PWSs:
Exhibit
4.11,
Column
I.
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
25
The
risk
assessment
model
also
accounts
for
the
number
of
days
per
year
that
individuals
in
the
affected
population
consume
water
from
the
different
types
of
systems
in
order
to
calculate
the
annual
risk
of
infection
and
illness,
as
described
above.
EPA's
model
accounts
for
exposure
to
people
served
by
Community
Water
Systems
(
CWSs),
Nontransient
Noncommunity
Water
Systems
(
NTNCWSs),
and
Transient
NCWSs
(
TNCWSs)
by
using
different
durations
of
exposure.
Exhibit
5.8
describes
the
three
types
of
PWSs
and
the
annual
number
of
days
of
exposures
modeled
for
individuals
served
by
these
water
systems.
The
risk
assessment
model
evaluates
each
system
type
(
CWS,
NTNCWS,
and
TNCWS)
and
corresponding
days
exposed
independently.

Exhibit
5.8
Individual
Exposure
by
System
Type
Type
of
Water
System
Individual
Exposure
(
days
per
year)
Individuals
Exposed
(
national
aggregate)
Defining
Characteristics
CWS
350
CWS
Population
CWSs
supply
water
to
the
same
population
year­
round.

NTNCWS
250
NTNCWS
Population
NTNCWSs
regularly
supply
water
to
at
least
25
of
the
same
people
for
at
least
6
months
per
year,
but
not
necessarily
yearround

TNCWS
10
A
triangular
distribution
with
a
mode
of
18
times
reported
TNCWS
population
(
peak
population
by
system)
and
a
range
of
9
to
27
times
population.
TNCWSs
supply
drinking
water
in
places
where
people
do
not
remain
for
long
periods
of
time.

Source:
EPA
estimate
National
exposure
derives
from
days
exposed
and
population
served.
These
components
are
addressed
in
two
separate
parts
of
the
risk
model.
The
first
part
of
the
model
incorporates
average
days
of
exposure,
allowing
individual
annual
risk
to
be
estimated
with
precision
for
each
type
of
PWS.
The
second
part
of
the
model
incorporates
population
served,
multiplying
it
by
individual
annual
risk
for
each
system
type.

Overall
exposure
is
not
directly
proportional
to
population
served
due
to
the
differences
in
average
days
exposed
for
the
three
system
types.
For
example,
100
people
served
by
a
CWS
will
have
a
greater
aggregate
exposure
than
100
people
served
by
a
TNCWS
due
to
the
greater
number
of
days
exposed
for
CWS.

To
account
for
the
transient
nature
of
the
population
served
by
TNCWSs,
population
is
multiplied
by
a
factor
between
9
and
27
(
drawn
from
the
triangular
uncertainty
distribution
described
in
Exhibit
5.8).
Without
this
adjustment,
the
reported
TNCWS
population
would
underestimate
the
true
population
exposed,
since
each
TNCWS
system
reports
only
a
peak­
season
population
and
each
individual
exposure
is
assumed
to
be
just
10
days.
The
lower
end
of
the
triangular
distribution
represents
a
3
month
peak
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
26
season
(
9
x
10
days
=
90
days)
and
the
upper
end
a
9
month
peak
season
(
27
x
10
=
270
days).
By
making
this
adjustment
to
the
aggregate
TNCWS
population,
rather
than
to
the
average
number
of
TNCWS
days
exposed,
the
analysis
of
individual
risk
remains
distinct
from
the
analysis
of
aggregate
national
impact.

5.2.5
Risk
Model
Structure
The
risk
assessment
model
integrates
dose­
response
and
exposure
assessment
components
discussed
above
into
a
Monte
Carlo
simulation
model
structured
to
characterize
(
1)
the
distribution
of
individual
risk
of
illness
and
mortality,
and
(
2)
the
total
number
of
national
illnesses
and
deaths
annually
due
to
Cryptosporidium
in
finished
drinking
water.
Modeling
is
carried
under
two
treatment
conditions:
Pre­
LT2ESWTR
and
Post­
LT2ESWTR,
the
former
representing
the
baseline
(
no
action)
conditions
and
the
latter
incorporating
assumptions
of
treatment
improvements
due
to
the
LT2ESTWR.
The
difference
between
the
results
obtained
for
the
improved
conditions
and
for
Pre­
LT2ESWTR
conditions
(
in
terms
of
the
cases
of
illness
and
death
avoided)
constitute
the
quantified
benefits
of
the
rule.

The
risk
assessment
modeling
is
carried
out
in
two
steps.
The
first
step
focuses
on
calculating
the
annual
risk
of
illness.
The
endpoints
of
this
step
are
(
1)
the
distribution
of
annual
risks
of
illness
experienced
by
different
individuals
in
the
population,
reflecting
variability
in
exposure
and
infectivity
conditions,
and
(
2)
the
estimated
average
annual
risk
for
the
population
as
a
whole,
with
confidence
intervals
on
that
estimate
to
reflect
uncertainty.
The
flowchart
in
Exhibit
5.9
summarizes
this
first
step
of
risk
modeling.

The
second
step
applies
the
estimated
average
annual
risk
of
illness,
including
secondary
spread,
to
the
overall
population
to
estimate
the
annual
cases
of
illness,
the
annual
number
of
deaths,
and
the
regulatory
benefits
in
terms
of
illnesses
and
deaths
avoided
due
to
each
proposed
rule.
Discussed
further
in
section
5.3,
the
second
step
also
integrates
information
about
the
monetized
values
of
illnesses
and
deaths
avoided
to
provide
a
dollar
value
for
the
overall
benefits
of
the
rule.
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
27
Pre­
LT2
Baseline
LT2
Reduction
Determine
Rule
Alternative
Treatment
Bin
(
and
Target
Log­
Removal)

Select
Treatment
Technology
from
Decision
Matrices
Filtered
Unfiltered
Select
Occurrence
Distribution
Select
Plant­
Mean
Source
Water
Concentration
of
Total
Oocysts
Determine
Maximum
Running
Annual
Average
(
RAA)
Select
Pre­
LT2
Log
Removal
Distribution
Calculate
Finished
Water
Concentration
of
Infectious
Oocysts
Calculate
Daily
Risk
of
Infection
Apply
Daily
Consumption
of
Drinking
Water
Apply
Cryptosporidiosis
Infectivity
Distribution
Calculate
Daily
Dose
of
Cryptosporidium
Calculate
Annual
Risk
of
Infection
Apply
Morbidity
Factor
Calculate
Annual
Risk
of
Illness
Simulate
24
Monthly
Total
Oocyst
Concentrations
Apply
Days
of
Exposure
per
Year
Apply
Pre­
LT2
Credit
Calculate
LT2
Log
Removal/
Inactivation
Risk
distributions
representing
the
different
system
types,
occurrence
datasets,
and
regulatory
conditions
Calculate
Pre­
LT2
Log
Removal/
Inactivation
Determine
Average
Concentration
Select
Infectious
Oocyst
Fraction
Exhibit
5.9
Flowchart
of
Risk
Model
 
Step
1:
Computing
Annual
Individual
Risk
of
Illness
Step
1
of
the
Risk
Characterization
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
28
Step
1
is
structured
as
a
two­
dimensional
Monte
Carlo
simulation
to
appropriately
address
variability
and
uncertainty
in
model
inputs.
The
basic
algorithm
for
Step
1
of
the
modeling
is:

P
M
=
M*(
1
­
[
exp((­
C*
v*
I)*
r)]
n)

or
more
simply,

=
M*(
1­
exp(­
C*
v*
I*
r*
n))
Where:
P
M
is
the
individual
annual
risk
of
illness
M
is
the
morbidity
factor,
or
the
probability
of
illness
given
an
infection
C
is
the
Cryptosporidium
concentration
in
finished
water
(
oocysts/
Liter)
v
is
the
ratio
of
the
percentage
of
infectious
oocysts
in
the
environment
to
the
percentage
of
infectious
oocysts
in
doses
tested
in
clinical
studies
I
is
mean
individual
daily
drinking
water
consumption
(
Liters)
r
is
the
infectivity
dose­
response
parameter
(
the
expected
probability
of
a
single
ingested
oocyst,
surviving
long
enough
to
reach
an
infection
site
in
the
body)
n
is
the
average
annual
days
per
year
of
exposure
This
formula
is
equivalent
to
that
presented
earlier
in
section
5.2.3
describing
the
dose­
response
function
for
calculating
the
risk
of
illness.
The
difference
is
that
the
d
variable
(
dose
in
infectious
oocysts/
day)
has
been
expanded
to
show
its
components,
namely
Cryptosporidium
occurrence
in
finished
water
(
measured
oocysts/
liter),
infectious
Cryptosporidium
rate
(
infectious
oocysts/
measured
oocyst),
and
drinking
water
consumption
(
liters/
day).

As
noted
above,
the
first
step
of
the
risk
assessment
model
was
structured
as
a
two­
dimensional
Monte
Carlo
simulation.
A
two­
dimensional
simulation
is
used
when
the
model
includes
both
uncertainty
and
variability
components
in
the
inputs,
and
where
it
is
necessary
to
clearly
distinguish
the
influences
of
these
elements
on
the
model
output.
SAS
v8.2
software
was
used
to
carry
out
the
analysis
(
see
Appendix
T
for
progamming
details).

In
the
risk
formula
shown
above,
the
uncertainty
and
variability
components
are:

Uncertainty:
Data
set
representing
source
water
concentration
(
ICR,
ICRSSL,
ICRSSM)
True
distribution
of
source
water
oocyst
concentration
True
distribution
of
Pre­
LT2ESWTR
removal
Morbidity
factor
(
the
probability
of
illness
given
an
infection)
(
M)
Infectious
oocyst
fraction
(
per
total
oocyst's
detected)
(
v)
True
mean
dose­
response
infectivity
parameter
(
r)

Variability:
Source
water
concentration
(
from
a
selected
distribution)
Pre­
LT2ESWTR
Cryptosporidium
removal
(
from
a
selected
distribution)
Predicted
binning
Earned
log
credit
for
enhanced
filtration
(
0
or
0.5
log)
Actual
log
reduction
achieved
due
to
LT2ESWTR
treatment
Variable
daily
ingestion
(
applied
to
step
1
results)
(
I)
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
29
The
form
and
values
for
these
variability
and
uncertainty
distributions
were
discussed
in
the
preceding
sections
on
hazard
identification
(
section
5.2.2)
and
exposure
assessment
(
section
5.2.4).
Exhibit
5.10
summarizes
the
model
parameters.

In
the
two­
dimensional
simulation
structure,
a
set
value
or
a
distribution
of
values
is
randomly
selected
for
each
of
the
uncertainty
parameters
identified
above.
These
uncertainty
parameters
are
then
"
frozen,"
and
a
specified
number
of
iterations
are
performed,
generating
randomly
selected
values
for
the
variability
parameters
(
referred
to
as
"
inner
loops").
These
results
are
stored,
and
a
second
set
of
uncertainty
parameters
is
chosen,
for
which
the
specified
number
of
iterations
are
again
run
for
the
variability
factors.
This
process
is
repeated
for
some
specified
number
of
sets
of
uncertainty
parameters
(
referred
to
as
"
outer
loops").

In
the
risk
model
used
here,
250
sets
of
uncertainty
parameters,
or
outer
loops,
were
generated,
each
with
1,000
variability
iterations,
or
inner
loops.
For
the
model
end­
point
of
this
step
of
the
analysis
"
P
M,
the
annual
individual
risk
of
illness,"
the
following
key
results
were
computed
for
each
of
the
250
uncertainty
iterations
based
on
the
results
of
the
1,000
variability
iterations
performed
within
each
of
those
uncertainty
loops:

°
Mean
°
Standard
Deviation
°
Minimum
°
Maximum
°
Percentiles
(
every
5th
percentile
between
the
5th
and
95th)

°
Percentage
of
population
having
individual
risk
levels
exceeding
10­
2,
10­
3,
10­
4,
10­
5,
10­
6,
and
10­
8
There
were
250
sets
of
these
estimates
for
individual
annual
risk
of
illness
produced,
each
reflecting
a
different
possible
combination
of
the
uncertainty
factors.
These
250
estimates
of
each
of
the
above
statistics
were
then
used
to
compute
an
overall
mean
value
for
each
simulation
group
and
confidence
bounds
on
those
mean
values.

By
structuring
the
first
step
of
the
risk
analysis
in
this
way,
it
was
possible
to
characterize
both
the
distribution
of
individual
annual
risk
of
illness
in
the
affected
population
(
reflecting
variability
in
Cryptosporidium
occurrence
levels
and
daily
water
consumption),
and
the
overall
population
average
annual
risk
of
illness.
This
latter
value,
and
the
associated
uncertainty
in
it
reflected
by
the
250
alternative
values
obtained,
was
then
used
in
the
second
step
of
the
modeling
to
compute
the
number
of
cases
of
illness
and
death
by
applying
these
population
average
risks
to
the
population
exposed.
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
30
Variable
Units
Values/
Range
Variability
or
Uncertainty
Depends
upon
Top
Level
Factors
Proposed
regulatory
alternative
A1,
A2,
A3,
A4
PWS
size
VS,
S,
M,
L
Variability
PWS
type
CWS,
NTNCWS,
TNCWS
Variability
PWS
filtration
status
Filtered,
Unfiltered
Variability
Cryptosporidium
occurrence
dataset
ICR,
SSL,
SSM
Uncertainty
Cryptosporidium
Exposure
plant
mean
oocysts/
liter
Est
[
5th,
95th]
%
tiles
Both
(
V
and
U)
Occurrence
dataset
[
0.0011,
2.7657]
ICR
Filtration
[
0.0033,
0.6765]
SSM
[
0.0059,
0.3460]
SSL
[
0.0004,
0.1177]
Unfiltered
rate/
probability
[
15%,
25%]
ICR
Uncertainty
Dataset
[
30%,
50%]
SSM,
SSL
L/
person/
day
Distribution
1,
Exhibit
5.6
Variability
[
0.047,
4.808]
mean
=
1.24
per
individual
10
(
TNCWS)
Constants
PWS
type
250
(
NTNCWS)
350
(
CWS)
Population
at
risk
(
pop)
per
PWS
size
and
type
see
Exhibit
5.8
Constants
PWS
type
and
size
TNCWS
peak
population
multiplier
multiplication
factor
[
9,
27]
Variability
PWS
type
(
TNCWS
only)

Dose­
Response
Model
Dose­
response
mean
infectivity
(
r)
rate/
probability
[
0,1]
See
Appendix
N
Uncertainty
Morbidity
rate
(
prob
of
illness
given
infection)
(
M)
rate/
probability
[
30%,
70%]
Uncertainty
Secondary
illness
factor
rate/
probability
[
10%,
40%]
Uncertainty
rate/
probability
1.06E­
04
(
Filtered)
Constants
Filtration
1.66E­
04
(
Unfiltered)
(
Population
characterisitics)
Cryptosporidium
Treatment
Reductions
Pre­
LT2ESWTR
Cryptosporidium
reduction
log(
oocysts/
liter)
2.0
to
4.5
log
Both
(
V
and
U)
PWS
size
Pre­
LT2ESWTR
enhanced
filtration
credit
log(
oocysts/
liter)
0
to
0.5
log
Variability
PWS
size
log(
oocysts/
liter)
0
to
2.5
log
Variability
Regulatory
alternative
Occurrence
dataset
Infectious
oocyst
rate
PWS
size
PWS
type
(
including
Filtration)
Predicted
LT2
technology
selection
LT2
toolbox
options
Variability
Predicted
treatment
bin
log(
oocysts/
liter)
0
to
3
log
Variability
Predicted
treatment
bin
Predicted
LT2
technology
selection
Source
water
Cryptosporidium
concentration
(
C)

Infectious
oocysts
per
oocyst
detected
(
v)

Drinking
water
consumption
(
I)

Average
annual
daily
exposures
(
n)

Mortality
rate
(
prob
of
death
given
illness)
(
F)

LT2
treatment
bin
removal
requirement
Actual
Cryptosporidium
reduction
due
to
LT2
treatment
Exhibit
5.10
Overview
of
Risk
Assessment
Model
Parameters
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
31
Step
2
of
the
Risk
Characterization
In
Step
2
of
the
risk
analysis,
the
key
algorithms
used
were:

C
M
=
P
M
x
Pop
C
F
=
C
M
x
F
Where:
C
M
is
the
cases
of
illness
in
the
affected
population
P
M
is
the
distribution
of
individual
annual
probability
of
illness
Pop
is
the
number
of
individuals
in
the
affected
population
(
a
probability
distribution
for
TNCWS)
C
F
is
the
count
of
fatalities
in
the
affected
population
F
is
the
probability
of
fatality
given
an
illness
(
two
values,
see
section
5.2.3)

Step
2
was
conducted
with
a
Monte
Carlo
simulation
(
separate
from
the
simulation
in
Step
1)
in
which
the
variable
P
M(
Avg)
is
treated
as
an
uncertainty
variable,
values
for
which
were
derived
from
a
custom
distribution
based
on
the
250
estimates
of
the
mean
individual
annual
risk
obtained
in
Step
1
of
the
analysis.
From
this
process,
estimates
of
the
cases
of
illness
and
mortality
were
computed,
as
well
as
the
confidence
bounds
on
those
estimates,
for
the
various
baseline
and
Post­
LT2ESWTR
assumptions
regarding
Cryptosporidium
removal
from
source
water.

Secondary
Spread
The
last
step
of
Step
2
is
to
adjust
the
number
of
illness
cases
to
account
for
secondary
spread
(
mortality
is
calculated
from
illness
and
this
is
also
affected
by
secondary
spread).
Secondary
spread
in
this
case
is
infection
passed
through
contact
with
an
individual
initially
infected
by
ingestion
of
contaminated
water.
Secondary
spread
is
quantified
by
the
ratio
of
secondary
cases
to
primary
cases.
The
ratio
varies
depending
on
a
number
of
factors,
such
as
whether
infected
persons
are
symptomatic
(
or
asymptomatic
"
carriers"),
the
age,
health
and
immune
status
of
the
exposed,
and
sanitary
conditions
within
the
household,
office,
or
day
care
centers.

The
secondary
spread
rate
associated
with
endemic
waterborne
cryptosporidiosis
is
estimated
using
data
on
secondary
spread
and
household
secondary
attack
rate
compiled
from
past
cryptosporidiosis
outbreaks
(
Exhibit
5.11).
The
outbreak
data
reported
in
the
exhibit
have
secondary
spread
or
household
secondary
attack
rates
ranging
from
4
percent
to
46
percent.
Most
(
8
of
11)
values
are
in
the
range
between
15
percent
and
37
percent.

In
analyzing
the
available
outbreak
data,
it
is
necessary
to
be
aware
of
at
least
three
potential
effects.
First,
infection
by
Cryptosporidium
appears
to
confer
limited
immunity,
so
the
secondary
spread
rate
may
be
affected
by
the
immune
status
(
previous
infection
history)
of
the
potential
secondarily
exposed
population.
Second,
the
number
of
secondary
cases
during
a
common
source
outbreak
may
be
limited
because
the
outbreak
is
so
large
that
most
people
are
affected
by
the
common
source,
so
many
fewer
people
are
available
to
be
exposed
through
secondary
spread.
Third,
secondary
spread
rates
associated
with
children
(
who
often
acquire
infection
in
day­
care
centers)
are
high
from
the
frequent
handling
of
soiled
diapers
and
training
pants
and
poor
toddler
hygiene
habits.
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
32
The
outbreak
data
in
Exhibit
5.11
suggest
a
triangular
distribution
for
the
range
of
possible
secondary
spread
rates
associated
with
endemic
exposure.
A
preponderance
of
the
rates
are
in
the
middle
of
the
distribution
rather
than
at
the
margins,
yet
the
unusual
scenarios
discussed
above
(
and
others)
will
occasionally
lead
to
extremes
on
both
the
high
and
low
side
of
the
typical
range.

To
capture
uncertainty,
a
triangular
distribution
was
used
with
a
low
at
10
percent,
a
high
at
40
percent
and
a
most
likely
value
of
25
percent.
The
peak
at
25
percent
reflects
the
average
value
of
secondary
spread
and
attack
rates
shown
in
Exhibit
5.11
The
high
value
of
40
percent
in
the
distribution
is
below
the
highest
reported
rate
in
Exhibit
5.11,
and
the
low
value
of
10
percent
is
above
the
lowest
reported
rate.
EPA
chose
to
eliminate
extreme
values
to
minimize
the
impact
of
any
hidden
bias
in
available
data.

With
the
basic
risk
model
described,
we
move
on
to
a
discussion
of
how
the
model
is
used
to
estimate
LT2ESWTR
benefits.
In
short,
the
model
is
used
to
estimate
baseline
conditions
(
Pre­
LT2)
and
then
expected
conditions
arising
under
each
of
the
proposed
rules.
The
differences
 
baseline
to
rule
 
are
the
basis
for
obtaining
the
"
cases
avoided"
(
and
confidence
bounds
on
those
estimates
of
cases
avoided).
As
further
discussed
in
section
5.3,
estimates
of
"
cases
avoided"
generated
in
Step
2
of
the
risk
modeling
are
integrated
with
estimated
costs
of
illness
and
mortality
to
produce
the
monetized
benefit
estimates
for
the
LT2ESWTR.
The
following
subsections
present
these
results.

5.2.6
Individual
Annual
Risk
Distributions
The
benefits
of
the
LT2ESWTR
regulatory
alternatives
have
been
explicitly
estimated
for
two
health
end­
points:
avoided
illnesses
and
avoided
deaths
due
to
endemic
cases
of
cryptosporidiosis.
These
benefits
are
measured
both
in
terms
of
the
number
of
cases
of
illness
and
death
avoided,
and
in
terms
of
the
monetized
value
of
those
avoided
cases.
This
section
focuses
on
the
reduction
in
risk
as
measured
by
anticipated
changes
in
the
distribution
of
individual
risks
in
the
exposed
population
(
Step
1
of
the
risk
characterization).
Section
5.2.7
shows
how
those
changes
in
the
distribution
of
individual
risks
aggregated
across
the
exposed
population
translates
into
the
reduction
in
cases
of
illness
and
death
from
Cryptosporidium
on
a
national
level
(
Step
2).
Results
for
the
small
population
of
unfiltered
systems
are
presented
first,
followed
by
filtered
systems.
Section
5.3
extends
the
benefits
analysis
to
address
the
monetization
of
those
avoided
cases.

The
reader
is
reminded
that
this
section
and
the
next
are
narrowly
focused
on
risks
and
benefits
related
to
endemic
cases
of
cryptosporidiosis.
Other
recognized
benefits
from
the
LT2ESWTR
that
are
not
explicitly
captured
in
the
analysis
presented
here
are
those
associated
with
avoided
Cryptosporidium
outbreaks,
as
well
as
benefits
of
avoided
endemic
and
outbreak
illnesses
and
deaths
from
waterborne
pathogens
other
than
Cryptosporidium
that
might
also
be
prevented
or
controlled
by
these
regulations.

As
summarized
in
the
previous
section,
a
key
output
of
the
risk
model
is
the
estimated
distribution
of
annual
individual
risks
of
endemic
cryptosporidiosis.
The
variability
in
individual
risks
in
the
exposed
population
reflects
the
differences
in
Cryptosporidium
concentration
from
one
location
to
the
next,
treatment
effectiveness,
and
individual
average
daily
water
consumption.
As
a
result,
the
endemic
risk
of
cryptosporidiosis
varies
substantially
across
individuals
in
the
population.
Characterizing
how
individual
risks
are
distributed
prior
to
the
LT2ESWTR
and
how
that
distribution
of
risks
is
expected
to
change
after
the
regulation
is
an
important
component
of
the
overall
benefits
analysis.
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
33
Exhibit
5.11
Secondary
Spread
and
Secondary
Attack
Rates
Associated
with
Cryptosporidiosis
Outbreaks
Tangerman
et
al.
1991
Millard
et
al.
1994
Heijbel
et
al.
1987
Willocks
et
al.
1998
MacKenzie
et
al.
1995a
Brown
et
al.
1989
Bridgman
et
al.
1995
Morgan
et
al.
1995
MacKenzie
et
al.
1995b
Secondary
Spread
[
1]
17/
50
34%
6/
26
23%
17%
­
24%
21%
37%

Household
Secondary
Attack
Rate[
2]
31/
101
31%
53/
353
15%
77/
204
38%
6/
118[
3]

5%
32/
69
46%
­
­
3/
69
4%

Number
of
Confirmed
Cases
(
Adults
and
Children)
39
50
35
345
339
39
47
64
22
Outbreak
Type
Day­
care
Food
Day­
care
Ground
Water
Surface
Water
Unknown
Ground
Water
Ground
Water
Recreational
Water
Location
and
Year
Atlanta,
1989
Maine,
1993
Tulsa,
1984
London,
UK,
1997
Milwaukee,
1993
Great
Yarmouth,
UK,
1986
Warrington,
UK,
1992
UK,
1993
Oshkosh,
1993
Notes:
[
1]
Ratio
and
percent
of
secondary
cases
to
primary
cases
for
those
with
laboratory­
confirmed
cryptosporidiosis.
[
2]
Ratio
and
percent
of
the
number
of
illness
cases
(
not
laboratory
confirmed)
to
the
total
number
of
people
potentially
exposed
in
the
household
of
laboratoryconfirmed
cryptosporidiosis
cases.
[
3]
Ratio
and
percent
of
the
number
of
illness
cases
(
not
laboratory
confirmed)
to
the
total
number
of
people
potentially
exposed
in
the
household
of
an
ill
visitor
to
Milwaukee
during
the
outbreak
(
two
visitors
had
laboratory
confirmed
cryptosporidiosis
and
were
associated
with
44
household
members).
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
34
Exhibits
5.12
and
5.13
present
the
distribution
of
individual
annual
endemic
illness
risk
for
populations
in
CWSs
that
filter
their
water,
and
for
populations
in
CWSs
that
do
not
filter
their
water,
using
the
ICR
occurrence
data
set.
These
exhibits
show
the
portion
of
the
population
that
has
individual
risk
levels
at
or
above
specified
values.
This
provides
a
means
of
focusing
on
the
portion
of
population
having
the
highest
individual
risks,
how
large
that
portion
of
the
population
is,
and
how
the
upper
tails
of
the
risk
distribution
change
as
a
result
of
the
regulatory
alternatives.
Appendix
C
presents
similar
exhibits
for
the
ICRSSM
and
ICRSSL
data
sets.
Appendix
S
further
analyzes
the
filtered
risk
distributions
according
to
bin
requirements.

The
filtered
CWS
individual
risk
distributions
vary
as
a
result
of
the
regulatory
alternatives
considered
for
this
EA.
Exhibit
5.12
shows
the
risk
decreases
from
regulatory
alternative
A4
to
A1.

In
Appendix
C,
the
filtered
risk
distributions
are
displayed
again
based
on
the
ICRSS
data
sets.
Using
different
occurrence
data
sets
produces
different
estimates
of
cases
avoided
and
cases
remaining.
The
cases
avoided
using
the
ICR
data
set
appear
greater
compared
to
using
the
ICRSS
data
sets
(
i.
e.,
the
distance
between
the
Pre­
LT2ESWTR
and
regulatory
alternatives'
distributions
is
greater
for
the
ICR
data
set).
This
is
because
the
ICR
data
predict
higher
concentrations
of
Cryptosporidium
in
source
water,
which
leads
to
more
systems
requiring
treatment.
Because
many
of
the
treatment
options
achieve
more
treatment
than
is
actually
needed
(
see
Exhibit
5.5),
there
is
a
greater
reduction
in
exposure
than
if
lower
levels
of
Cryptosporidium
are
assumed.
Thus,
analyses
using
higher
predicted
Cryptosporidium
levels
will
show
that
proportionally
more
cases
are
avoided
(
because
of
especially
efficient
technologies).

5.2.7
General
Population
Risk
 
Number
of
Cases
Avoided
This
analysis
uses
a
two­
dimensional
Monte
Carlo
simulation
to
develop
estimates
of
the
range
of
individual
risks
of
infection
and
illness
experienced
in
the
general
population,
and
of
the
number
of
annual
infections
and
illnesses
resulting
from
those
risks.
The
algorithms
used
for
calculating
individual
risk
and
the
resulting
number
of
infections
and
illnesses
in
the
overall
population
at
risk,
as
well
as
the
details
on
the
forms
of
the
distributions
used
in
the
Monte
Carlo
simulation
employing
those
algorithms,
have
been
described
previously
in
this
chapter.

This
section
summarizes
the
reduction
in
general
population
risk
for
unfiltered
systems
followed
by
filtered
systems.
Summary
results
presented
in
this
chapter
include
all
system
categories.
More
results
are
presented
in
Appendix
C.
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
35
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%

1.00E­
06
1.00E­
05
1.00E­
04
1.00E­
03
1.00E­
02
1.00E­
01
1.00E+
00
Individual
Risk
Level
(
Illness
Rate)
Percent
of
Population
Exceeding
Risk
Level
Pre­
LT2
Opt
A1
Opt
A2
Opt
A3
Opt
A4
Example:
Under
Pre­
LT2
conditions,
about
46
percent
of
the
population
served
by
filtered
CWSs
have
annual
individual
risk
greater
than
0.0001
(
one
in
a
ten
thousand),
based
upon
the
ICR
occurrence
data
set.

Example:
Under
the
Preferred
Regulatory
Alternative
(
A3),
an
estimated
23
percent
of
the
population
served
by
filtered
CWSs
have
annual
individual
risk
greater
than
0.0001
(
one
in
ten
thousand),
based
upon
ICR
occurrence
data.
Exhibit
5.12
Annual
Individual
Risk
Distributions
Based
Upon
ICR
Occurrence
Data
Set,
Filtered
CWSs
Source:
Risk
Assessment
Model.
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
36
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%

1.00E­
06
1.00E­
05
1.00E­
04
1.00E­
03
1.00E­
02
1.00E­
01
1.00E+
00
Individual
Risk
Level
(
Illness
Rate)
Percent
of
Population
Exceeding
Risk
Level
Pre­
LT2
Post­
LT2
Example:
Under
Pre­
LT2
conditions,
essentially
the
entire
population
served
by
unfiltered
CWSs
have
annual
individual
risk
greater
than
0.0001
(
one
in
ten
thousand).

Example:
Under
the
Preferred
Regulatory
Alternative
(
A3),
an
estimated
39
percent
of
the
population
served
by
unfiltered
CWSs
have
annual
individual
risk
greater
than
0.0001
(
one
in
ten
thousand).
Exhibit
5.13
Annual
Individual
Risk
Distributions
Based
Upon
ICR
Occurrence
Data
Set,
Unfiltered
CWSs
Source:
Risk
Assessment
Model.

5.2.7.1
Unfiltered
Systems
Exhibit
5.14
summarizes
modeled
results
for
unfiltered
systems.
It
gives
both
the
baseline
estimate
of
illnesses
and
deaths,
prior
to
the
LT2ESWTR,
and
the
estimated
number
of
illnesses
and
deaths
avoided
as
a
result
of
implementing
the
LT2ESWTR.
Results
are
presented
for
small
and
large
(
those
serving
10,000
people
or
more)
systems
separately.
Population
figures
are
provided
for
reference.

The
ICRSSM
and
ICRSSL
data
sets
did
not
have
adequate
unfiltered
data
to
generate
modeled
source
water
occurrence.
Therefore,
the
unfiltered
benefits
are
estimated
indirectly,
using
the
ICR
benefit
estimates.
Because
of
the
differences
in
occurrence
data
sets,
the
unfiltered
ICR
cases
of
illness
and
death
could
not
be
directly
added
to
the
filtered
ICRSSM
and
ICRSSL
data
sets.
Instead,
ICR
unfiltered
cases
were
adjusted
by
the
ratio
of
filtered
cases,
ICRSSL
to
ICR.
For
example,
the
ICRSSL­
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
37
estimated
number
of
Pre­
LT2ESWTR
illnesses
for
unfiltered
systems
was
derived
as
follows
(
values
taken
from
Exhibits
5.14
and
5.15):

ICRSSL
UNFILTERED
(
160,903)
=
ICR
UNFILTERED(
550,397)
x
ICRSSL
FILTERED(
126,806)/
ICR
FILTERED(
433,764)
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
38
Systems
Serving
<
10,000
Systems
Serving
>
10,000
All
Systems
Lower
(
5th
%
tile)
Upper
(
95th
%
tile)
Population
at
Risk
179,117
11,560,660
11,739,777
A
B
C
D
E
Pre­
LT2ESWTR
7,601
540,839
548,440
118,006
1,031,806
Post­
LT2ESWTR
41
1,657
1,698
180
4,429
Illnesses
Avoided
7,560
539,182
546,742
117,826
1,027,377
Pre­
LT2ESWTR
2,494
166,271
168,765
36,308
317,527
Post­
LT2ESWTR
11
362
372
22
1,105
Illnesses
Avoided
2,483
165,909
168,393
36,286
316,421
Pre­
LT2ESWTR
4,294
289,552
293,846
63,216
552,868
Post­
LT2ESWTR
19
670
689
47
2,012
Illnesses
Avoided
4,275
288,882
293,157
63,169
550,857
Pre­
LT2ESWTR
1
90
91
20
172
Post­
LT2ESWTR
0
0
0
0
1
Deaths
Avoided
1
90
91
20
171
Pre­
LT2ESWTR
0
28
28
6
53
Post­
LT2ESWTR
0
0
0
0
0
Deaths
Avoided
0
28
28
6
53
Pre­
LT2ESWTR
1
48
49
11
92
Post­
LT2ESWTR
0
0
0
0
0
Deaths
Avoided
1
48
49
11
92
ICR
Data
ICRSSL
Data
ICRSSM
Data
90%
Confidence
Bound
for
All
Systems
Illnesses
Mean
Deaths
ICR
Data
ICRSSL
Data
ICRSSM
Data
Exhibit
5.14
Annual
Cases
of
Illness
and
Deaths
Avoided
for
the
LT2ESWTR,
Preferred
Alternative,
Unfiltered
Systems,
by
Data
Set
Note:
Detail
may
not
add
due
to
independent
rounding.
Sources:
Population
at
risk
from
the
risk
assessment
model.
[
A]
Pre­
LT2ESWTR
data
from
Appendix
C,
Exhibit
C.
3,
Columns
A
and
B,
Row
­
Small
Systems,
ICR,
ICRSSL,
and
ICRSSM
Unfiltered.
Illnesses
Avoided
data
from
Appendix
C,
Exhibit
C.
7,
Columns
A
and
D,
Row
­
Small
Systems,
ICR,
ICRSSL,
and
ICRSSM
[
B]
Pre­
LT2ESWTR
data
from
Appendix
C,
Exhibit
C.
3,
Columns
A
and
B,
Row
­
Large
Systems,
ICR,
ICRSSL,
and
ICRSSM
Unfiltered.
Illnesses
Avoided
data
from
Appendix
C,
Exhibit
C.
7,
Columns
A
and
D,
Row
­
Large
Systems,
ICR,
ICRSSL,
and
ICRSSM
[
C]
Pre­
LT2ESWTR
data
from
Appendix
C,
Exhibit
C.
3,
Columns
A
and
B,
Row
­
All
Systems,
ICR,
ICRSSL,
and
ICRSSM
Unfiltered.
Illnesses
Avoided
data
from
Appendix
C,
Exhibit
C.
7,
Columns
A
and
D,
Row
­
All
Systems,
ICR,
ICRSSL,
and
ICRSSM
[
D]
Pre­
LT2ESWTR
data
from
Appendix
C,
Exhibit
C.
3,
Columns
B
and
F,
Row
­
All
Systems,
ICR,
ICRSSL,
and
ICRSSM
Unfiltered.
Illnesses
Avoided
data
from
Appendix
C,
Exhibit
C.
7,
Columns
B
and
E,
Row
­
All
Systems,
ICR,
ICRSSL,
and
ICRSSM
[
E]
Pre­
LT2ESWTR
data
from
Appendix
C,
Exhibit
C.
3,
Columns
C
and
G,
Row
­
All
Systems,
ICR,
ICRSSL,
and
ICRSSM
Unfiltered.
Illnesses
Avoided
data
from
Appendix
C,
Exhibit
C.
7,
Columns
C
and
F,
Row
­
All
Systems,
ICR,
ICRSSL,
and
ICRSSM
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
39
Systems
Serving
<
10,000
Systems
Serving
>
10,000
All
Systems
Lower
(
5th
%
tile)
Upper
(
95th
%
tile)
Population
at
Risk
9,368,257
151,162,846
160,531,102
A
B
C
D
E
Pre­
LT2ESWTR
56,528
446,900
503,429
49,493
1,418,729
Post­
LT2ESWTR
4,094
27,161
31,256
3,432
89,300
Illnesses
Avoided
52,434
419,739
472,173
46,061
1,329,429
Pre­
LT2ESWTR
18,484
137,382
155,866
16,366
451,166
Post­
LT2ESWTR
8,366
59,719
68,085
8,301
187,670
Illnesses
Avoided
10,117
77,663
87,780
8,065
263,496
Pre­
LT2ESWTR
31,841
239,570
271,411
25,576
836,388
Post­
LT2ESWTR
8,326
57,879
66,205
7,011
188,790
Illnesses
Avoided
23,516
181,691
205,206
18,565
647,598
Pre­
LT2ESWTR
6
47
53
5
150
Post­
LT2ESWTR
0
3
3
0
9
Deaths
Avoided
6
44
50
5
140
Pre­
LT2ESWTR
2
28
16
2
48
Post­
LT2ESWTR
1
19
7
1
20
Deaths
Avoided
1
8
9
1
28
Pre­
LT2ESWTR
3
25
29
3
88
Post­
LT2ESWTR
1
6
7
1
20
Deaths
Avoided
2
19
22
2
68
ICRSSM
Data
ICR
Data
ICRSSL
Data
Illnesses
ICR
Data
ICRSSL
Data
Deaths
ICRSSM
Data
90%
Confidence
Bound
for
All
Systems
Mean
Exhibit
5.15
Annual
Cases
of
Illness
and
Deaths
Avoided
Due
to
the
LT2ESWTR,
Preferred
Alternative,
All
Filtered
Systems,
by
Data
Set
Note:
Detail
may
not
add
due
to
independent
rounding.
Sources:
Population
at
risk
from
the
risk
assessment
model.
[
A]
Pre­
LT2ESWTR
data
from
Appendix
C,
Exhibit
C.
3,
Columns
A
and
B,
Row
­
Small
Systems,
ICR,
ICRSSL,
and
ICRSSM
Filtered.
Illnesses
Avoided
data
from
Appendix
C,
Exhibit
C.
6,
Columns
A
and
D,
Row
­
Small
Systems,
ICR,
ICRSSL,
and
ICRSSM
[
B]
Pre­
LT2ESWTR
data
from
Appendix
C,
Exhibit
C.
3,
Columns
A
and
B,
Row
­
Large
Systems,
ICR,
ICRSSL,
and
ICRSSM
Filtered.
Illnesses
Avoided
data
from
Appendix
C,
Exhibit
C.
6,
Columns
A
and
D,
Row
­
Large
Systems,
ICR,
ICRSSL,
and
ICRSSM
[
C]
Pre­
LT2ESWTR
data
from
Appendix
C,
Exhibit
C.
3,
Columns
A
and
B,
Row
­
All
Systems,
ICR,
ICRSSL,
and
ICRSSM
Filtered.
Illnesses
Avoided
data
from
Appendix
C,
Exhibit
C.
6,
Columns
A
and
D,
Row
­
All
Systems,
ICR,
ICRSSL,
and
ICRSSM
[
D]
Pre­
LT2ESWTR
data
from
Appendix
C,
Exhibit
C.
3,
Columns
B
and
F,
Row
­
All
Systems,
ICR,
ICRSSL,
and
ICRSSM
Filtered.
Illnesses
Avoided
data
from
Appendix
C,
Exhibit
C.
6,
Columns
B
and
E,
Row
­
All
Systems,
ICR,
ICRSSL,
and
ICRSSM
[
E]
Pre­
LT2ESWTR
data
from
Appendix
C,
Exhibit
C.
3,
Columns
C
and
G,
Row
­
All
Systems,
ICR,
ICRSSL,
and
ICRSSM
Filtered.
Illnesses
Avoided
data
from
Appendix
C,
Exhibit
C.
6,
Columns
C
and
F,
Row
­
All
Systems,
ICR,
ICRSSL,
and
ICRSSM
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
40
This
method
is
applicable
to
the
Pre­
LT2ESWTR
of
cases
avoided
estimates.
However,
when
estimating
the
number
of
illnesses
avoided
due
to
the
LT2ESWTR,
estimates
are
dependent
on
the
regulatory
alternative.
In
order
to
carry
out
the
calculation
above,
an
ICRSSL/
ICR
illnesses­
avoided
ratio
for
filtered
systems
is
required,
but
which
ratio
is
best?
There
are
four
regulatory
alternatives
for
filtered
systems,
and
four
corresponding
ICRSSL/
ICR
ratios.
(
This
question
does
not
arise
for
unfiltered
systems
because
there
is
only
one
regulatory
alternative.)

To
address
this
issue,
EPA
evaluated
all
regulatory
alternatives
to
identify
the
one
with
filtered
treatment
requirements
most
similar
to
the
unfiltered
requirements.
EPA
concluded
that
alternative
A1
was
most
similar,
in
terms
of
expected
reductions
in
Cryptosporidium.
Alternative
A1
is
an
across­
theboard
2
log
reduction
that
impacts
all
filtered
systems
regardless
of
initial
monitoring
results,
as
opposed
to
alternatives
A2
through
A4
that
are
expected
to
impact
fewer
than
half
of
filtered
systems
(
see
Appendix
B).

Therefore,
in
the
equation
above,
the
alternative
A1
ratio,
ICRSSL
to
ICR,
was
used
to
compute
the
expected
number
of
illnesses
avoided
for
unfiltered
systems
based
on
the
ICRSSL
data
set.
This
single
estimate
of
ICRSSL
unfiltered
illnesses
avoided
then,
was
added
to
the
each
of
four
ICRSSL
estimates
for
filtered
system
illnesses
avoided
to
obtain
a
national
estimate
for
each
regulatory
alternative.

5.2.7.2
Filtered
Systems
Exhibit
5.15
summarizes
the
estimated
Pre­
LT2ESWTR
cases
of
illness
and
deaths
associated
with
endemic
Cryptosporidium
occurrence,
and
estimated
cases
of
illness
and
deaths
avoided
for
populations
served
by
filtered
water
systems
as
a
result
of
the
LT2ESWTR.
Results
are
presented
for
small
and
large
systems
(
those
serving
at
least
10,000
people)
separately
and
for
the
three
occurrence
distribution
data
sets.
Population
figures
are
provided
for
reference.

5.2.8
Reduction
in
Sensitive
Subpopulation
Risk
Morbidity
risk
in
this
analysis
is
based
on
studies
of
infectivity
and
morbidity
done
on
healthy
volunteers.
No
data
currently
exist
that
would
give
a
differential
infectivity
or
morbidity
for
the
immunocompromised
and
other
sensitive
subpopulations.
Therefore,
this
analysis
has
not
accounted
for
possible
elevated
infectivity
and
morbidity
in
these
populations.
The
mortality
risk
from
Cryptosporidium
in
this
analysis
is
expressed
as
the
probability
of
death
given
an
illness,
derived
from
the
study
of
the
1993
Milwaukee
outbreak.
The
majority
of
the
fatalities
due
to
cryptosporidiosis
in
that
outbreak
were
AIDS
patients,
and
the
remainder
were
elderly.
Since
all
observed
mortality
has
been
in
sensitive
subpopulations,
all
of
the
quantified
deaths
avoided
due
to
the
LT2EWSTR
are
presumed
to
be
lives
saved
in
sensitive
populations.

5.3
Monetized
Benefits
from
Reduction
in
Exposure
to
Cryptosporidium
Resulting
from
the
LT2ESWTR
Once
the
annual
endemic
illnesses
and
deaths
avoided
as
a
result
of
the
LT2ESWTR
are
estimated
using
the
risk
model
described
in
the
previous
sections,
monetary
unit
values
can
be
applied
to
these
estimates
to
establish
the
monetary
benefits
attributable
to
the
rule.
Because
the
quantified
6
Although
estimates
of
averting
expenditures
incurred
during
Cryptosporidium
outbreaks
are
available,
the
value
of
avoiding
these
expenditures
is
not
quantified
in
this
analysis.
The
risk
assessment
focuses
on
endemic
risk
rather
than
risk
from
outbreaks,
and
averting
behaviors
would
be
less
likely
to
occur
in
the
absence
of
a
publicized
outbreak.

Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
41
projection
of
illnesses
avoided
is
underestimated
(
risks
from
outbreaks
and
risks
from
other
pathogens
are
not
quantified),
the
projection
of
monetized
benefits
are
similarly
underestimated.
Monetary
benefits
are
estimated
using
different
methodologies
for
illnesses
and
deaths
avoided.
In
addition,
two
alternative
values
are
used
for
estimates
of
the
cost
of
illness
(
COI)
avoided
 
the
Enhanced
COI
and
Traditional
COI.
The
methodologies
and
the
resulting
monetary
benefits
estimates
are
presented
in
the
subsections
that
follow.

5.3.1
Value
of
Reduction
in
Cryptosporidiosis
Cases
5.3.1.1
Value
of
Illnesses
Avoided
The
goal
of
this
analysis
is
to
provide
as
complete
an
accounting
as
possible
of
the
social
welfare
impacts
of
the
regulatory
options
under
consideration.
In
this
context,
based
on
the
principles
of
welfare
economics,
the
preferred
approach
for
valuing
reductions
in
the
risk
of
cryptosporidiosis­
related
morbidity
is
to
rely
on
estimates
of
willingness
to
pay
for
these
risk
reductions.
However,
a
review
of
the
literature
indicates
that
the
available
studies
address
illnesses
with
significantly
different
effects
from
those
associated
with
cryptosporidiosis,
hence
estimates
from
this
literature
are
inappropriate
here.
This
analysis
instead
estimates
the
value
of
averted
morbidity
risks
based
on
(
1)
the
avoided
medical
costs
and
(
2)
the
value
of
averted
time
losses.
The
rationale
for,
and
limitations
of,
this
approach
are
introduced
below
and
discussed
in
greater
detail
in
Appendix
K.
Appendix
L
describes
the
calculations
used
in
the
appendix.
Appendix
P
contains
a
sensitivity
analysis
using
alternative
values
for
two
key
inputs
to
the
Enhanced
COI.

The
calculation
of
medical
costs
includes
the
costs
of
medical
services
and
medications
received
by
ill
individuals.
The
assumption
behind
using
these
costs
as
a
benefit
measure
is
that
reduced
incidence
of
illness
will
yield
benefits
at
minimum
level
equal
to
the
costs
saved.
However,
COI
estimates
may
significantly
underestimate
individual
willingness
to
pay,
for
a
variety
of
reasons.
In
particular,
these
estimates:
(
1)
may
not
fully
address
the
value
of
avoiding
pain
and
suffering;
(
2)
do
not
include
costs
that
individuals
incur
to
avoid
the
illness
(
i.
e.,
defensive
or
averting
expenditures);
6
(
3)
do
not
reflect
aversion
to
risk
(
the
fear
of
becoming
ill);
(
4)
do
not
consider
ex
ante
values
(
they
are
based
on
ex
post
costs);
and
(
5)
do
not
consider
whether
treatment
returns
individuals
to
the
original
state
of
health
(
i.
e.,
is
equivalent
to
avoiding
the
illness
entirely).

A
number
of
researchers
have
explored
the
relationship
between
the
COI
and
individual
willingness
to
pay
for
risk
reductions
for
illnesses
other
than
cryptosporidiosis.
This
research
suggests
that
the
ratio
of
these
two
quantities
varies
greatly
depending
on
the
nature
of
the
health
effect,
the
characteristics
of
the
individuals
studied,
and
the
factors
included
in
the
construction
of
each
estimate.
Comparison
studies
result
in
ratios
of
willingness
to
pay
to
cost
of
illness
that
range
from
7
See
Appendix
B
of
EPA's
Handbook
for
Non­
Cancer
Health
Effects
Valuation
(
USEPA
2000h)
for
a
review
of
these
studies.

8
Paid
care
is
included
in
the
medical
cost
component
of
the
analysis
and
hence
is
not
discussed
in
the
discussion
of
time
losses.

Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
42
about
a
factor
of
2
to
as
much
as
a
factor
of
79
(
in
one
case);
many
of
the
ratios
are
between
3
and
6.7
In
other
words,
the
cost
of
illness
estimates
were
typically
one­
third
to
one­
sixth
of
the
willingness
to
pay
estimates,
but
the
ratio
varied
greatly.

In
some
cases,
COI
studies
include
indirect
(
i.
e.,
work
time
for
which
payment
is
received),
as
well
as
direct
costs.
These
indirect
costs
usually
include
lost
earnings
due
to
missed
market
work
time,
and
may
also
include
costs
associated
with
reduced
productivity
while
at
work
and/
or
lost
nonmarket
work
time
(
e.
g.,
child
care
or
housekeeping).
Typically,
these
costs
are
estimated
using
the
human
capital
approach,
which
focuses
on
the
value
of
goods
and
services
that
are
bought
and
sold
in
the
marketplace
and
ignores
other
aspects
of
time
use
that
affect
individual
well­
being.

The
analysis
of
cryptosporidiosis­
related
morbidity
uses
two
measures
of
the
COI
 
Traditional
and
Enhanced.
Both
approaches
include
of
the
direct
medical
costs
and
the
value
of
lost
work
time,
but
differ
in
the
assessment
of
value
of
lost
work
time.
They
both
consider
the
impact
of
time
losses
on
foregone
market
production,
which
affects
the
individual
worker
(
e.
g.,
in
terms
of
lost
income)
as
well
as
other
members
of
society
(
who
benefit
from
the
availability
of
the
goods
or
services
produced
as
well
as
the
taxes
paid),
and
foregone
nonmarket
(
household
and
volunteer)
production,
which
affects
the
individual
and
other
household
members
and
often
has
impacts
outside
the
home.
The
Traditional
COI
includes
nonmarket
(
unpaid)
work
time
based
on
replacement
costs.
The
other
approach,
the
Enhanced
COI,
values
nonmarket
work
time
based
on
opportunity
costs.
Both
approaches
also
include
values
for
the
unpaid
time
lost
by
friends
or
family
members
caring
for
those
who
are
sick,
8
but
the
approaches
use
different
values
for
this
lost
time.

The
Enhanced
COI
also
includes
the
value
of
lost
leisure
time
and
lost
productivity
 
the
reduced
utility
(
or
sense
of
well­
being)
associated
with
decreased
enjoyment
of
time
spent
in
both
work
and
nonwork
activities.
The
Enhanced
COI
is
an
attempt
to
more
completely
measure
of
the
loss
of
welfare
from
an
illness.

Regarding
how
to
best
value
lost
work
time,
a
search
of
the
literature
suggests
that
researchers
have
not
attempted
to
directly
estimate
(
e.
g.,
through
surveys)
the
difference
between
the
value
of
time
in
a
well
state
compared
to
an
ill
state.
Hence,
this
analysis
instead
relies
on
wage
and
compensation
data
to
estimate
the
opportunity
costs
of
time
usage.
This
approach
recognizes
that,
because
resources
are
limited,
any
decision
to
use
resources
for
one
purpose
means
that
they
cannot
be
used
for
other
purposes.
Therefore,
the
value
of
the
resource
can
be
determined
based
on
the
value
of
its
next
best
use.

The
application
of
the
opportunity
cost
approach
to
paid
work
time
is
relatively
clear,
since
compensation
can
be
used
to
estimate
these
costs.
More
precisely,
lost
market
work
is
valued
at
the
median
gross
(
pre­
tax)
wage
rate
plus
benefits,
also
referred
to
as
total
compensation
or
employer's
costs.
This
approach
is
most
representative
of
the
full
social
impact
of
lost
work
time
because
it
incorporates
both
the
income
loss
to
the
individual
and
the
loss
to
society
in
terms
of
reduced
tax
revenue
or
decreased
production
of
goods
and
services.
9
A
sensitivity
analysis
that
uses
alternative
values
for
nonwork
time
in
the
calculation
of
the
Enhanced
COI
is
included
in
Appendix
P.

10
A
sensitivity
analysis
that
uses
alternative
values
for
lost
productivity
in
the
calculation
of
the
Enhanced
COI
is
included
in
Appendix
P.

11
A
pioneering
example
of
this
approach
is
Rice
1966;
a
more
recent
example
is
Thamer
et
al.
1998.

12
A
number
of
other
simplifying
assumptions
inherent
in
this
approach
may
lead
it
to
under­
or
overstate
the
value
of
time
losses.
These
relate
to
factors
such
as
the
functioning
of
the
labor
market,
the
treatment
of
individuals
who
are
not
labor
force
participants,
the
use
of
average
or
median
(
rather
than
marginal)
earnings
data,
and
the
possibility
that
substitute
activities
(
e.
g.,
watching
TV
instead
of
normal
activities)
have
some
positive
value
if
not
offset
by
the
utility
losses
from
the
discomfort
and
stress
of
being
sick.
It
is
unclear
whether,
in
total,
these
practical
limitations
serve
to
increase
or
decrease
the
bias
that
results
from
the
sources
discussed
in
this
paragraph.

Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
43
For
unpaid
time
spent
in
nonmarket
work
and
leisure
time,
wage
data
are
also
used.
The
Enhanced
COI
assumes
that
(
at
the
margin)
the
wage
represents
the
opportunity
cost
of
engaging
in
such
activities.
Lost
nonwork
time
(
including
nonmarket
work
and
leisure)
is
valued
at
the
median
net
(
posttax
wage
rate.
This
approach
reflects
the
assumption
that,
at
the
margin,
an
individual
will
choose
to
engage
in
nonmarket
work
or
leisure
activities
only
if
the
value
of
these
activities
exceeds
the
post­
tax
wage
rate
that
the
individual
would
otherwise
earn.
9
These
values
are
applied
to
both
complete
losses
of
time
(
time
spent
in
illness­
related
activities
rather
than
normal
activities)
as
well
as
to
partial
losses
(
time
spent
in
normal
activities
that
is
less
productive
or
pleasurable
than
in
the
absence
of
illness).
In
the
later
case,
however,
the
dollar
value
of
the
loss
is
prorated
to
reflect
the
fact
that
the
individual
does
not
completely
lose
the
productivity
or
utility
associated
with
the
activity.
These
values
are
applied
to
unpaid
caretakers
whose
normal
activities
are
affected
by
illness
as
well
as
to
time
losses
accruing
to
the
ill
individual.
10
For
unpaid
time
spent
in
nonmarket
work,
the
Traditional
COI
also
uses
wage
data.
The
value
used
is
half
of
the
after­
tax
wage.
The
use
of
50
percent
of
the
wage
rate
is
consistent
with
the
common
practice
in
the
human
capital
literature
of
valuing
nonmarket
work
time
at
the
market
rate
for
domestic
workers.
11
This
literature
uses
replacement
costs
as
a
measure
of
the
productivity
of
nonmarket
work,
rather
than
focusing
on
the
opportunity
costs
(
or
utility
loss)
for
the
individual
who
chooses
to
engage
in
nonmarket
work.
In
support
of
the
use
of
50
percent
of
the
after­
tax
wage
rate,
the
median
weekly
earnings
of
private
household
workers
in
the
service
industry
were
$
261
per
week
in
2000,
about
45
percent
of
the
median
weekly
earnings
of
$
576
for
all
workers
(
U.
S.
Census
Bureau
Table
621,
2001).
Private
household
workers
include
childcare
workers,
cleaners,
and
servants.
The
Traditional
COI
does
not
include
values
for
lost
leisure
time
or
lost
productivity.

Sleep
time
presents
special
problems
in
this
analysis,
in
part
because
data
on
the
effect
cryptosporidiosis­
related
morbidity
on
the
amount
or
quality
of
sleep
time
is
not
available.
Thus,
this
analysis
conservatively
assumes
that
lost
sleep
time
has
zero
value.

The
use
of
medical
costs
and
the
opportunity
cost
of
time
to
value
cryptosporidiosis­
related
morbidity
may
understate
the
value
of
these
risk
reductions
for
a
variety
of
reasons.
12
As
noted
earlier,
COI
estimates
generally
understate
willingness
to
pay
for
a
variety
of
reasons,
e.
g.,
because
they
may
not
fully
consider
the
value
of
avoided
pain
and
suffering
or
of
risk
aversion.
In
addition,
the
use
of
wage
and
compensation
data
to
value
lost
time
may
understate
the
utility
of
time
spent
in
its
preferred
use.
The
use
of
wage
rates
may
understate
the
total
utility
associated
with
an
activity
even
in
the
case
of
paid
work,
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
44
because
individuals
may
derive
intrinsic
pleasure
from
the
activity
above
and
beyond
the
income
they
receive.
For
nonmarket
work
and
leisure,
the
value
of
the
activity
to
the
individual
may
exceed
the
opportunity
cost
for
similar
reasons.
In
addition,
nonmarket
work
and
other
activities
can
provide
benefits
to
other
members
of
society
that
are
not
reflected
in
the
individual
wage
rate.
Finally,
neither
the
Tradtional
or
Enhanced
COI
approach
includes
the
value
of
lost
sleep
time.

In
addition,
relying
on
wage
data
for
valuing
lost
time
presents
difficulties
in
the
case
of
individuals
for
whom
these
data
do
not
exist,
such
as
children,
the
unemployed
who
are
seeking
employment,
and
those
out
of
the
labor
market.
The
approach
taken
in
this
analysis
is
to
value
all
time
losses
at
the
rates
applicable
to
adult
wage
earners.
It
is
unclear
whether
this
approach
under­
or
overstates
the
value
of
times
losses
for
the
individuals
in
these
other
categories,
given
the
lack
of
information
on
these
values.

COI
Calculations
The
primary
risk
of
illness
that
LT2ESWTR
addresses
is
from
endemic
exposure
to
Cryptosporidium
and
the
resulting
cases
of
cryptosporidiosis.
Direct
measurements
of
many
elements
of
the
COI
were
made
as
part
of
an
investigation
of
the
1993
Milwaukee
Cryptosporidium
outbreak.
Some
of
the
data
collected
during
that
outbreak
have
been
reported
in
Corso
et
al.
(
2003),
whose
data
sources
include
the
original
epidemiological
investigation
by
CDC
and
State
personnel
that
included
telephone
surveys
and
a
review
of
hospital
records.
The
epidemiological
investigation
collected
information
on
the
duration
of
illness,
types
of
medication
taken,
medical
care
sought,
if
any,
and
the
costs
associated
with
these
services.
The
data
from
that
report
and
other
sources
of
information
used
in
the
analysis
that
follows
are
shown
in
Appendix
L.

The
computation
of
COI
involves
two
broad
categories
 
direct
medical
costs
and
the
value
of
lost
time
(
Exhibit
5.16).
The
components
are
updated
to
a
common
month
and
year
(
December
2000),
which
is
used
as
the
starting
point
for
projecting
benefits
into
future
time
periods.
For
each
of
these
components,
separate
estimates
are
made
based
on
the
severity
of
the
illness.
Illnesses
are
sorted
into
three
severity
categories
as
follows:

°
Mild:
the
person
did
not
seek
professional
medical
care
for
the
illness.

°
Moderate:
the
person
had
one
or
more
outpatient
visits
to
a
physician
or
emergency
room,
but
the
person
was
ultimately
not
hospitalized.

°
Severe:
the
person
was
hospitalized
one
or
more
times.

The
percentage
of
people
in
each
of
the
severity
classifications
is
used
to
derive
an
average
weighted
cost
per
patient.
The
average
loss
per
case
of
cryptosporidiosis
incorporating
all
categories
is
approximately
$
745
for
the
Enhanced
COI
and
$
245
for
the
Traditional
COI.
Of
those
totals,
approximately
$
94
is
for
direct
medical
costs
(
Exhibit
5.16).
The
details
of
the
computations
are
discussed
in
the
next
three
subsections.

Direct
Medical
Costs
As
Exhibit
5.16
shows,
the
weighted
average
of
direct
medical
costs
per
case
of
illness
is
$
93.82.
Costs
for
doctor
visits,
emergency
room
(
ER)
visits,
hospital
stays,
ambulance
costs,
and
costs
of
medication
comprise
the
direct
medical
costs.
(
As
mentioned
earlier,
medications
can
help
relieve
some
13
Bureau
of
Labor
Statistics,
264.8
(
Dec2000$)/
205.2
(
Dec1993$)
=
1.29
CPI­
U
medical
cost
update
factor.

Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
45
Mild
Moderate
Severe
Mild
Moderate
Severe
Mild
(
88%)
Moderate
(
11%)
Severe
(
1%)

Doctor
Visits
NA
$
45.00
$
45.00
NA
$
58.05
$
58.05
NA
95%*$
58.05
=$
55.15
29%*$
58.05
=$
16.83
Emergency
Room
Visits
NA
$
224.00
$
224.00
NA
$
288.96
$
288.96
NA
5%*$
288.96
=$
14.45
71%*$
288.96
=$
205.16
Hospital
Stays
NA
NA
$
6,152.96
NA
NA
$
7,937.32
NA
NA
100%*
7,937.32
=$
7,937.32
Ambulance
NA
$
228.00
$
228.00
NA
$
294.12
$
294.12
NA
4.9%*
5%*
$
294.12
=$
0.72
16.3%*$
294.12
=$
47.94
Medication
$
5.73
$
5.92
$
6.74
$
7.39
$
7.64
$
8.69
30%*$
7.39
=$
2.22
30%*$
7.64
=$
2.29
29%*$
8.69
=$
2.52
Medication
after
Health
Care
NA
$
8.91
$
70.52
NA
$
11.49
$
90.97
NA
54%*$
11.49
=$
6.21
48%*$
90.97=
$
43.67
Medication
Taken
upon
Reoccurrence
$
2.44
$
2.44
$
2.44
$
3.15
$
3.15
$
3.15
21%*$
3.15
=$
0.66
21%*$
3.15
=$
0.66
21%*$
3.15
=$
0.66
Totals
$
2.88
$
79.47
$
8,254.10
Weighted
Total
$
93.82
December
2000$
December
2000$
Average
Cost
[
1]

Medical
Cost
1993$
Average
Cost
Per
Patient
symptoms,
but
do
not
help
cure
infection).
All
direct
medical
costs
are
obtained
in
December
1993
dollars
and
updated
by
a
1.29
cost
per
illness
(
CPI­
U)
update
factor
to
December
2000
dollars.
13
The
Corso
et
al.
(
2003)
report
states
that
costs
for
doctor
visits
are
not
applicable
to
those
with
mild
cases
but
are
for
95
percent
of
moderate
cases
and
29
percent
of
severe
cases.
Part
of
the
reason
for
the
greater
percentage
of
doctor
visits
for
moderate
cases
lies
in
the
definition
of
doctor
visits.
This
category
includes
those
patients
whose
primary
medical
attention
was
a
non­
ER
physician.
Conversely,
5
percent
of
moderate
cases
and
71
percent
of
severe
cases
went
to
the
ER.
Multiplying
through
by
the
respective
costs
of
doctor
visits
($
58.05)
and
ER
visits
($
288.96)
yields
average
costs
of
$
55.15
per
moderate
case
and
$
16.83
per
severe
case
for
doctor
visits,
and
an
average
of
$
14.45
per
moderate
case
and
$
205.16
per
severe
case
per
ER
visit.

Exhibit
5.16
Direct
Medical
Costs
of
a
Case
of
Cryptosporidiosis
Notes:
Detail
may
not
add
to
totals
due
to
independent
rounding.
[
1]
All
direct
medical
costs
are
obtained
in
December
1993$
and
updated
by
a
1.29
CPI­
U
update
factor
to
December
2000$.
Bureau
of
Labor
Statistics,
264.8
(
Dec2000$)/
205.2
(
Dec1993$)
=
1.29.
Sources:
1993$
average
cost
data
from
Corso
et
al.
(
2003).

The
average
cost
of
a
hospital
stay
was
$
7,937.32.
This
figure
is
applied
only
to
severe
cases.
The
only
medication
costs
that
were
the
same
for
all
levels
of
severity
are
medication
taken
upon
reoccurrence.
The
average
costs
of
medication
for
reoccurrence
was
$
3.15
per
person.
Of
the
people
infected
with
cryptosporidiosis,
21
percent
had
at
least
one
reoccurrence.
Otherwise,
medication
costs
varied
depending
on
severity
of
illness.
For
medication
used
before
receiving
medical
attention,
costs
are
similar
across
severity
groups;
$
7.39
for
mild,
$
7.64
for
moderate
and
$
8.69
for
severe
cases
of
cryptosporidiosis.
Medication
costs
after
health
care
varied.
In
addition,
these
costs
only
applied
to
non­
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
46
mild
cases;
costs
amounted
to
$
11.49
for
moderate
cases
and
$
90.97
for
severe
cases,
but
are
weighted
based
on
the
percentage
of
people
taking
medication.
Summing
all
of
the
above
costs
and
obtaining
a
weighted
average
by
severity
level
yields
an
overall
weighted
average
of
$
93.82
for
direct
medical
costs
per
illness.

Value
of
Lost
Time
Per
Day
The
value
of
lost
time
is
derived
through
several
steps
shown
in
summary
below,
and
discussed
in
detail
in
Appendix
L.
First,
the
number
of
days
lost
and
days
with
lessened
productivity
(
for
the
Enhanced
COI)
must
be
generated.
Exhibit
5.17
shows
the
days
lost
by
severity
of
illness,
and
Exhibit
5.18
calculates
the
average
days
lost
weighted
by
percent
of
cases
with
each
severity
of
illness.
For
the
21
percent
of
cases
with
a
2­
day
reoccurrence
of
the
illness
(
Corso
et
al.,
2003),
the
analysis
assumes
these
have
at
least
2
days
reduced
productivity
(
Appendix
L).

Exhibit
5.17
Days
Lost
and
Days
with
Lost
Productivity,
by
Severity
of
Illness
Illness
Severity
Time
Category
Mean
Duration
of
Illness
Days
Lost
to
Illness
Lost
Productivity
Days
A
B
C=
A­
B
Mild
4.7
1.3
3.4
Moderate
9.4
3.8
5.6
Severe
34.0
5.6
20.5
Source:
Exhibit
L.
5.

Exhibit
5.18
Weighted
Average
Days
Lost
for
Work,
Caregivers,
and
Productivity
Severity
Days
Lost
Weight
Weighted
Average
Days
Work
(
Patients)
Mild
1.3
88%
1.144
Moderate
3.8
11%
0.418
Severe
13.5
1%
0.135
Total
1.697
Caregivers
Mild
0.1
88%
0.088
Moderate
1.3
11%
0.143
Severe
3.9
1%
0.039
Total
0.270
Productivity
Losses
Mild
3.4
88%
2.992
Moderate
5.6
11%
06.616
Severe
20.5
1%
0.205
Reoccurrence
2.0
21%
0.420
Total
4.233
Source:
Exhibit
L.
6
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
47
Second,
the
value
of
time
must
be
estimated.
Exhibit
5.19
presents
the
hours
lost
per
day
of
illness.
These
values
are
based
on
estimates
of
the
number
of
hours
worked,
adjusted
by
the
percent
of
the
population
engaged
in
paid
and
unpaid
work,
and
assumes
8
hours
of
sleep
per
day
per
person.
Details
of
the
sources,
calculations,
and
assumptions
to
derive
these
values
are
provided
in
Appendix
L.
5.

Exhibit
5.19
also
presents
the
per
hour
value
of
lost
time.
These
values
all
derive
from
reported
usual
weekly
earnings
of
$
576
(
U.
S.
Census
Bureau
Table
621,
2001).
To
value
lost
work
time,
this
figure
is
increased
to
reflect
employers'
costs
(
adding
in
benefits
paid).
Because
employers
are
willing
to
pay
for
workers'
time
at
this
level,
it
is
the
best
measure
of
the
value
of
that
lost
time
($
18.47
per
hour).
For
the
Enhanced
COI,
the
value
of
lost
unpaid
work
time
is
the
median
after
tax
wage.
To
derive
this
figure,
the
weekly
earnings
estimate
of
$
576
was
adjusted
downward
to
reflect
after
tax
wages
(
to
$
10.92).
For
the
Traditional
COI,
half
of
this
figure
is
used
(
or
$
5.46),
as
discussed
above,
and
in
detail
in
Appendices
K
and
L.
Details
of
the
sources,
calculations,
and
assumptions
to
derive
these
values
are
provided
in
Appendix
K
and
Appendix
L.
6.

Exhibit
5.19
also
multiplies
these
dollar­
per­
hour
values
by
the
time
allocations
to
determine
the
weighted
average
value
of
time
per
hour
and
per
day.

Exhibit
5.19
Value
of
Time,
2000
Time
Loss
Category
Hours
Per
Day
of
Illness
Per
Hour
Value
Per
Day
Value
(
weighted
by
time)

Enhanced
COI
Lost
Work
Time
3.505
$
18.47
$
64.75
Lost
Unpaid
Work
Time
2.183
$
10.92
$
23.83
Lost
Leisure
Time
10.312
$
10.92
$
112.56
Subtotal
16.000
Lost
Caregiver
Day
Sum
of
weighted
lost
paid
and
unpaid
work
and
leisure
days
$
201.14
Traditional
COI
Lost
Work
Day
3.505
$
18.47
$
64.75
Lost
Unpaid
Work
Day
2.183
$
5.4
$
11.91
Subtotal
5.688
Lost
Caregiver
Day
Sum
of
weighted
lost
paid
and
unpaid
work
days
$
76.66
Note:
Rounded
data
are
shown
here,
but
full
precision
was
used
in
all
calculations.
Source:
Exhibit
L.
9.
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
48
Total
Morbidity
Cost
of
Illness
There
are
two
major
components
of
the
total
value
of
the
morbidity
cost
of
avoiding
a
case
of
cryptosporidiosis
 
direct
medical
costs
and
lost
time.
As
discussed
above,
the
total
direct
medical
costs
are
$
93.82
per
illness.
Lost
time
estimates
are
derived
from
the
estimate
of
the
average
days
lost
and
the
value
of
each
day
lost.
For
the
Enhanced
COI,
productivity
losses
are
included.
Because
patients
are
only
fractionally
as
productive
at
work
as
well
people,
the
loss
associated
with
the
less
productive
days
is
a
portion
of
the
value
of
a
full
lost
day,
specifically
30
percent
(
rounded
from
Harrington
1991).
For
example,
this
may
be
the
result
of
frequent
trips
to
the
bathroom,
reduced
concentration
on
tasks,
or
preparation
of
special
meals.
Summing
the
subcategories
of
total
value
of
lost
time
yields
a
weighted
cost
of
$
651.07
(
Enhanced
COI)
and
$
150.80
(
Traditional
COI)
per
case
of
cryptosporidiosis
in
2000.
Exhibit
5.20
shows
these
calculations.
The
total
loss
per
case
in
2000$
is
$
744.89
for
the
Enhanced
COI
and
about
a
third
of
that,
or
$
244.62,
for
the
Traditional
COI.
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
49
Exhibit
5.20
Total
Loss
Per
Case,
Enhanced
and
Traditional
COI,
2000
Loss
Category
Average
Days
Lost
Per
Illness
Value
Per
Day
(
weighted
by
time)
Total
Loss
Per
Case
Enhanced
COI
Traditional
COI
Enhanced
COI
Traditional
COI
A
B
C
D=
A*
B
E=
A*
C
Direct
Medical
Costs
$
93.82
$
93.82
Lost
Paid
Work
Days
1.697
$
64.75
$
64.75
$
109.88
$
109.88
Lost
Unpaid
Work
Days
$
23.83
$
11.91
$
40.44
$
20.22
Lost
Leisure
Time
$
112.56
­
$
191.02
­

Lost
Caregiver
Days
0.270
$
201.14
$
76.66
$
54.31
$
20.70
Lost
Leisure
Productivity
4.233
$
112.56
x
30%
­
$
142.94
­

Lost
Productivity
at
Work
($
64.75
+
$
23.83)
x
30%
­
$
112.49
­

Lost
Time
Subtotal
$
651.07
$
150.80
Total
$
744.89
$
244.62
Notes:
Detail
may
not
calculate
to
totals
due
to
independent
rounding.
The
Traditional
COI
only
includes
valuation
for
medical
costs
and
lost
work
time
(
including
some
portion
of
unpaid
household
production).
The
Enhanced
COI
also
factors
in
valuations
for
lost
personal
time
(
non­
work
time)
such
as
child
care
and
homemaking
(
to
the
extent
not
covered
by
the
traditional
COI),
time
with
family,
and
recreation,
and
lost
productivity
at
work
on
days
when
workers
are
ill
but
go
to
work
anyway.
Source:
Exhibit
L.
10.

The
value
of
lost
time
can
increase
or
decrease
over
time,
depending
on
the
change
in
real
income.
Those
annual
changes
in
income
growth
and,
therefore,
the
value
of
time,
are
shown
in
Exhibit
5.21.
These
changes
(
mostly
increases)
in
income
growth
mean
that
the
loss
due
to
an
illness
would
increase
over
time
because
lost
time
is
recovered
by
wage
rates
or
their
equivalent.
In
the
benefits
model,
the
cases
avoided
in
each
year
are
valued
as
shown
in
Exhibit
5.21
(
the
model
uses
unrounded
data).
Benefits
derived
from
medical
costs
are
not
adjusted
for
changes
in
income
over
time,
because
medical
costs
do
not
necessarily
have
a
direct
or
indirect
link
with
income.
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
50
Exhibit
5.21
Yearly
Total
Loss
Per
Case,
Enhanced
and
Traditional
COI
Year
Annual
Percent
Change
in
Income
(
Real
GDP
per
Capita)
Lost
Time
Direct
Medical
Costs
Total
Loss
Per
Case
Enhanced
COI
Traditional
COI
Enhanced
COI
Traditiona
l
COI
A
B=(
1+
A)*
previous
year
C=(
1+
A)*
previous
year
D
E=
B+
D
F=
C+
D
2000
Base
Year
$
651.07
$
150.80
$
93.82
$
744.89
$
244.62
2001
0.1%
$
651.67
$
150.94
$
93.82
$
802.61
$
244.76
2002
­
0.1%
$
651.02
$
150.79
$
93.82
$
801.81
$
244.61
2003
3.2%
$
671.74
$
155.59
$
93.82
$
827.32
$
249.41
2004
2.4%
$
687.90
$
159.33
$
93.82
$
847.23
$
253.15
2005
2.4%
$
704.55
$
163.19
$
93.82
$
867.74
$
257.01
2006
2.4%
$
721.69
$
167.16
$
93.82
$
888.85
$
260.98
2007
2.4%
$
739.31
$
171.24
$
93.82
$
910.55
$
265.06
2008
2.3%
$
755.96
$
175.10
$
93.82
$
931.06
$
268.92
2009
2.3%
$
773.04
$
179.05
$
93.82
$
952.09
$
272.87
2010
2.3%
$
790.55
$
183.11
$
93.82
$
973.66
$
276.93
2011
2.3%
$
808.49
$
187.26
$
93.82
$
995.75
$
281.08
2012
2.3%
$
826.81
$
191.51
$
93.82
$
1,018.32
$
285.33
2013
2.3%
$
845.55
$
195.84
$
93.82
$
1,041.39
$
289.66
2014
2.3%
$
864.71
$
200.28
$
93.82
$
1,065.00
$
294.10
2015
2.3%
$
884.34
$
204.83
$
93.82
$
1,089.17
$
298.65
2016
2.3%
$
904.44
$
209.49
$
93.82
$
1,113.93
$
303.31
2017
2.3%
$
925.04
$
214.26
$
93.82
$
1,139.29
$
308.08
2018
2.3%
$
946.16
$
219.15
$
93.82
$
1,165.30
$
312.97
2019
2.3%
$
967.82
$
224.16
$
93.82
$
1,191.18
$
317.98
2020
2.3%
$
990.04
$
229.31
$
93.82
$
1,219.36
$
323.13
2021
2.3%
$
1,012.81
$
234.59
$
93.82
$
1,247.40
$
328.41
2022
2.3%
$
1,036.11
$
239.98
$
93.82
$
1,276.09
$
333.80
2023
2.3%
$
1,059.95
$
245.50
$
93.82
$
1,305.45
$
339.32
2024
2.3%
$
1,084.35
$
251.15
$
93.82
$
1,335.50
$
344.97
2025
2.3%
$
1,109.31
$
256.94
$
93.82
$
1,366.25
$
350.76
2026
2.3%
$
1,134.88
$
262.86
$
93.82
$
1,397.74
$
356.68
2027
2.3%
$
1,161.07
$
268.93
$
93.82
$
1,429.99
$
362.75
Note:
Full
precision
is
used
in
model
calculations.
Rounded
data
are
shown
here.
The
Traditional
COI
only
includes
valuation
for
medical
costs
and
lost
work
time
(
including
some
portion
of
unpaid
household
production).
The
Enhanced
COI
also
factors
in
valuations
for
lost
personal
time
(
non­
work
time)
such
as
child
care
and
homemaking
(
to
the
extent
not
covered
by
the
traditional
COI),
time
with
family,
and
recreation,
and
lost
productivity
at
work
on
days
when
workers
are
ill
but
go
to
work
anyway.
Source:
Exhibit
L.
1.

The
sensitivity
of
different
assumptions
on
the
Enhanced
COI
was
tested
and
is
described
in
Appendix
P.
Alternative
values
for
several
underlying
variables
were
considered,
but
only
the
following
two
were
judged
important
to
test
in
a
sensitivity
analysis:
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
51
°
The
hourly
value
of
nonwork
time
(
nonmarket
work
and
leisure
time).
The
Enhanced
COI
uses
a
value
of
$
10.92
per
hour
(
in
2000),
but
an
alternative
value
of
$
5.46
per
hour
was
used
in
the
"
low
estimate"
sensitivity
analysis,
and
a
value
of
$
16.37
was
used
in
the
"
high
estimate"
sensitivity
analysis.
These
bounds
represent
the
effect
of
(
1)
assuming
that
all
nonwork
time
is
valued
at
a
rate
higher
than
the
best
estimate
and
(
2)
assuming
that
some
of
the
nonwork
time
lost
is
an
incomplete
loss
of
utility.
The
basis
for
using
these
bounds
at
50
percent
and
150
percent
of
the
best
estimate
is
discussed
in
detail
in
Appendix
P.

°
The
percent
decrease
in
productivity
for
days
in
which
work
could
be
performed,
but
effects
of
the
illness
prevented
full
productivity.
The
Enhanced
COI
uses
30
percent,
and
the
bounds
for
the
sensitivity
analysis
are
20
percent
and
40
percent.
Related
studies
and
the
basis
for
selecting
these
levels
for
a
sensitivity
analysis
is
discussed
in
Appendix
P.

The
total
value
of
lost
time
using
the
Enhanced
COI
is
$
651.07
(
in
2000$),
and
using
these
alternative
values
in
the
derivation
of
that
estimate
would
lower
that
value
to
$
374.06
(
57
percent
of
the
Enhanced
COI)
and
raise
that
value
to
$
985.82
(
151
percent
of
the
Enhanced
COI).
The
overall
effect
on
total
benefits
is
less
pronounced
because
of
the
value
of
fatalities,
and
fixed
direct
medical
costs.
Appendix
P
calculates
the
effect
of
using
these
alternatives
on
the
total
COI
and
total
benefits.

5.3.1.2
Value
of
Avoiding
Fatal
Cases
of
Cryptosporidiosis
Benefits
of
the
LT2ESWTR
also
derive
from
avoiding
fatal
cases
of
cryptosporidiosis.
The
Value
of
a
Statistical
Life
(
VSL)
is
used
to
measure
the
value
of
these
benefits.
The
VSL
represents
an
estimate
of
the
monetary
value
of
reducing
risks
of
premature
death.
The
VSL,
therefore,
is
not
an
estimate
of
the
value
of
saving
a
particular
individual's
life.
The
value
of
a
"
statistical"
life
represents
the
sum
of
the
values
placed
on
small
individual
risk
reductions
across
an
exposed
population.
For
example,
if
a
regulation
were
to
reduce
the
risk
of
premature
death
from
cryptosporidiosis
by
1/
1,000,000
for
1
million
exposed
individuals,
the
regulation
would
"
save"
one
statistical
life
(
1,000,000
x
1/
1,000,000).
If
each
of
the
1,000,000
people
were
willing
to
pay
$
5
to
achieve
the
individual
risk
reduction
anticipated
from
the
regulation,
the
VSL
would
be
$
5
million
($
5
x
1,000,000).

An
EPA
study
characterized
the
range
of
possible
VSL
values
as
a
Weibull
distribution
with
a
mean
of
$
4.8
million
(
1990
price
level),
based
on
26
individual
study
estimates
(
USEPA
1997b).
This
represents
the
value
recommended
for
use
in
benefits
analyses
in
EPA's
Guidelines
for
Preparing
Economic
Analyses
(
USEPA
2000e)
and
endorsed
by
the
Science
Advisory
Board
(
SAB)
Arsenic
review
panel
(
USEPA
2001e).
For
purposes
of
the
LT2ESWTR
benefits
analysis,
the
VSL
Weibull
distribution
(
with
parameters
of
location
=
0,
scale
=
5.32,
shape
=
1.51)
was
incorporated
into
the
benefits
model
Monte
Carlo
simulation.
This
enables
quantification
of
the
uncertainty
surrounding
benefits
estimates
derived
from
the
VSL.
The
VSL
was
also
updated
to
a
year
2000
price
level
using
a
CPI
adjustment
factor
(
see
Appendix
P)
and
the
distribution
in
2000
has
a
mean
of
$
6.3
million,
median
of
$
5.5
million,
a
5th
percentile
value
of
$
1.0
million
and
a
95th
percentile
value
of
$
14.5
million.

5.3.1.3
Measuring
Benefits
Over
the
LT2ESWTR
Implementation
Schedule
In
order
to
extract
benefits
data
from
the
model
and
present
these
benefits
in
comparable
terms
to
a
similarly
calculated
stream
of
costs,
it
is
necessary
to
calculate
the
present
value
of
all
benefits
over
14
Ideally,
income
elasticity
and
income
growth
measurements
would
be
calculated
using
real
per
capita
personal
income
growth.
However,
real
per
capita
GDP
is
used
as
a
proxy
for
real
per
capita
personal
income
growth
due
to
lack
of
appropriate
data
projections
for
real
personal
income
growth.
Historical
data
suggests
that
GDP
and
personal
income
grow
at
similar
rates
(
i.
e.,
Table
B­
31
of
the
2002
Economic
Report
of
the
President
shows
that
both
real
per
capita
GDP
and
real
per
capita
disposable
personal
income
grew
at
an
average
annual
rate
of
2.3
percent
between
1959
and
2000).

15
See
Appendix
A
of
Kleckner
and
Neumann
(
2000)
for
additional
information
on
the
derivation
and
application
of
income
elasticity
adjustments.

Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
52
the
lifetime
of
the
implementation
schedule.
LT2ESWTR
implementation
occurs
over
several
years
as
States
and
PWSs
learn
the
requirements,
inform
their
staffs,
and
perform
monitoring.
A
25­
year
horizon
was
chosen
for
this
analysis
because
systems
have
several
years
to
begin
treatment
associated
with
LT2ESWTR,
and
many
technologies
in
this
analysis
have
a
20­
year
life­
cycle.
Calculations
using
this
time
frame
allows
the
analysis
to
capture
all
of
the
period
when
technologies
would
be
installed
and
avoids
the
complications
that
would
be
necessary
to
estimate
rehabilitation
or
replacement
costs
for
installed
equipment.
This
time
frame
also
matches
that
used
in
other
recent
analyses
such
as
the
one
for
the
Stage
2
DBPR.
A
complete
schedule
of
when
costs
are
expected
to
be
incurred
and
benefits
obtained
is
presented
in
Appendix
O.

5.3.1.4
Adjustment
for
Income
Elasticity
Although
the
price
level
(
year
2000)
is
held
constant
across
all
benefits
projections,
values
in
future
years
are
adjusted
to
reflect
changes
in
the
valuation
of
avoiding
health
effects
associated
with
changes
in
income
over
time.
Estimates
of
how
valuation
varies
with
income
growth
(
i.
e,
income
elasticities)
are
available
from
the
economic
literature,
and
in
those
cases
income
elasticities
are
combined
with
estimates
of
income
growth.
Benefits
based
on
potentially
fatal
health
effects,
which
are
based
on
willingness
to
pay
estimates
that
vary
with
income,
are
adjusted
using
estimates
of
income
elasticity
and
income
growth.
This
section
describes
how
this
adjustment
is
carried
out.

In
the
case
of
avoided­
death
benefits,
income
elasticity
adjustments
are
applied
to
values
in
future
years.
In
general,
income
elasticity
represents
changes
in
valuation
in
relation
to
changes
in
real
income.
For
example,
if,
for
every
1
percent
increase
in
real
income,
a
particular
consumer's
willingness
to
pay
for
a
particular
item
increases
by
1
percent,
this
would
be
represented
by
an
income
elasticity
of
1.
For
most
willingness
to
pay
estimates,
income
elasticity
values
are
less
than
1,
reflecting
slower
growth
in
willingness
to
pay
than
in
income.

In
order
to
apply
the
income
elasticity
values
in
the
benefits
model,
they
must
be
combined
with
projections
of
real
income
growth
over
the
time
frame
for
analysis.
To
accomplish
this,
population
and
real
gross
domestic
product
(
GDP)
projections
are
combined
to
calculate
per­
capita
real
GDP
values14
(
see
Exhibit
C.
12).
Percent
changes
in
these
values
over
time
can
then
be
combined
with
income
elasticity
figures
to
derive
a
single
adjustment
factor.
15
Given
any
two
points
in
time,
this
factor
is
calculated
as
follows:

Income
elasticity
adjustment
factor
=
(
eI
1
­
eI
2
­
I
2
­
I
1)
/
(
eI
2
­
eI
1
­
I
2
­
I
1)
16
The
distribution
of
VSL
values
used
in
this
EA
is
derived
based
on
a
meta­
analysis
of
26
different
VSL
studies,
all
representing
different
year
price
levels.
These
price
levels
were
updated
to
a
common
1990
price
level
as
part
of
the
analysis
in
"
The
Benefits
and
Costs
of
the
Clean
Air
Act,
1970­
1990"
(
USEPA
1997b),
from
which
the
distribution
used
in
this
EA
is
taken.

Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
53
where:
e
=
income
elasticity
I
1
=
real
income
(
per­
capita
GDP)
in
the
base
year
I
2
=
real
income
(
per­
capita
GDP)
in
the
year
of
analysis
When
applying
this
formula,
income
elasticity
adjustment
factors
are
calculated
from
the
same
base
year
as
the
values
subject
to
adjustment.
In
this
case,
income
elasticity
factors
for
fatal
cryptosporidiosis
cases
are
calculated
from
a
1990
base
year
(
I
1
=
1990
in
the
above
formula)
because
that
is
the
base
year
used
in
the
study
from
which
VSL
estimates
are
derived.
16
Kleckner
and
Neumann
(
2000)
identified
published
studies
from
which
elasticity
values
could
be
derived
for
potentially
fatal
health
effects.
They
suggest
a
triangular
distribution
with
a
mode
of
0.40,
and
endpoints
at
0.08
and
1.00.
In
the
Monte
Carlo
simulation
that
assigns
dollar
values
to
benefits,
income
elasticity
values
(
e
in
the
above
equation)
are
drawn
from
this
probability
distribution.
Based
on
this
formula
and
inputs,
income
elasticity
factors
are
computed
and
applied
to
avoided­
death
benefits
in
future
years.
At
the
average
income
elasticity
value
(
0.49),
the
income
elasticity
factors
applied
range
from
1.185
(
2008)
to
1.460
(
2027).
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
54
Exhibit
5.22
Mean
of
Yearly
Values
for
a
Statistical
Life
($
Million,
rounded)

Year
Mean
of
10,000
Values
2008
$
7.57
2009
$
7.66
2010
$
7.74
2011
$
7.83
2012
$
7.92
2013
$
8.01
2014
$
8.10
2015
$
8.19
2016
$
8.28
2017
$
8.37
2018
$
8.46
2019
$
8.56
2020
$
8.65
2021
$
8.75
2022
$
8.85
2023
$
8.95
2024
$
9.05
2025
$
9.15
2026
$
9.25
2027
$
9.35
Source:
Exhibit
C.
13
The
estimates
for
the
value
of
a
statistical
life
derive
from
a
distribution
of
the
value
of
statistical
life
(
discussed
in
section
5.3.1.2),
and
adjustments
for
income
elasticity
(
discussed
in
this
section).
In
the
benefits
model,
each
year
from
2008
to
2027
has
a
vector
of
10,000
values
for
a
statistical
life,
and
each
value
is
used
exactly
once.
The
complete
distributions
(
200,000
values)
are
documented
in
the
model,
but
to
illustrate
how
these
distributions
are
affected
by
the
income
elasticity
adjustments,
the
mean
of
each
year's
distribution
of
values
is
shown
above
in
Exhibit
5.22.
The
values
are
shown
starting
in
the
year
2008
because
that
is
the
first
year
with
benefits.
Appendix
C
has
additional
discussion
of
the
derivation
of
these
data.

5.3.1.5
Present
Value
of
Future
Benefits
To
allow
comparison
of
future
streams
of
costs
and
benefits,
it
is
common
practice
to
adjust
both
streams
to
a
present
value
(
PV)
using
a
social
discount
rate.
This
process
takes
into
account
the
time
preference
that
society
places
on
expenditures
and
benefits
and
allows
comparison
of
cost
and
benefit
17
See
EPA's
Guidelines
for
Preparing
Economic
Analyses
(
USEPA
2000e)
for
a
full
discussion
of
the
use
of
social
discount
rates
in
the
evaluation
of
policy
decisions.

Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
55
streams
that
vary
over
a
given
time
period.
17
A
present
value
for
any
future
period
can
be
calculated
using
the
following
equation:

PV
=
V(
t)
/
(
1
+
R)
t
Where:
t
=
The
number
of
years
from
the
reference
period
(
year
0
of
the
benefits
stream)
R
=
Social
discount
rate
V(
t)
=
The
benefits
occurring
t
years
from
the
reference
period
The
present
values
presented
in
this
EA
are
the
sum
of
the
PVs
for
each
year.

There
is
much
discussion
among
economists
of
the
proper
social
discount
rate
to
use
for
policy
analysis.
Therefore,
for
this
EA,
PV
calculations
are
made
using
two
social
discount
rates
thought
to
best
represent
current
policy
evaluation
methodologies,
3
and
7
percent.
The
rate
of
3
percent
is
based
on
rates
of
return
on
relatively
risk­
free
investments,
as
described
in
the
Guidelines
for
Preparing
Economic
Analyses
(
USEPA
2000e).
The
rate
of
7
percent
is
a
recommendation
of
OMB
as
an
estimate
of
"
before­
tax
rate
of
return
to
incremental
private
investment"
(
USEPA
1996b).
To
present
results
on
an
annual
basis,
the
total
PV
of
benefits
are
annualized
using
the
same
social
discount
rates.

5.3.1.6
Summary
of
Quantified
Benefits
of
LT2ESWTR
The
risk
assessment
methodology
described
in
this
chapter
estimates
the
quantified
benefits
of
reducing
endemic
cryptosporidiosis
as
applied
to
each
of
the
regulatory
alternatives
considered
for
LT2ESWTR.
These
alternatives
 
which
are
described
in
detail
in
Chapter
3
 
were
evaluated
to
provide
EPA
with
information
on
different
approaches
for
implementation
of
these
regulations.
Exhibits
5.23
and
5.24
provide
summaries
of
the
cumulative
monetary
benefits
estimated
for
the
Preferred
Alternative
using
three
occurrence
distribution
data
sets.
Exhibit
5.23
presents
benefits
calculated
using
the
Enhanced
COI
and
Exhibit
5.24
presents
benefits
calculated
using
the
Traditional
COI.
These
benefit
estimates
were
compared
to
costs
used
by
EPA
as
one
criterion
(
of
many)
to
decide
upon
a
preferred
rule
alternative.
Costs
for
rule
alternatives
are
presented
in
Chapter
6,
and
cost/
benefit
comparisons
are
evaluated
in
Chapter
8.
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
56
Serving
<
10,000
Serving
>
10,000
All
Systems
Lower
(
5th
%
ile)
Upper
(
95th
%
ile)
A
B
C
D
E
3%
Discount
Rate
Illnesses
Avoided
ICR
36
$
648
$
684
$
114
$
1,564
$
ICRSSL
7
$
165
$
172
$
30
$
376
$
ICRSSM
17
$
318
$
335
$
57
$
789
$
Deaths
Avoided
ICR
32
$
729
$
761
$
51
$
2,340
$
ICRSSL
7
$
195
$
202
$
14
$
613
$
ICRSSM
15
$
366
$
381
$
25
$
1,179
$
7%
Discount
Rate
Illnesses
Avoided
ICR
29
$
550
$
579
$
96
$
1,323
$
ICRSSL
6
$
140
$
146
$
26
$
319
$
ICRSSM
13
$
270
$
283
$
48
$
668
$
Deaths
Avoided
ICR
26
$
624
$
651
$
44
$
2,001
$
ICRSSL
6
$
167
$
172
$
12
$
524
$
ICRSSM
12
$
313
$
325
$
22
$
1,010
$
90%
Confidence
Bound
(
All
Systems)

Data
Set
Mean
Exhibit
5.23
Annualized
Benefits
of
Illnesses
and
Deaths
Avoided,
Preferred
Alternative,
Enhanced
Cost
of
Illness
[
1]

($
Millions/
Year,
2000$)

Notes:
[
1]
The
traditional
COI
only
includes
valuation
for
medical
costs
and
lost
work
time
(
including
some
portion
of
unpaid
household
production).
The
enhanced
COI
also
factors
in
valuations
for
lost
personal
time
(
non­
work
time)
such
as
child
care
and
homemaking
(
to
the
extent
not
covered
by
the
traditional
COI),
time
with
family,
and
recreation,
and
lost
productivity
at
work
on
days
when
workers
are
ill
but
go
to
work
anyway.
Sources:
[
A]
Appendix
C,
Exhibits
C.
4a
and
C.
5a,
Columns
G
and
J,
Row
­
Small
Systems,
A3,
ICR,
ICRSSL,
and
ICRSSM.
[
B]
Appendix
C,
Exhibits
C.
4a
and
C.
5a,
Columns
G
and
J,
Row
­
Large
Systems,
A3,
ICR,
ICRSSL,
and
ICRSSM.
[
C]
Appendix
C,
Exhibit
C.
4a,
Columns
J
and
M,
Row
­
All
Systems,
A3,
ICR,
ICRSSL,
and
ICRSSM.
[
D]
Appendix
C,
Exhibit
C.
4a,
Columns
K
and
N,
Row
­
All
Systems,
A3,
ICR,
ICRSSL,
and
ICRSSM.
[
E]
Appendix
C,
Exhibit
C.
4a,
Columns
O
and
P,
Row
­
All
Systems,
A3,
ICR,
ICRSSL,
and
ICRSSM.
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
57
Serving
<
10,000
Serving
>
10,000
All
Systems
Lower
(
5th
%
ile)
Upper
(
95th
%
ile)
A
B
C
D
E
3%
Discount
Rate
Illnesses
Avoided
ICR
11
$
195
$
206
$
34
$
470
$
ICRSSL
2
$
50
$
52
$
9
$
113
$
ICRSSM
5
$
96
$
101
$
17
$
237
$
Deaths
Avoided
ICR
32
$
729
$
761
$
51
$
2,340
$
ICRSSL
7
$
175
$
202
$
14
$
613
$
ICRSSM
15
$
366
$
381
$
25
$
1,179
$
7%
Discount
Rate
Illnesses
Avoided
ICR
9
$
166
$
175
$
29
$
400
$
ICRSSL
2
$
42
$
44
$
8
$
96
$
ICRSSM
4
$
82
$
97
$
16
$
234
$
Deaths
Avoided
ICR
26
$
624
$
651
$
44
$
2,001
$
ICRSSL
6
$
167
$
172
$
12
$
524
$
ICRSSM
12
$
313
$
325
$
22
$
1,010
$
Mean
Data
Set
90%
Confidence
Bound
(
All
Systems)
Exhibit
5.24
Annualized
Benefits
of
Illnesses
and
Deaths
Avoided,
Preferred
Alternative,
Traditional
Cost
of
Illness
[
1]

($
Millions/
Year,
2000$)

Notes:
[
1]
The
traditional
COI
only
includes
valuation
for
medical
costs
and
lost
work
time
(
including
some
portion
of
unpaid
household
production).
The
enhanced
COI
also
factors
in
valuations
for
lost
personal
time
(
non­
work
time)
such
as
child
care
and
homemaking
(
to
the
extent
not
covered
by
the
traditional
COI),
time
with
family,
and
recreation,
and
lost
productivity
at
work
on
days
when
workers
are
ill
but
go
to
work
anyway.
Sources:
[
A]
Appendix
C,
Exhibits
C.
4b
and
C.
5b,
Columns
G
and
J,
Row
­
Small
Systems,
A3,
ICR,
ICRSSL,
and
ICRSSM.
[
B]
Appendix
C,
Exhibits
C.
4b
and
C.
5b,
Columns
G
and
J,
Row
­
Large
Systems,
A3,
ICR,
ICRSSL,
and
ICRSSM.
[
C]
Appendix
C,
Exhibit
C.
4b,
Columns
J
and
M,
Row
­
All
Systems,
A3,
ICR,
ICRSSL,
and
ICRSSM.
[
D]
Appendix
C,
Exhibit
C.
4b,
Columns
K
and
N,
Row
­
All
Systems,
A3,
ICR,
ICRSSL,
and
ICRSSM.
[
E]
Appendix
C,
Exhibit
C.
4b,
Columns
O
and
P,
Row
­
All
Systems,
A3,
ICR,
ICRSSL,
and
ICRSSM.
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
58
5.3.2
Monetization
of
Benefits
to
Sensitive
Subpopulations
The
infectivity
estimates
used
in
this
analysis
are
derived
from
clinical
studies
performed
on
healthy
adults
and
are
applied
to
all
populations.
No
separate
monetization
of
benefits
for
sensitive
subpopulations
was
therefore
performed.
The
morbidity
estimates
are
based
on
a
large
population
 
those
affected
by
the
Milwaukee
outbreak
 
that
included
a
mix
of
the
general
population
and
sensitive
subpopulations.
By
using
morbidity
factors
from
that
outbreak,
the
monetization
of
benefits
to
sensitive
subpopulations
is
included,
but
could
not
be
separately
itemized.
The
mortality
rates
used
in
this
study
are
derived
from
data
from
the
Milwaukee
outbreak,
where
46
of
the
54
deaths
were
persons
with
AIDS;
the
other
fatalities
were
elderly
and
some
had
other
illnesses.
For
normally
healthy
adults,
cryptosporidiosis
is
not
considered
a
fatal
disease.
Therefore,
all
the
mortality
benefits
estimated
for
this
rule
are
deaths
avoided
within
sensitive
subpopulations.
Avoided
illnesses
and
deaths
were
not
separately
quantified
for
children,
so
monetization
of
these
benefits
is
not
shown
separately.
Further
discussion
of
the
impact
of
the
rule
on
sensitive
populations
is
in
Chapter
7,
section
7.9.

5.4
Other
Benefits
of
LT2ESWTR
Provisions
Section
5.4
describes
qualitative
benefits
of
the
LT2ESWTR
provisions.
Although
sufficient
information
was
not
available
to
quantify
these
benefits
of
LT2ESWTR
implementation,
the
benefits
 
in
terms
of
both
health
and
monetary
value
 
are
thought
to
be
significant.

5.4.1
Reduction
in
Outbreak
Risk
Besides
reducing
the
endemic
risk
of
cryptosporidiosis,
the
LT2ESWTR
will
reduce
the
likelihood
of
major
outbreaks,
such
as
occurred
in
Milwaukee.
The
economic
value
of
reducing
the
risk
of
outbreaks
could
be
quite
high
when
the
magnitude
of
potential
costs
is
considered.
For
example,
if
the
$
745
COI
per
cryptosporidiosis
infection
estimate
is
applied
to
the
estimated
2,000
cases
attributed
to
a
sewage­
contaminated
well
in
Braun
Station,
Texas
(
Craun
et
al.
1998),
health
damages
could
reach
$
1.5
million.
Other
costs
associated
with
outbreaks
include
spending
by
Local,
State,
and
national
public
health
agencies;
emergency
corrective
actions
by
utilities;
and
possible
legal
costs
if
liability
is
a
factor.
Affected
water
systems
and
local
governments
may
incur
costs
through
provision
of
alternative
water
supplies
and
issuing
customer
water
use
warnings
and
health
alerts.
Commercial
establishments
(
e.
g.,
restaurants)
and
their
customers
may
incur
costs
due
to
interrupted
and
lost
service.
Local
businesses,
institutions,
and
households
may
incur
costs
associated
with
undertaking
averting
and
defensive
actions.
To
the
extent
that
LT2ESWTR
reduces
the
likelihood
of
waterborne
disease
outbreaks,
avoided
response
costs
are
potentially
numerous
and
significant.

During
outbreaks,
consumers
and
businesses
may
use
alternative
water
sources
or
may
adopt
behaviors
to
reduce
risk,
such
as
boiling
water.
If
the
rule
reduces
the
need
for
these
risk­
averting
behaviors,
an
economic
benefit
will
accrue.
To
give
a
source
of
the
possible
scale,
during
an
outbreak
of
giardiasis,
a
disease
with
gastrointestinal
symptoms
similar
to
cryptosporidiosis,
expenditures
on
averting
behaviors,
such
as
hauling
in
safe
water,
boiling
water,
and
purchasing
bottled
water,
were
estimated
(
in
2000$)
between
$
1.74
to
$
5.53
per
person
per
day
during
the
Milwaukee
outbreak
(
Harrington
et
al.
1991).
If
these
figures
are
applied
to
a
small
drinking
water
system
serving
10,000
customers,
total
expenditures
on
risk­
averting
behavior
during
a
Cryptosporidium
outbreak
could
range
between
$
17,400
and
$
55,300
per
day
(
2000$).
Determining
the
reduction
in
outbreak
risk
and
the
resulting
benefits
from
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
59
avoiding
averting
behaviors
is
not
possible
given
current
information,
but
potential
benefits
are
expected
to
be
substantial.

Five
studies
were
identified
that
used
the
averting
cost
approach
to
estimate
household
and
other
costs
attributable
to
short­
term
contamination
of
drinking
water
supplies
(
Abdalla
1990;
Abdalla
et
al.
1992;
Harrington
et
al.
1991;
Sun
et
al.
1992;
Van
Houtven
et
al.
1997).
The
most
relevant
of
these
for
the
LT2ESWTR
analysis
is
a
study
by
Harrington
et
al.
(
1991),
that
analyzes
the
costs
associated
with
drinking
water
contamination
by
Giardia
in
Luzerne
County,
Pennsylvania.
The
December
1983
outbreak
resulted
in
366
confirmed
giardiasis
cases
resulting
from
sewage
leaking
into
the
unfiltered
source
water.
The
total
affected
population
was
75,000
individuals
across
Pittston
Borough
and
17
other
municipalities.
The
Harrington
study
also
developed
a
theoretical
and
empirical
example
of
how
outbreak
costs
are
incurred,
based
on
the
Luzerne
County
example.

The
four
steps
associated
with
a
waterborne
outbreak
that
may
impose
costs
on
society
are
discovery,
survey
and
testing,
reaction,
and
aftermath.
(
Harrington
et
al.
1991).
These
are
described
below:

°
Discovery.
Health
care
providers
or
State,
Local,
or
hospital
laboratory
technicians
send
reports
to
State
authorities
notifying
them
of
the
need
for
further
investigation
when
the
rate
of
new
cases
suddenly
increases
above
the
normal
rate.

°
Survey
and
testing.
Epidemiological
surveys
may
be
conducted,
along
with
tests
of
the
water
supply,
once
a
few
cases
are
confirmed.

°
Reaction.
Local
authorities
and
the
water
system
may
issue
boil­
water
advisories
or
other
warnings
to
reduce
exposure
once
a
link
is
made
between
the
drinking
water
supply
and
the
disease
outbreak.
Businesses,
as
well
as
households,
may
be
affected
by
such
action,
requiring
government
agencies
to
begin
surveillance
and
enforcement
activities
and,
in
some
cases,
provide
alternative
water
sources.

°
Aftermath.
Long­
term
solutions
to
the
problem
are
discussed,
as
well
as
how
the
costs
of
the
outbreak
and
prevention
of
future
ones
may
be
shared.
These
discussions
can
only
take
place
once
the
outbreak
is
contained
by
actions
taken
during
the
previous
phase.

In
the
Luzerne
County
outbreak,
individuals
took
actions
to
avoid
exposure
to
the
contaminated
water
and
those
actions
resulted
in
estimated
losses
between
$
20.8
million
and
$
61.8
million
(
2000$).
The
predominant
cost
was
due
to
the
need
to
boil
water
and
the
associated
time
lost.
Losses
due
to
averting
actions
for
restaurants,
bars,
schools
and
other
businesses
during
the
outbreak
exceeded
$
1.0
million.
The
burden
for
government
agencies
was
$
230,000
and
the
outbreak
cost
the
water
supply
utility
$
1.8
million.
These
costs
are
in
year
2000$
and
do
not
include
legal
fees,
adverse
effects
on
businesses
(
which
were
not
investigated),
leisure
activities,
or
net
losses
due
to
substituting
more
expensive
beverages
for
tap
water.
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
60
5.4.2
Costs
to
Households
to
Avert
Infection
In
addition
to
averting
actions
taken
with
regard
to
outbreaks,
a
reduction
of
everyday
averting
behaviors
can
be
achieved.
Many
households
may
undertake
on
a
daily
basis
the
same
averting
actions
that
they
take
during
an
outbreak
(
e.
g.,
buying
bottled
water,
boiling
water,
installing
point­
of­
use
(
POU)
filtration).
To
the
extent
that
the
LT2ESWTR
can
be
expected
to
reduce
a
household's
perceptions
of
the
health
risks
associated
with
drinking
water,
regulatory
action
may
reduce
the
frequency
of
such
averting
actions
and
their
costs.

5.4.3
Enhanced
Aesthetic
Water
Quality
Some
treatment
improvements
resulting
from
the
implementation
of
the
rule
are
likely
to
improve
the
aesthetic
quality
of
the
drinking
water.
Consumers,
presumably,
would
be
willing
to
pay
to
protect
the
aesthetic
quality
of
drinking
water,
and
therefore,
these
benefits
should
result
in
an
economic
benefit.
However,
the
benefits
from
such
water
quality
improvements
due
to
the
rule
are
not
quantified
for
this
analysis.

5.4.4
Risk
Reduction
from
Co­
occurring
and
Emerging
Pathogens
While
the
benefits
analysis
for
the
LT2ESWTR
only
includes
reductions
in
illness
and
mortality
attributable
to
Cryptosporidium,
the
rule
is
expected
to
reduce
exposure
to
other
pathogens
(
e.
g.,
Giardia
or
other
waterborne
bacterial
or
viral
pathogens
such
as
Cyclospora
and
Microsporidium).
For
example,
membrane
processes
that
remove
Cryptosporidium
are
also
shown
to
achieve
equivalent
log
removal
of
Giardia
under
worst­
case
and
normal
operating
conditions,
and
nanofiltration
also
shows
similar
removal
of
Giardia
as
Cryptosporidium
(
USEPA
2003c).
Goodrich
and
Lykins
(
1995)
evaluated
bag
filters
and
concluded
that
any
microbe
or
object
greater
than
4.5
microns
in
size
would
be
subject
to
0.5
to
2.0
log
removal.
Strengthened
regulatory
requirements
will
translate
into
increased
removal
of
additional
pathogens
and
a
resulting
reduction
in
risk.
This
may
prove
valuable
in
reducing
overall
risk
because
the
impact
of
emerging
pathogens,
although
not
well
established,
could
be
significant.

5.4.5
Benefits
from
Other
Rule
Provisions
The
benefit
estimates
discussed
in
this
chapter
result
from
increased
treatment
requirements
that
improve
the
consumers'
water
quality.
However,
other
provisions
of
the
LT2ESWTR
not
directly
involving
changes
to
treatment
practices
will
also
provide
benefits
to
water
consumers.
Due
to
data
constraints,
EPA
was
not
able
to
quantify
these
benefits.
Instead,
a
qualitative
discussion
of
these
benefits
is
provided
below.

Benefits
of
Source
Water
Monitoring
While
source
water
monitoring
does
not
provide
any
direct
monetary
benefits,
the
information
gained
from
turbidity,
Cryptosporidium,
and
E.
coli
testing
may
provide
benefits
to
the
water
systems
and
ultimately
to
their
customers.
Although
some
large
systems
currently
monitor
their
source
water
for
these
contaminants,
many
do
not.
Most
small
systems
do
not
monitor
source
water.
Monitoring
allows
systems
to
better
understand
variations
in
their
source
water
and
to
adjust
their
operations
accordingly.
For
example,
if
a
system
discovers
that
pathogen
levels
are
elevated
in
the
spring,
they
could
plan
to
add
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
61
more
coagulant
or
bring
another
sedimentation
basin
online
during
that
period.
Systems
that
find
little
or
no
Cryptosporidium
will
be
able
to
boost
consumer
confidence
in
their
water,
providing
benefits
through
fewer
home
treatment
devices
and
less
time
spent
in
dealing
with
customer
complaints.
Systems
that
detect
Cryptosporidium
can
use
that
information
for
public
education
about
source
water
protection
and
watershed
management.
These
can
help
bring
about
changes
in
watershed
protection
that
will
ultimately
result
in
better
source
water
quality.
Improved
source
water
quality
can
produce
cost
savings
for
treatment.

Benefits
of
Covered
Finished
Water
Reservoirs
The
quality
of
water
in
finished
water
reservoirs
is
subject
to
similar
environmental
influences
as
surface
water,
including
deposition
of
airborne
chemicals,
surface
water
runoff,
animal
carcasses,
animal
or
bird
droppings,
and
growth
of
algae
and
other
aquatic
organisms.
In
one
study,
gulls
contaminated
a
10
million
gallon
reservoir
and
increased
bacteriological
growth;
in
another,
waterfowl
were
found
to
elevate
coliform
levels
in
small
recreational
lakes
by
20
times
their
normal
levels
(
Morra
1979).
Algal
growth
increases
the
biomass
in
the
reservoir,
which
reduces
dissolved
oxygen
and
thereby
increases
the
release
of
iron,
manganese,
and
nutrients
from
the
sediments.
This,
in
turn,
supports
more
algal
growth
(
Cooke
and
Carlson
1989).
Algae
can
cause
taste
and
odor
problems.
Further,
uncovered
finished
water
reservoirs
may
be
subject
to
contamination
by
illegal
swimming
and
dumping.
Documented
water
quality
problems
in
open
finished
water
reservoirs
include
increased
algal
cells;
heterotrophic
plate
count
(
HPC)
bacteria;
turbidity;
color;
particle
counts;
biomass;
and
decreased
chlorine
residuals
(
Pluntze
1974;
AWWA
1983;
Silverman
et
al.
1983;
LeChevallier
et
al.
1997b).

Finished
water
is
usually
not
treated
or
tested
again
prior
to
consumption,
so
any
contamination
in
the
uncovered
reservoir
may
be
passed
directly
to
the
customer.
Therefore,
requirements
to
cover
all
finished
water
reservoirs
or
to
treat
the
effluent,
unless
other
adequate
risk
management
measures
can
be
demonstrated,
will
reduce
the
risk
of
contamination
and
result
in
positive
health
benefits.
Covering
reservoirs
or
providing
additional
treatment
of
finished
water
will
also
provide
some
additional
protection
from
possible
acts
of
terrorism.
Data
are
not
available,
however,
to
quantify
the
benefits
associated
with
covering
all
finished
water
reservoirs.

5.4.6
Summary
of
Nonquantified
Benefits
As
explained
above,
several
types
of
potential
benefits
were
not
included
in
the
quantitative
analysis.
Exhibit
5.25
shows
how
the
rule
provisions
that
have
not
been
quantified
would
be
expected
to
affect
the
overall
benefits
derived
from
LT2ESWTR.
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
62
Exhibit
5.25
Summary
of
Nonquantified
Benefits
Benefit
Type
Potential
Effect
on
Benefits
Comments
Reducing
outbreak
risks
and
response
costs
Increase
Some
outbreaks
are
caused
by
human
or
equipment
failures
that
may
occur
even
with
the
proposed
new
requirements;
however,
by
adding
barriers
of
protection
for
some
systems,
the
rule
will
reduce
the
possibility
of
such
failures
leading
to
outbreaks.

Reducing
averting
behavior
(
e.
g.,
boiling
tap
water
or
purchasing
bottled
water)
Increase
/
No
Change
Averting
behavior
is
associated
with
both
out­
ofpocket
costs
(
e.
g.,
purchase
of
bottled
water)
and
opportunity
costs
(
e.
g.,
time
required
to
boil
water)
to
the
consumer.
Reductions
in
averting
behavior
are
expected
to
have
a
positive
impact
on
benefits
from
the
rule.

Improving
aesthetic
water
quality
Increase
Some
technologies
installed
for
this
rule
(
e.
g.,
ozone)
are
likely
to
reduce
taste
and
odor
problems.

Reducing
risk
from
co­
occurring
and
emerging
pathogens
Increase
Although
focused
on
removal
of
Cryptosporidium
from
drinking
water,
systems
that
change
treatment
processes
will
also
increase
removal
of
pathogens
that
the
rule
does
not
specifically
regulate.
Additional
benefits
will
accrue.

Increased
source
water
monitoring
Increase
The
greater
understanding
of
source
water
quality
that
results
from
monitoring
may
enhance
the
ability
of
plants
to
optimize
treatment
operations
in
ways
other
than
those
addressed
in
this
rule.

Covering
all
finished
water
reservoirs
Increase
Although
insufficient
data
were
available
to
quantify
benefits,
the
reduction
of
contaminants
introduced
through
uncovered
finished
water
reservoirs
would
produce
positive
public
health
benefits.

5.5
Summary
of
Uncertainties
As
described
in
previous
sections
within
this
chapter,
many
sources
of
uncertainty
in
the
benefits
analysis
were
incorporated
in
the
risk
assessment
model.
Exhibit
5.26
summarizes
those
uncertainties
and
describes
how
they
may
affect
the
benefit
analysis.
It
is
not
known
whether
most
of
these
uncertainties
would
lead
to
an
underestimation
or
over­
estimation
of
benefits.

5.6
Comparison
of
Regulatory
Alternatives
Although
the
models
were
run
separately
for
combinations
of
PWS
type
and
size
and
different
Cryptosporidium
occurrence
data
sets,
all
four
regulatory
alternatives
(
and
the
baseline)
were
always
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
63
computed
within
a
given
model
run.
This
method
effectively
"
blocked"
the
variability
and
uncertainty
from
other
sources
so
that
direct,
more
precise
comparisons
could
be
made
among
all
the
regulatory
alternatives
in
a
given
simulation.

The
five
modeled
regulatory
conditions
are
(
see
Chapter
3
for
descriptions
of
these):

°
Pre­
LT2
Baseline
(
Regulatory
Alternative
A0)

°
Regulatory
Alternative
A1
°
Regulatory
Alternative
A2
°
Regulatory
Alternative
A3
(
the
Preferred
Regulatory
Alternative)

°
Regulatory
Alternative
A4
Exhibit
5.27
summarizes
the
estimated
number
of
illnesses
and
deaths
avoided
by
filtered
and
unfiltered
systems
for
each
regulatory
alternative.
Exhibit
5.28
follows
with
the
monetization
by
regulatory
alternative.
Quantified
benefits
do
not
vary
substantially
among
alternatives
for
two
reasons.
First,
roughly
half
of
the
benefits
are
attributed
to
the
unfiltered
systems
and
the
requirements
for
those
systems
are
the
same
for
each
alternative.
Second,
UV
is
the
least
expensive
technology
for
most
systems
(
thus
the
most
selected
technology)
and
provides
3
log
treatment
regardless
of
the
treatment
requirements.
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
64
Exhibit
5.26
Summary
of
Uncertainties
Affecting
LT2ESWTR
Benefits
Estimates
Uncertainty
Section
with
Discussion
of
Uncertainty
Effect
on
Benefits
Estimates
Underestimate
Overestimate
Under
or
Over
Estimate
Quantifying
only
cases
of
endemic
illness
of
cryptosporidiosis
5.2.3
X
Infectivity
for
C.
parvum
estimated
from
three
known
isolates
5.2.3
X
Morbidity
based
on
triangular
distribution
5.2.3
X
Mortality
based
primarily
on
deaths
of
patients
with
AIDS
5.2.3
X
Source
water
concentrations
estimated
using
three
data
sets,
calculation
of
central
tendencies
and
bounds
5.2.4.1
X
Proportion
of
measured
oocysts
that
were
infectious,
represented
by
triangular
distribution
5.2.4.1
X
Binning
assignments
4.5.6,
Appendix
B
X
Estimate
of
plant
implementation
of
enhanced
filtration
5.2.4.1
X
Pre­
LT2
removal/
Inactivation
using
triangular
distributions
(
with
uncertain
modes)
5.2.4.1
X
LT2
treatment
log
reduction
achieved
5.2.4.1
X
Morbidity
benefits
based
on
COI
data
5.3.1
X
Benefits
from
other
rule
provisions
EPA
was
not
able
to
quantify
5.4.5
X
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
65
Systems
Serving
<
10,000
Systems
Serving
>
10,000
All
Systems
Lower
(
5th
%
tile)
Upper
(
95th
%
tile)
9,547,374
162,723,506
172,270,879
A
B
C
D
E
A1
63,866
984,717
1,048,583
172,409
2,410,265
A2
62,168
969,548
1,031,716
170,643
2,367,993
A3
­
Preferred
Alt.
59,994
958,921
1,018,915
169,358
2,331,467
A4
51,723
888,847
940,570
161,366
2,084,877
A1
20,919
302,996
323,915
53,589
757,802
A2
16,027
259,589
275,616
47,681
619,109
A3
­
Preferred
Alt.
12,601
243,572
256,173
45,292
560,648
A4
9,786
209,034
218,820
41,215
457,470
A1
36,026
527,904
563,930
92,474
1,367,370
A2
31,647
488,854
520,501
87,782
1,244,961
A3
­
Preferred
Alt.
27,791
470,572
498,363
84,724
1,177,415
A4
23,006
410,514
433,520
78,386
984,238
A1
7
137
144
25
316
A2
7
135
142
25
311
A3
­
Preferred
Alt.
7
134
141
25
308
A4
6
127
133
24
282
A1
2
42
44
8
99
A2
2
37
39
7
84
A3
­
Preferred
Alt.
1
36
37
7
78
A4
1
32
33
7
67
A1
4
73
77
14
177
A2
4
69
73
13
164
A3
­
Preferred
Alt.
3
67
70
13
157
A4
3
61
64
12
136
90%
Confidence
Bound
for
All
Systems
Illnesses
Avoided
Population
at
Risk
Mean
ICR
Data
ICRSSL
Data
Deaths
Avoided
ICR
Data
Regulatory
Alternative
Regulatory
Alternative
ICRSSM
Data
Regulatory
Alternative
ICRSSM
Data
Regulatory
Alternative
Regulatory
Alternative
Regulatory
Alternative
ICRSSL
Data
Exhibit
5.27
Annual
Cases
of
Illness
and
Deaths
Avoided
from
the
LT2ESWTR
for
Regulatory
Alternatives
Sources:
[
A]
Appendix
C,
Exhibit
C.
4,
Columns
A
and
D,
Row
­
Small
Systems,
A1­
A4,
ICR,
ICRSSL,
and
ICRSSM.
[
B]
Appendix
C,
Exhibit
C.
4,
Columns
A
and
D,
Row
­
Large
Systems,
A1­
A4,
ICR,
ICRSSL,
and
ICRSSM.
[
C[
Appendix
C,
Exhibit
C.
4,
Columns
A
and
D,
Row
­
All
Systems,
A1­
A4,
ICR,
ICRSSL,
and
ICRSSM.
[
D]
Appendix
C,
Exhibit
C.
4,
Columns
B
and
E,
Row
­
All
Systems,
A1­
A4,
ICR,
ICRSSL,
and
ICRSSM.
[
E]
Appendix
C,
Exhibit
C.
4,
Columns
C
and
F,
Row
­
All
Systems,
A1­
A4,
ICR,
ICRSSL,
and
ICRSSM.
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
66
A
B
C
D
E
F
G
A1
38
$
666
$
704
$
34
$
744
$
778
$
1,482
$
A2
37
$
656
$
693
$
33
$
735
$
768
$
1,461
$
A3
­
Preferred
Alt.
36
$
648
$
684
$
32
$
729
$
761
$
1,445
$
A4
31
$
601
$
632
$
28
$
689
$
717
$
1,349
$

A1
12
$
205
$
217
$
11
$
228
$
239
$
457
$
A2
10
$
176
$
185
$
9
$
204
$
212
$
397
$
A3
­
Preferred
Alt.
7
$
165
$
172
$
7
$
195
$
202
$
374
$
A4
6
$
141
$
147
$
6
$
175
$
180
$
328
$

A1
21
$
357
$
378
$
19
$
398
$
418
$
796
$
A2
19
$
331
$
349
$
17
$
376
$
393
$
742
$
A3
­
Preferred
Alt.
17
$
318
$
335
$
15
$
366
$
381
$
715
$
A4
14
$
278
$
291
$
13
$
331
$
344
$
635
$

A1
31
$
565
$
596
$
28
$
637
$
665
$
1,260
$
A2
30
$
556
$
586
$
27
$
630
$
657
$
1,243
$
A3
­
Preferred
Alt.
29
$
550
$
579
$
26
$
624
$
651
$
1,230
$
A4
25
$
510
$
535
$
23
$
590
$
613
$
1,148
$

A1
10
$
174
$
184
$
9
$
196
$
205
$
389
$
A2
8
$
149
$
157
$
7
$
174
$
182
$
338
$
A3
­
Preferred
Alt.
6
$
140
$
146
$
6
$
167
$
172
$
318
$
A4
5
$
120
$
125
$
5
$
150
$
154
$
279
$

A1
17
$
303
$
320
$
16
$
341
$
357
$
677
$
A2
15
$
280
$
296
$
14
$
322
$
336
$
632
$
A3
­
Preferred
Alt.
13
$
270
$
283
$
12
$
313
$
325
$
609
$
A4
11
$
235
$
247
$
10
$
284
$
294
$
541
$
3%
Discount
Rate
7%
Discount
Rate
Total
Value,
All
Systems
Rule
Alternative
Data
Set
All
Systems
ICR
Estimated
Value
of
Cases
of
Illnesses
Avoided
($
Millions)
Estimated
Value
of
Deaths
Avoided
($
Millions)

ICRSSL
ICR
ICRSSL
ICRSSM
Serving
>
10,000
Serving
<
10,000
Serving
>
10,000
All
Systems
Serving
<
10,000
ICRSSM
Exhibit
5.28a
Summary
of
Estimated
Present
Values
of
Annual
Illnesses
and
Deaths
Avoided
from
LT2ESWTR
for
Regulatory
Alternatives,
Enhanced
Cost
of
Illness
[
1]
($
Millions,
2000$)

Notes:
1]
The
traditional
COI
only
includes
valuation
for
medical
costs
and
lost
work
time
(
including
some
portion
of
unpaid
household
production).
The
enhanced
COI
also
factors
in
valuations
for
lost
personal
time
(
non­
work
time)
such
as
child
care
and
homemaking
(
to
the
extent
not
covered
by
the
traditional
COI),
time
with
family,
and
recreation,
and
lost
productivity
at
work
on
days
when
workers
are
ill
but
go
to
work
anyway.
Sources:
[
A]
Appendix
C,
Exhibits
C.
4a
and
C.
5a,
Column
G,
Row
­
Small
Systems,
A1­
A4,
ICR,
ICRSSL,
and
ICRSSM.
[
B]
Appendix
C,
Exhibits
C.
4a
and
C.
5a,
Column
G,
Row
­
Large
Systems,
A1­
A4,
ICR,
ICRSSL,
and
ICRSSM.
[
C]
Appendix
C,
Exhibits
C.
4a
and
C.
5a,
Column
G,
Row
­
All
Systems,
A1­
A4,
ICR,
ICRSSL,
and
ICRSSM.
[
D]
Appendix
C,
Exhibits
C.
4a
and
C.
5a,
Column
M,
Row
­
Small
Systems,
A1­
A4,
ICR,
ICRSSL,
and
ICRSSM.
[
E]
Appendix
C,
Exhibits
C.
4a
and
C.
5a,
Column
M,
Row
­
Large
Systems,
A1­
A4,
ICR,
ICRSSL,
and
ICRSSM.
[
F]
Appendix
C,
Exhibits
C.
4a
and
C.
5a,
Column
M,
Row
­
All
Systems,
A1­
A4,
ICR,
ICRSSL,
and
ICRSSM.
[
G]
Appendix
C,
Exhibits
C.
4a
and
C.
5a,
Column
P,
Row
­
Small
Systems,
A1­
A4,
ICR,
ICRSSL,
and
ICRSSM.
Economic
Analysis
for
the
LT2ESWTR
Proposal
June
2003
5­
67
A
B
C
D
E
F
G
A1
11
$
200
$
212
$
34
$
744
$
778
$
989
$
A2
11
$
197
$
208
$
33
$
735
$
768
$
977
$
A3
­
Preferred
Alt.
11
$
195
$
206
$
32
$
729
$
761
$
967
$
A4
9
$
181
$
190
$
28
$
689
$
717
$
907
$

A1
4
$
62
$
65
$
11
$
228
$
239
$
305
$
A2
3
$
53
$
56
$
9
$
204
$
212
$
268
$
A3
­
Preferred
Alt.
2
$
50
$
52
$
7
$
195
$
202
$
253
$
A4
2
$
43
$
44
$
6
$
175
$
180
$
225
$

A1
6
$
107
$
114
$
19
$
398
$
418
$
531
$
A2
6
$
99
$
105
$
17
$
376
$
393
$
498
$
A3
­
Preferred
Alt.
5
$
96
$
101
$
15
$
366
$
381
$
481
$
A4
4
$
83
$
88
$
13
$
331
$
344
$
432
$

A1
9
$
171
$
180
$
28
$
637
$
665
$
845
$
A2
9
$
168
$
177
$
27
$
630
$
657
$
834
$
A3
­
Preferred
Alt.
9
$
166
$
175
$
26
$
624
$
651
$
826
$
A4
7
$
154
$
162
$
23
$
590
$
613
$
775
$

A1
3
$
53
$
56
$
9
$
196
$
205
$
260
$
A2
2
$
45
$
47
$
7
$
174
$
182
$
229
$
A3
­
Preferred
Alt.
2
$
42
$
44
$
6
$
167
$
172
$
216
$
A4
1
$
36
$
38
$
5
$
150
$
154
$
192
$

A1
5
$
91
$
97
$
16
$
341
$
357
$
454
$
A2
5
$
85
$
89
$
14
$
322
$
336
$
425
$
A3
­
Preferred
Alt.
4
$
82
$
86
$
12
$
313
$
325
$
411
$
A4
3
$
71
$
74
$
10
$
284
$
294
$
369
$
7%
Discount
Rate
ICR
ICRSSL
3%
Discount
Rate
ICR
ICRSSL
Total
Value,
All
Systems
Serving
<
10,000
Serving
>
10,000
All
Systems
Serving
<
10,000
Serving
>
10,000
All
Systems
ICRSSM
ICRSSM
Data
Set
Rule
Alternative
Estimated
Value
of
Cases
of
Illnesses
Avoided
($
Millions)
Estimated
Value
of
Deaths
Avoided
($
Millions)
Exhibit
5.28b
Summary
of
Estimated
Present
Values
of
Annual
Illnesses
and
Deaths
Avoided
from
LT2ESWTR
for
Regulatory
Alternatives,
Traditional
Cost
of
Illness
[
1]
($
Millions,
2000$)

Notes:
1]
The
traditional
COI
only
includes
valuation
for
medical
costs
and
lost
work
time
(
including
some
portion
of
unpaid
household
production).
The
enhanced
COI
also
factors
in
valuations
for
lost
personal
time
(
non­
work
time)
such
as
child
care
and
homemaking
(
to
the
extent
not
covered
by
the
traditional
COI),
time
with
family,
and
recreation,
and
lost
productivity
at
work
on
days
when
workers
are
ill
but
go
to
work
anyway.
Sources:
[
A]
Appendix
C,
Exhibits
C.
4b
and
C.
5b,
Column
J,
Row
­
Small
Systems,
A1­
A4,
ICR,
ICRSSL,
and
ICRSSM.
[
B]
Appendix
C,
Exhibits
C.
4b
and
C.
5b,
Column
J,
Row
­
Large
Systems,
A1­
A4,
ICR,
ICRSSL,
and
ICRSSM.
[
C]
Appendix
C,
Exhibits
C.
4b
and
C.
5b,
Column
J,
Row
­
All
Systems,
A1­
A4,
ICR,
ICRSSL,
and
ICRSSM.
[
D]
Appendix
C,
Exhibits
C.
4b
and
C.
5b,
Column
M,
Row
­
Small
Systems,
A1­
A4,
ICR,
ICRSSL,
and
ICRSSM.
[
E]
Appendix
C,
Exhibits
C.
4b
and
C.
5b,
Column
M,
Row
­
Large
Systems,
A1­
A4,
ICR,
ICRSSL,
and
ICRSSM.
[
F]
Appendix
C,
Exhibits
C.
4b
and
C.
5b,
Column
M,
Row
­
All
Systems,
A1­
A4,
ICR,
ICRSSL,
and
ICRSSM.
[
G]
Appendix
C,
Exhibits
C.
4b
and
C.
5b,
Column
P,
Row
­
Small
Systems,
A1­
A4,
ICR,
ICRSSL,
and
ICRSSM.
Economic
Analysis
for
the
LT2ESWTR
Proposal
6­
1
June
2003
6.
Cost
Analysis
6.1
Introduction
This
chapter
presents
estimates
of
the
total
national
costs
for
the
four
LT2ESWTR
regulatory
alternatives.
Total
national
costs
include
the
costs
of
rule
implementation,
monitoring
for
bin
classification,
additional
treatment,
and
future
monitoring.
These
costs
are
summarized
first,
and
then
individual
methodologies
and
cost
details
are
provided.
This
chapter
also
summarizes
per­
household
costs
for
all
systems
covered
by
the
rule.

The
estimated
costs
of
this
rule,
as
presented
in
this
chapter,
are
highly
dependent
on
the
estimated
occurrence
of
Cryptosporidium
in
source
water.
As
discussed
in
Chapter
4,
EPA
has
three
different
occurrence
data
sets,
the
ICR,
ICRSSL,
and
ICRSSM.
Each
has
uncertainties
surrounding
its
applicability
to
nationwide
occurrence.
The
occurrence
distributions
modeled
from
these
data
sets
and
their
underlying
uncertainties
are
discussed
in
more
detail
in
Chapter
4.
As
there
is
no
clearly
superior
occurrence
data
set,
separate
cost
analyses
were
conducted
using
each
of
them.
Additionally,
section
6.12
presents
a
sensitivity
analysis
of
occurrence
distributions,
using
distributions
outside
the
ranges
of
the
highest
(
ICR)
and
lowest
(
ICRSSL)
data
sets.
Costs
are
also
presented
for
each
of
the
four
regulatory
alternatives
evaluated,
denoted
as
A1
through
A4
(
described
in
Chapter
3).

Section
6.1.1
describes
the
assumptions
used
to
estimate
national
costs
of
the
LT2ESWTR.
The
remaining
sections
detailed
below
present
the
estimated
costs
of
the
LT2ESWTR.

6.2
Summary
of
the
National
Costs
of
the
LT2ESWTR
6.3
Rule
Implementation
Costs
6.3.1
PWSs
6.3.2
States
and
Other
Primacy
Agencies
6.4
Source
Water
Monitoring
for
Initial
Bin
Classification
Costs
6.4.1
PWSs
6.4.2
State
and
Other
Primacy
Agency
Costs
6.5
Treatment
Costs
6.5.1
Toolbox
Options
and
Unit
Costs
6.5.2
Compliance
Forecast
and
Technology
Selection
6.5.3
Capital
and
Annual
Costs
6.6
Costs
of
Treatment
for
Unfiltered
Plants
6.7
Costs
for
Benchmarking
and
Technology
Reporting
Requirements
6.8
Costs
of
Treatment
for
Uncovered
Finished
Water
Reservoirs
6.8.1
Unit
Costs
6.8.2
Compliance
Forecast
and
Technology
Selection
6.8.3
Total
Annual
Treatment
Costs
6.9
Future
Source
Water
Monitoring
6.10
Household
Costs
6.11
Unquantifiable
Costs
6.12
Summary
of
Uncertainties
and
Sensitivity
Analyses
6.12.1
Cryptosporidium
Occurrence
Data
Sets
6.12.2
Sensitivity
Analysis
of
Influent
Bromide
Levels
on
Technology
Selection
for
Filtered
Plants
1
For
purposes
of
analyses
in
this
EA,
all
present
value
figures
are
presented
at
a
year
2000
price
level.
Present
value
calculations
are
performed
to
the
expected
year
of
rule
implementation
(
2003).

2
See
EPA's
Guidelines
for
Preparing
Economic
Analyses
(
USEPA
2000e)
for
a
full
discussion
of
the
use
of
social
discount
rates
in
the
evaluation
of
policy
decisions.

Economic
Analysis
for
the
LT2ESWTR
Proposal
6­
2
June
2003
6.13
Comparison
of
Regulatory
Alternatives
Appendix
D
offers
a
comprehensive
explanation
of
the
laboratory
and
labor
costs
for
rule
implementation,
E.
coli
and
Cryptosporidium
monitoring
for
initial
bin
classification,
and
future
monitoring.
Appendices
E,
F,
G,
H,
and
Q
support
the
cost
estimates
(
unit
costs,
methodology
for
technology
selection,
summary
of
technology
selection,
and
summary
of
treatment
costs,
respectively).
In
addition,
Appendices
G
and
H
provide
results
for
sensitivity
analyses
and
regulatory
alternatives.
Appendix
I
presents
unit
costs
for
uncovered
finished
water
reservoirs,
and
Appendix
J
summarizes
the
per­
household
cost
estimation
methodology.

6.1.1
Cost
Description
and
Assumptions
To
estimate
the
total
national
costs
of
the
rule,
EPA
estimated
costs
to
be
incurred
by
public
water
systems
(
PWSs)
and
States
or
other
Primacy
Agencies.
For
PWSs,
these
include
the
costs
of
installing
treatment,
the
costs
to
administer
the
program
and
understand
the
rule,
and
monitoring
costs.
State
and
Primacy
Agency
costs
include
estimates
of
the
labor
burdens
that
these
agencies
will
face,
such
as
training
employees
on
the
requirements
of
the
LT2ESWTR,
reviewing
PWS
reports
and
responding
to
inquiries,
and
recordkeeping.

EPA
estimated
costs
for
these
activities
using
cost
models,
equipment
price
lists
and
quotes,
wage
rates
from
the
Bureau
of
Labor
Statistics,
stakeholder
inputs,
and
assumptions
used
in
economic
analyses
performed
for
earlier
drinking
water
rules.
This
section
discusses
the
assumptions
on
discount
rates,
wage
rates,
laboratory
fees,
and
uncertainty
parameters.
Section
6.5.1
describes
calculations
of
technology
costs.
More
details
on
cost
assumptions
and
results
are
in
Appendices
D
through
J.
EPA
expresses
costs
as
annualized
values
over
25
years,
based
on
the
present
value
of
the
stream
of
costs
that
occur
over
time.
All
cost
estimates
are
expressed
in
year
2000
dollars.

Scheduling
and
Discounting
Assumptions
for
National
Costs
Nominal
cost
estimates
for
both
non­
treatment
and
treatment
activities
are
of
two
kinds:
(
1)
onetime
costs
that
occur
near
the
beginning
of
the
rule
implementation
period,
and
(
2)
annual
"
steady­
state"
costs
that
systems
and
States/
Primacy
Agencies
will
incur
after
systems
have
made
necessary
changes
to
treatment
and/
or
monitoring
to
comply
with
the
LT2ESWTR.
For
the
purposes
of
this
Economic
Analysis
(
EA),
one­
time
and
steady­
state
costs
were
projected
over
a
25­
year
time
period
to
coincide
with
the
estimated
life
span
of
capital
equipment
(
typically
estimated
as
20
years
for
most
technologies)
and
an
average
time­
lag
of
up
to
five
years
for
technology
installation
after
rule
promulgation.
The
projected
schedules
for
all
rule
activities
are
summarized
in
Appendix
O.

As
described
previously
in
the
Chapter
5
discussion
of
benefits,
it
is
common
practice
to
adjust
benefits
and
costs
to
a
present
value1
using
a
social
discount
rate
so
that
they
can
be
compared
to
one
another.
This
process
takes
into
account
the
time
preference
that
society
places
on
expenditures
and
allows
comparison
of
cost
and
benefit
streams
that
are
variable
over
a
given
time
period.
2
Similar
to
Economic
Analysis
for
the
LT2ESWTR
Proposal
6­
3
June
2003
calculating
the
present
value
of
benefits
(
see
section
5.3.1.5),
the
present
value
of
costs
for
any
future
period
can
be
calculated
using
the
following
equation:

PV
=
V(
t)
/
(
1
+
R)
t
Where:
t
=
The
number
of
years
from
the
reference
period
R
=
Social
discount
rate
V(
t)
=
The
cost
occurring
t
years
from
the
reference
period
The
present
values
presented
in
this
EA
are
the
sum
of
the
PVs
for
each
year.

There
is
much
discussion
among
economists
of
the
proper
social
discount
rate
to
use
for
policy
analysis.
Therefore,
for
LT2ESWTR
cost
analyses,
present
value
calculations
are
made
using
two
social
discount
rates
thought
to
best
represent
current
policy
evaluation
methodologies,
3
and
7
percent.
Historically,
the
use
of
a
3
percent
is
based
on
rates
of
return
on
relatively
risk­
free
investments,
as
described
in
the
Guidelines
for
Preparing
Economic
Analyses
(
USEPA
2000e).
The
rate
of
7
percent
is
a
recommendation
of
the
Office
of
Management
and
Budget
(
OMB)
as
an
estimate
of
"
before­
tax
rate
of
return
to
incremental
private
investment"
(
USEPA
1996b).
For
any
future
cost,
the
higher
the
discount
rate,
the
lower
the
present
value.
Specifically,
a
future
cost
(
or
stream
of
costs)
evaluated
at
a
7
percent
social
discount
rate
will
always
result
in
a
lower
total
present
value
cost
than
the
same
future
cost
evaluated
at
a
3
percent
rate.

To
allow
evaluation
on
an
annual
basis,
the
total
present
value
costs
are
annualized
using
the
same
social
discount
rates
(
3
and
7
percent)
over
25
years.
Unlike
the
total
present
value,
the
relationship
between
annualized
costs
at
3
and
7
percent
is
dependent
on
the
time
frame
for
annualization,
as
well
as
when
the
costs
are
incurred
(
as
set
forth
in
the
schedule
of
rule
activities,
Exhibit
O.
2).
When
applying
social
discount
rates
to
annualize
costs,
a
higher
discount
rate
will
yield
lower
annualized
cost
in
the
earlier
years.
Given
a
long
enough
time
frame,
the
7
percent
annualized
value
will
eventually
be
greater
than
the
3
percent
annualized
value.

Labor
Rates
EPA
estimates
the
labor
needs
and
hourly
labor
rates
of
systems
and
States
or
Primacy
Agencies,
for
two
labor
categories:
managerial
and
technical.
For
PWSs,
all
analyses
use
labor
rates
representative
of
national
averages
as
reported
by
the
Bureau
of
Labor
Statistics
(
BLS).
For
technical
labor,
the
year
2000
mean
hourly
wage
rate
of
$
15.60
for
Standard
Occupational
Classification
(
SOC)
51­
8031,
"
Water
and
Liquid
Waste
Treatment
Plant
and
System
Operators"
is
used.
For
managerial
labor,
the
year
2000
mean
hourly
wage
rate
of
$
28.07
for
SOC
17­
2051,
"
Civil
Engineers"
is
used.
To
account
for
the
cost
of
fringe
benefits,
a
60
percent
loading
factor
is
applied
to
each
of
the
BLS
rates,
resulting
in
a
technical
rate
of
$
24.96
per
hour
and
a
managerial
rate
of
$
44.91
per
hour.

EPA
recognizes
that
there
may
be
significant
variation
in
labor
rates
among
PWSs.
However,
data
are
not
currently
available
that
would
allow
statistically
valid
assignment
of
labor
rates
to
specific
PWSs
based
on
characteristics
such
as
size,
classification,
or
geographical
region.
In
the
absence
of
such
data
and
because
analyses
in
this
EA
are
performed
on
a
national
level,
the
BLS
data
are
used.

PWS
cost
estimates
presented
in
this
EA
for
implementation
and
for
E.
coli
and
Cryptosporidium
monitoring
reflect
the
labor
rates
given
above.
Treatment
technology
costs
as
Economic
Analysis
for
the
LT2ESWTR
Proposal
6­
4
June
2003
Laboratory
Cost
Per
Sample
Total
Range
Laboratory
Microorganism
Average
Min
Max
Cost
per
Plant
A
B
C
D
Cryptosporidium
530
$
389
$
713
$
E.
coli
Utility
Analysis
21
12
38
Commercial
Lab
70
60
85
$
1,820
$
546
$
13,767
summarized
in
the
EPA
document,
"
Technologies
and
Costs
for
Control
of
Microbial
Contaminants
and
Disinfection
By­
Products"
(
T&
C
document)
(
USEPA
2003a),
include
operational
and
maintenance
(
O&
M)
costs
and,
therefore,
also
use
the
BLS
rates
listed
above.

Labor
costs
attributable
to
States
or
other
Primacy
Agencies
are
estimated
based
on
an
average
annual
full
time
equivalent
(
FTE)
labor
cost,
including
fringe
benefits,
of
$
55,000
(
1997$).
This
rate
was
established
based
on
State
input
during
the
development
of
the
State
Workload
Model
in
1997.
For
use
in
the
LT2ESWTR
EA
analyses,
the
$
55,000
annual
rate
was
updated
to
a
year
2000
price
level
($
60,086)
and
converted
to
an
hourly
basis
(
1
FTE
=
2,080
hours)
to
establish
a
State/
Primacy
Agency
rate
of
$
28.89
per
hour.
(
This
is
a
salary
rate
that
includes
fringe
benefits;
therefore,
no
additionalloading
factor
was
applied
as
with
the
PWS
rates.)

Laboratory
Fees
A
laboratory
fee,
expressed
as
a
cost
per
sample,
is
associated
with
E.
coli
and
Cryptosporidium
monitoring
and
future
monitoring
for
bin
reclassification.
Exhibit
6.1
summarizes
the
range
of
laboratory
fees
estimated
from
a
survey
of
laboratories
and
EPA's
experience
during
the
ICR
and
ICRSSs.
Cost
calculations
for
this
EA
used
the
average
laboratory
costs
for
water
sample
analyses.
Costs
are
calculated
on
a
per­
plant
basis
to
be
consistent
with
costs
for
treatment.
Appendix
D
provides
a
more
detailed
derivation
of
the
laboratory
costs.

Some
of
the
factors
that
could
cause
the
cost
per
sample
to
differ
from
one
system
to
another
are
regional
variations
in
laboratory
fees,
the
number
of
samples
processed
(
quantity
discounts),
and
laboratory
capacity
for
Cryptosporidium
analysis.

Exhibit
6.1
E.
coli
and
Cryptosporidium
Laboratory
Costs
Source:
Appendix
D,
described
in
sections
D.
4.1
and
D.
4.2.
[
A]
Cryptosporidium
­
Exhibit
D.
14a,
column
F;
E.
coli
­
Exhibit
D.
12,
columns
F
and
I.
[
B]
and
[
C]
Cryptosporidium
­
DynCorp
(
2002);
E.
coli
­
DynCorp
(
2000).
[
D]
Column
[
A]
multiplied
by
the
number
of
samples
(
26)
(
biweekly
samples
for
E.
coli,
and
24
regular
samples
for
Cryptosporidium
plus
2
spiked
samples).

Modeled
Uncertainty
Parameters
As
noted
throughout
this
EA,
EPA
recognizes
that
there
is
variability
among
many
of
the
input
parameters
to
the
LT2ESWTR
cost
model
(
e.
g.,
plants
per
system,
population
served,
flow
per
population,
labor
rates,
occurrence
distributions).
In
most
cases,
there
is
insufficient
information
to
fully
characterize
Economic
Analysis
for
the
LT2ESWTR
Proposal
6­
5
June
2003
the
distribution
of
variability
on
a
national
scale.
EPA
believes
that
mean
values
for
the
various
input
parameters
are
adequate
to
generate
EPA's
best
estimate
of
national
costs
for
the
rule.

EPA
also
recognizes
that
there
is
uncertainty
in
the
national
cost
estimates,
and
has
characterized
the
uncertainty
around
the
mean
unit
technology
costs
(
as
described
in
section
6.5.1.2)
in
the
LT2ESWTR
cost
model.
Thus,
national
costs
of
treatment
(
initial
capital
and
operational
and
maintenance
(
O&
M))
are
presented
with
confidence
bounds
in
this
EA.
Another
dimension
of
uncertainty
is
reflected
in
the
presentation
of
costs
separately
for
each
of
the
three
modeled
occurrence
distributions.

6.2
Summary
of
the
National
Costs
of
the
LT2ESWTR
This
section
presents
estimates
of
the
number
of
systems
expected
to
incur
costs
as
a
result
of
the
LT2ESWTR
and
the
national
cost
of
the
rule.
Costs
are
broken
out
in
several
ways,
including
by
type
of
system
or
plant
subject
to
rule
provisions,
system
size,
and
nature
of
cost
(
one­
time
or
annual).
All
costs
presented
in
this
section
are
for
the
Preferred
Regulatory
Alternative
for
the
LT2ESWTR.
Cost
estimates
for
the
other
regulatory
alternatives
are
presented
in
section
6.13.

Systems
Subject
to
Rule
Provisions
and
Activities
The
LT2ESWTR
applies
to
all
surface
water
and
ground
water
under
the
direct
influence
of
surface
water
(
GWUDI)
systems
that
are
classified
as
community
water
systems
(
CWSs),
nontransient
noncommunity
water
systems
(
NTNCWSs),
or
transient
noncommunity
water
systems
(
TNCWSs),
as
described
in
Chapter
4.
Unfiltered
and
filtered
systems
are
subject
to
different
rule
provisions,
as
are
systems
with
uncovered
finished
water
reservoirs.
It
is
estimated
that
all
unfiltered
systems
will
add
treatment
to
meet
rule
requirements,
and
all
systems
with
uncovered
finished
water
reservoirs
will
cover
their
reservoirs
or
treat
the
discharge
to
inactivate
viruses.
See
Chapter
4,
sections
4.4
and
4.6
for
baseline
numbers
of
unfiltered
systems
and
uncovered
finished
water
reservoirs
subject
to
the
rule.

Exhibit
6.2
shows
the
estimated
number
of
systems
and
plants
that
are
subject
to
rule
implementation,
source
water
monitoring,
treatment,
and
benchmarking
costs.
All
nonpurchased
plants
will
incur
rule
implementation
costs.
Plants
achieving
5.5
log
treatment
of
Cryptosporidium
will
not
incur
treatment
costs
nor,
possibly,
source
water
monitoring
costs,
depending
on
when
they
meet
the
5.5
log
treatment
with
respect
to
LT2ESWTR
promulgation.

EPA
assumes
no
unfiltered
plant
currently
achieves
2
log
inactivation
of
Cryptosporidium
as
required
by
the
LT2ESWTR;
therefore,
all
unfiltered
plants
will
incur
costs
for
adding
treatment.
Filtered
plants
with
source
water
monitoring
results
of
0.075
oocysts/
L
or
greater
will
be
required
to
provide
additional
treatment
for
Cryptosporidium.
All
systems
with
uncovered
finished
reservoirs
must
cover
or
treat
the
effluent.
Purchased
systems
do
not
have
direct
treatment
costs
and,
thus,
are
not
included
in
Exhibit
6.2
(
unless
they
have
uncovered
reservoirs;
however
EPA
recognizes
that
they
will
likely
incur
indirect
costs
through
rate
increases
by
the
seller).
Economic
Analysis
for
the
LT2ESWTR
Proposal
6­
6
June
2003
Dataset
Initial
E.
Coli
Monitoring
Initial
Crypto
Monitoring
Future
E.
coli
Monitoring
Future
Crypto
Monitoring
A
B
C
D
E
F
G
<
10,000
5,682
5,792
2,016
5,122
1,782
2,251
32
>
10,000
1,384
1,774
1,774
1,273
1,273
733
106
Total
7,066
7,565
3,789
6,395
3,056
2,984
138
<
10,000
1,295
5,409
1,209
1,463
>
10,000
1,774
1,453
1,453
481
Total
3,069
6,862
2,662
1,944
<
10,000
1,575
5,347
1,454
1,768
>
10,000
1,774
1,386
1,386
578
Total
3,349
6,733
2,840
2,346
Systems
with
Uncovered
Reservoirs
Nonpurchased
Systems
and
Plants
System
Size
(
population
served)
Systems
Incurring
Implementation
Costs
Source
Water
Monitoring
­
Plants
Plants
Adding
Treatment
ICR
Same
as
ICR
ICRSSL
ICRSSM
Same
as
ICR
Exhibit
6.2
Number
of
Systems
and
Plants
Expected
to
Incur
Costs,
Preferred
Alternative
Note:
Detail
may
not
add
to
totals
due
to
independent
rounding.
Plants
adding
treatment
in
column
F
include
purchased
unlinked
plants
(
see
section
4.3.2).
Sources:
[
A]
Appendix
D,
Exhibit
D.
4,
column
A.
[
B]
Appendix
D,
Exhibit
D.
4,
column
D.
[
C]
ICR
 
Appendix
D,
Exhibit
D.
4,
column
F;
ICRSSL
 
Appendix
D,
Exhibit
D.
6,
column
F;
ICRSSM
 
Appendix
D,
Exhibit
D.
5,
column
F.
[
D]
ICR
 
Appendix
D,
Exhibit
D.
4,
column
I;
ICRSSL
 
Appendix
D,
Exhibit
D.
6,
column
I;
ICRSSM
 
Appendix
D,
Exhibit
D.
5,
column
I.
[
E]
ICR
 
Appendix
D,
Exhibit
D.
4,
column
J;
ICRSSL
 
Appendix
D,
Exhibit
D.
6,
column
J.
[
F]
ICR
 
Appendix
G,
Exhibits
G.
37­
G.
39,
column
A;
ICRSSL
 
Appendix
G,
Exhibit
G.
43­
G.
45,
column
A;
ICRSSM
 
Appendix
G,
Exhibit
G.
49­
G.
51,
column
A.
All
include
Exhibit
4.5,
column
C.
[
G]
Exhibit
4.22
and
assuming
one
reservoir
per
system.

One­
Time
Activity
Costs
Total
national
costs
incurred
by
PWSs
include
costs
for
rule
implementation,
E.
coli
and
Cryptosporidium
monitoring,
and
adding
treatment.
Implementation
costs
arise
from
reading
the
rule
and
training
employees
on
its
provisions.
E.
coli
and
Cryptosporidium
monitoring
costs
include
laboratory
sample
analysis
and
labor
for
collecting
the
samples.
Monitoring
costs
are
incurred
at
two
stages
 
initial
monitoring
and
future
reassessment
of
bin
assignment.
Treatment
costs
represent
the
capital
infrastructure
requirement
(
O&
M
costs
are
continuous
and
not
included
in
one­
time
activity
costs).

State/
Primacy
Agency
costs
include
those
associated
with
rule
implementation,
E.
coli
monitoring,
Cryptosporidium
monitoring,
benchmarking,
and
future
monitoring
for
bin
reclassification.
Costs
associated
with
rule
implementation
are
attributable
to
regulation
adoption
and
program
development;
training
State
or
Primacy
Agency
staff,
and
PWS
staff;
providing
technical
assistance;
and
updating
the
data
management
system.
E.
coli
and
Cryptosporidium
monitoring
activities
that
generate
one­
time
costs
for
States
or
other
Primacy
Agencies
include
analyzing
PWS
monitoring
reports,
making
determinations
based
on
the
reports,
responding
to
PWSs,
and
monitoring
recordkeeping,
which
includes
data
entry
of
results
and
decisions.
These
costs
are
also
expected
to
be
small
because
States
will
gain
Primacy
after
medium
and
large
systems
complete
their
initial
monitoring
(
EPA
is
managing
the
data
collection
for
medium
and
large
systems
directly).
Economic
Analysis
for
the
LT2ESWTR
Proposal
6­
7
June
2003
Exhibit
6.3
summarizes
the
estimated
total
initial
capital
and
one­
time
costs
of
the
LT2ESWTR,
which
include
costs
of
rule
implementation
and
all
monitoring
activities.
This
summary
exhibit
(
and
Exhibits
6.4a
and
6.4b)
contain
costs
derived
from
ICR
and
ICRSSL
occurrence
distributions
to
show
the
estimated
range
of
costs.
The
remainder
of
exhibits
show
costs
for
all
three
occurrence
distributions
(
ICR,
ICRSSL,
and
ICRSSM).
Economic
Analysis
for
the
LT2ESWTR
Proposal
6­
8
June
2003
Mean
ICRSSL
Mean
ICR
Mean
ICRSSL
Mean
ICR
Mean
ICRSSL
Mean
ICR
5th
Percentile
ICRSSL
95th
Percentile
ICR
5th
Percentile
ICRSSL
95th
Percentile
ICR
5th
Percentile
ICRSSL
95th
Percentile
ICR
A
B
C
D
E
=
A
+
C
F
=
B
+
D
G
H
I
J
K
=
G
+
I
L
=
H
+
J
Treatment
Mean
$
68.72
$
122.30
$
951.14
$
1,469.50
$
1,019.86
$
1,591.80
$
45.49
$
132.02
$
643.23
$
1,597.51
$
688.72
$
1,729.52
Treatment
Uncertainty
Bound
[
2]
$
59.87
$
135.98
$
805.46
$
1,690.02
$
865.33
$
1,826.00
$
39.48
$
146.63
$
542.26
$
1,840.04
$
581.74
$
1,986.67
Uncovered
Reservoirs
Mean
$
0.07
$
0.07
$
64.28
$
64.28
$
64.35
$
64.35
$
0.07
$
0.07
$
64.28
$
64.28
$
64.35
$
64.35
Uncovered
Reservoirs
Uncertainty
Bound
[
2]
$
0.05
$
0.08
$
52.09
$
76.56
$
52.15
$
76.65
$
0.05
$
0.08
$
52.09
$
76.56
$
52.15
$
76.65
Implementation
$
1.23
$
1.23
$
0.33
$
0.33
$
1.56
$
1.56
$
1.23
$
1.23
$
0.33
$
0.33
$
1.56
$
1.56
Monitoring
$
28.90
$
39.05
$
26.88
$
26.88
$
55.78
$
65.93
$
21.67
$
42.07
$
26.88
$
26.88
$
48.55
$
68.95
Future
Monitoring
$
26.99
$
34.54
$
22.01
$
19.29
$
49.01
$
53.83
$
20.46
$
37.02
$
23.80
$
18.55
$
44.26
$
55.57
Benchmarking
$
0.02
$
0.06
$
0.05
$
0.08
$
0.08
$
0.14
$
0.02
$
0.06
$
0.03
$
0.09
$
0.05
$
0.15
State
Implementation
$
6.68
$
6.68
$
6.68
$
6.68
State
Monitoring
$
10.45
$
10.45
$
10.45
$
10.45
State
Benchmarking
$
0.04
$
0.07
$
0.03
$
0.08
$
1,207.79
$
1,794.80
$
864.63
$
1,937.30
System
[
3]
$
125.93
$
197.24
$
1,064.70
$
1,580.37
$
1,190.63
$
1,777.60
$
88.93
$
212.47
$
758.55
$
1,707.63
$
847.48
$
1,920.10
$
17.17
$
17.20
$
17.15
$
17.20
Total
Total
Serving
<
10k
Range
of
Occurrence
Means
[
1]
with
Cost
Uncertainty
[
2]
($
Millions,
2000$)

Serving
<
10k
Serving
>
10K
Serving
>
10K
Bounds
of
Occurrence
Range
[
1]
with
Cost
Uncertainty
[
2]
($
Millions,
2000$)

Type
of
Cost
TOTALS
National
(
System
+
State)

State/
Primacy
Agency
[
4]
Exhibit
6.3
Initial
Capital
and
One­
Time
Nominal
Costs,
Preferred
Alternative
($
Millions,
2000$)

Notes:
Detail
may
not
add
to
totals
due
to
independent
rounding.
[
1]
The
ICR
(
highest
occurrence)
and
ICRSSL
(
lowest
occurrence)
represent
the
ranges
of
occurrence
means
for
all
three
data
sets.
The
bounds
of
occurrence
range
are
the
5th
percentile
of
the
ICRSSL
data
set
and
the
95th
percentile
of
the
ICR
data
set.
Appendix
O
(
page
O­
1)
describes
the
cost
distributions
derived
from
both
the
ranges
and
bounds
of
occurrence.
[
2]
The
Treatment
Uncertainty
Bound
and
Uncovered
Reservoirs
Uncertainty
Bound
represent
the
5th
percentile
(
shown
in
the
ICRSSL
columns)
and
95th
percentile
(
shown
in
the
ICR
columns)
of
"
Range
of
Occurrence
Means"
and
"
Bounds
of
Occurrence
Range."
Appendix
O
(
Exhibit
O.
1)
describes
these
cost
distributions,
labeled
as
"
low"
and
"
high."
[
3]
Total
system
cost
is
the
sum
of
treatment
mean,
uncovered
reservoir
mean,
implementation,
monitoring,
future
monitoring,
and
benchmarking.
[
4]
Total
State/
Primacy
Agency
cost
is
the
sum
of
State
implementation,
State
monitoring,
and
State
benchmarking.
Sources:
All
data
from
Appendix
O,
and
from
Alternative
A3.
Implementation,
Monitoring,
and
Benchmarking
data
from
Exhibit
O.
3a.
Treatment
and
Uncovered
Reservoir
data
from
Exhibit
O.
4a,
`
Range
of
Occurrence
Means'
uses
ICRSSL
and
ICR
rows,
and
`
Bounds
of
Occurrence
Range'
uses
Low
and
High
rows.
"
Treatment
Mean":
(
All
columns
from
Exhibit
O.
4a,
A3,
Row­
Mean;
Listed
in
order
of
ICRSSL,
ICR,
Low,
and
High)
[
A],
[
B],
[
G],
and
[
H]
­
Sum
of
columns
A
and
C
[
C],
[
D],
[
I],
and
[
J]
­
Sum
of
columns
G
and
I
"
Treatment
Uncertainty
Bound":
(
All
columns
from
Exhibit
O.
4a,
A3,
Row­
5th
and
95th;
Listed
as
ICRSSL,
ICR,
Low,
and
High,
respectively)
[
A],
[
B],
[
G],
and
[
H]
­
Sum
of
columns
A
and
C
[
C],
[
D],
[
I],
and
[
J]
­
Sum
of
columns
G
and
I
"
Uncovered
Reservoirs
Mean":
(
All
columns
from
Exhibit
O.
4a,
A3,
Row­
Mean;
Listed
as
ICRSSL,
ICR,
Low,
and
High,
respectively)
[
A],
[
B],
[
G],
and
[
H]
­
Column
E
[
C],
[
D],
[
I],
and
[
J]
­
Column
K
"
Uncovered
Reservoirs
Uncertainty
Bound":
(
All
columns
from
Exhibit
O.
4a,
A3,
Row­
5th
and
95th;
Listed
as
ICRSSL,
ICR,
Low,
and
High,
respectively)
[
A],
[
B],
[
G],
and
[
H]
­
Column
E
[
C],
[
D],
[
I],
and
[
J]
­
Column
K
State
data
from
Exhibit
O.
5a.
Economic
Analysis
for
the
LT2ESWTR
Proposal
6­
9
June
2003
Annualized
Costs
Exhibits
6.4a­
b
display
annualized
costs
for
systems
and
States
or
other
Primacy
Agencies
at
3
and
7
percent
discount
rates.
The
present
values
of
implementation,
monitoring
(
E.
coli
and
Cryptosporidium),
benchmarking,
compliance
reporting,
capital,
and
O&
M
cost
(
treatment
and
uncovered
finished
reservoir
costs
include
O&
M)
are
annualized
at
the
3
and
7
percent
discount
rates
over
25
years.
For
States/
Primacy
Agencies,
implementation,
monitoring,
benchmarking,
and
compliance
reporting
costs
are
also
annualized
to
estimate
total
annual
State
or
Primacy
Agency
costs.
PWS
and
State/
Primacy
Agency
annualized
costs
are
added
together
to
estimate
total
national
annualized
costs
for
the
rule
($
74
to
$
108
million
at
a
3
percent
discount
rate
and
$
81
to
$
118
million
at
a
7
percent
discount
rate,
based
on
ICRSSL
and
ICR
modeled
Cryptosporidium
occurrence
distributions).
Sections
6.3
through
6.8
provide
detailed
cost
estimates
for
systems
and
States/
Primacy
Agencies
according
to
activity.
Economic
Analysis
for
the
LT2ESWTR
Proposal
6­
10
June
2003
Mean
ICRSSL
Mean
ICR
Mean
ICRSSL
Mean
ICR
Mean
ICRSSL
Mean
ICR
5th
Percentile
ICRSSL
95th
Percentile
ICR
5th
Percentile
ICRSSL
95th
Percentile
ICR
5th
Percentile
ICRSSL
95th
Percentile
ICR
A
B
C
D
E
=
A
+
C
F
=
B
+
D
G
H
I
J
K
=
G
+
I
L
=
H
+
J
Treatment
Mean
$
5.32
$
9.56
$
56.79
$
88.82
$
62.10
$
98.38
$
3.48
$
10.33
$
38.40
$
96.48
$
41.88
$
106.81
Treatment
Uncertainty
Bound
[
3]
$
4.83
$
10.29
$
49.69
$
99.56
$
54.51
$
109.85
$
3.15
$
11.12
$
33.46
$
108.26
$
36.62
$
119.38
Uncovered
Reservoirs
Mean
$
0.01
$
0.01
$
5.40
$
5.40
$
5.40
$
5.40
$
0.01
$
0.01
$
5.40
$
5.40
$
5.40
$
5.40
Uncovered
Reservoirs
Uncertainty
Bound
[
4]
$
0.00
$
0.01
$
4.60
$
6.20
$
4.61
$
6.20
$
0.00
$
0.01
$
4.60
$
6.20
$
4.61
$
6.20
Implementation
$
0.07
$
0.07
$
0.02
$
0.02
$
0.09
$
0.09
$
0.07
$
0.07
$
0.02
$
0.02
$
0.09
$
0.09
Monitoring
$
1.48
$
1.99
$
1.48
$
1.48
$
2.95
$
3.46
$
1.11
$
2.14
$
1.48
$
1.48
$
2.59
$
3.62
Future
Monitoring
$
1.06
$
1.34
$
0.94
$
0.82
$
2.00
$
2.16
$
0.81
$
1.44
$
1.02
$
0.79
$
1.82
$
2.24
Benchmarking
$
0.00
$
0.00
$
0.00
$
0.00
$
0.00
$
0.01
$
0.00
$
0.00
$
0.00
$
0.00
$
0.00
$
0.01
State
Implementation
$
0.38
$
0.38
$
0.38
$
0.38
State
Monitoring
$
0.47
$
0.47
$
0.47
$
0.47
State
Benchmarking
$
0.00
$
0.00
$
0.00
$
0.00
State
Technology
Reporting
$
0.09
$
0.15
$
0.07
$
0.15
$
73.48
$
110.50
$
52.70
$
119.16
System
[
5]
$
7.92
$
12.96
$
64.62
$
96.54
$
72.54
$
109.51
$
5.48
$
13.98
$
46.31
$
104.17
$
51.79
$
118.16
$
0.94
$
0.99
$
0.92
$
1.00
Type
of
Cost
TOTALS
National
(
System
+
State)

State/
Primacy
Agency
[
6]
Range
of
Occurrence
Means
[
1]
with
Cost
Uncertainty
[
2]
($
Millions,
2000$)

Serving
<
10k
Serving
>
10K
Total
Bounds
of
Occurrence
Range
[
1]
with
Cost
Uncertainty
[
2]
($
Millions,
2000$)

Serving
<
10k
Serving
>
10K
Total
Mean
ICRSSL
Mean
ICR
Mean
ICRSSL
Mean
ICR
Mean
ICRSSL
Mean
ICR
5th
Percentile
ICRSSL
95th
Percentile
ICR
5th
Percentile
ICRSSL
95th
Percentile
ICR
5th
Percentile
ICRSSL
95th
Percentile
ICR
A
B
C
D
E
=
A
+
C
F
=
B
+
D
G
H
I
J
K
=
G
+
I
L
=
H
+
J
Treatment
Mean
$
5.07
$
9.10
$
62.26
$
97.13
$
67.33
$
106.24
$
3.33
$
9.83
$
42.10
$
105.53
$
45.43
$
115.36
Treatment
Uncertainty
Bound
[
3]
$
4.58
$
9.84
$
54.11
$
109.46
$
58.69
$
119.30
$
2.99
$
10.63
$
36.44
$
119.06
$
39.43
$
129.69
Uncovered
Reservoirs
Mean
$
0.01
$
0.01
$
6.41
$
6.41
$
6.42
$
6.42
$
0.01
$
0.01
$
6.41
$
6.41
$
6.42
$
6.42
Uncovered
Reservoirs
Uncertainty
Bound
[
4]
$
0.01
$
0.01
$
5.41
$
7.42
$
5.41
$
7.43
$
0.01
$
0.01
$
5.41
$
7.42
$
5.41
$
7.43
Implementation
$
0.09
$
0.09
$
0.03
$
0.03
$
0.12
$
0.12
$
0.09
$
0.09
$
0.03
$
0.03
$
0.12
$
0.12
Monitoring
$
1.90
$
2.54
$
2.08
$
2.08
$
3.99
$
4.63
$
1.44
$
2.74
$
2.08
$
2.08
$
3.53
$
4.82
Future
Monitoring
$
0.97
$
1.22
$
0.96
$
0.84
$
1.93
$
2.06
$
0.74
$
1.31
$
1.04
$
0.81
$
1.78
$
2.12
Benchmarking
$
0.00
$
0.00
$
0.00
$
0.00
$
0.00
$
0.01
$
0.00
$
0.00
$
0.00
$
0.01
$
0.00
$
0.01
State
Implementation
$
0.55
$
0.55
$
0.55
$
0.55
State
Monitoring
$
0.52
$
0.52
$
0.52
$
0.52
State
Benchmarking
$
0.00
$
0.00
$
0.00
$
0.00
State
Technology
Reporting
$
0.08
$
0.12
$
0.06
$
0.13
$
80.94
$
120.67
$
58.41
$
130.05
System
[
5]
$
8.04
$
12.97
$
71.74
$
106.50
$
79.78
$
119.47
$
5.61
$
13.98
$
51.66
$
114.86
$
57.28
$
128.84
$
1.15
$
1.20
$
1.13
$
1.21
Serving
>
10K
Total
Type
of
Cost
State/
Primacy
Agency
[
6]
Range
of
Occurrence
Means
[
1]
with
Cost
Uncertainty
[
2]
($
Millions,
2000$)
Bounds
of
Occurrence
Range
[
1]
with
Cost
Uncertainty
[
2]
($
Millions,
2000$)

Serving
>
10K
TOTALS
National
(
System
+
State)
Serving
<
10k
Serving
<
10k
Total
Exhibit
6.4a
Annualized
Total
Costs,
Preferred
Alternative,
at
3
Percent
($
Millions,
2000$)

Exhibit
6.4b
Annualized
Total
Costs,
Preferred
Alternative,
at
7
Percent
($
Millions,
2000$)
Economic
Analysis
for
the
LT2ESWTR
Proposal
6­
11
June
2003
Notes
for
6.4a
and
6.4b:
Detail
may
not
add
to
totals
due
to
independent
rounding.
[
1]
The
ICR
(
highest
occurrence)
and
ICRSSL
(
lowest
occurrence)
represent
the
ranges
of
occurrence
means
for
all
three
data
sets.
The
bounds
of
occurrence
range
are
the
5th
percentile
of
the
ICRSSL
data
set
and
the
95th
percentile
of
the
ICR
data
set.
Appendix
O
(
page
O­
1)
describes
the
cost
distributions
derived
from
both
the
ranges
and
bounds
of
occurrence.
[
2]
The
Treatment
Uncertainty
Bound
and
Uncovered
Reservoirs
Uncertainty
Bound
represent
the
5th
percentile
(
shown
in
the
ICRSSL
columns)
and
95th
percentile
(
shown
in
the
ICR
columns)
of
"
Range
of
Occurrence
Means"
and
"
Bounds
of
Occurrence
Range."
Appendix
O
(
Exhibit
O.
1)
describes
these
cost
distributions,
labeled
as
"
low"
and
"
high."
[
3]
Includes
O&
M
and
systems
technology
reporting
costs.
[
4]
Includes
O&
M
costs.
[
5]
Total
system
cost
is
the
sum
of
treatment
mean,
uncovered
reservoir
mean,
implementation,
monitoring,
future
monitoring,
and
benchmarking.
[
6]
Total
State/
Primacy
Agency
cost
is
the
sum
of
State
implementation,
State
monitoring,
State
benchmarking,
and
State
technology
reporting.
Sources:
All
data
from
Appendix
O,
and
from
Alternative
A3.
Implementation,
Monitoring,
and
Benchmarking
data
from
Exhibits
O.
3d
and
O.
3e.
Treatment
and
Uncovered
Reservoir
data
from
Exhibits
O.
4d
and
O.
4e,
`
Range
of
Occurrence
Means'
uses
ICRSSL
and
ICR
rows,
and
`
Bounds
of
Occurrence
Range'
uses
Low
and
High
rows.
"
Treatment
Mean":
(
All
columns
from
Exhibits
O.
4d
and
O.
4e,
A3,
Row
­
Mean;
Listed
in
order
of
ICRSSL,
ICR,
Low,
and
High)
[
A],
[
B],
[
G],
and
[
H]
­
Sum
of
columns
A,
B,
C,
and
D.
[
C],
[
D],
[
I],
and
[
J]
­
Sum
of
columns
G,
H,
I,
and
J.
"
Treatment
Uncertainty
Bound":
(
All
columns
from
Exhibits
O.
4d
and
O.
4e,
A3,
Row
­
5th
and
95th;
Listed
in
order
of
ICRSSL,
ICR,
Low,
and
High)
[
A],
[
B],
[
G],
and
[
H]
­
Sum
of
columns
A,
B,
C,
and
D.
[
C],
[
D],
[
I],
and
[
J]
­
Sum
of
columns
G,
H,
I,
and
J.
"
Uncovered
Reservoirs
Mean":
(
All
columns
from
Exhibits
O.
4d
and
O.
4e,
A3,
Row
­
Mean;
Listed
in
order
of
ICRSSL,
ICR,
Low,
and
High)
[
A],
[
B],
[
G],
and
[
H]
­
Sum
of
columns
E
and
F
[
C],
[
D],
[
I],
and
[
J]
­
Sum
of
columns
K
and
L.
"
Uncovered
Reservoirs
Uncertainty
Bound":
(
All
columns
from
Exhibits
O.
4d
and
O.
4e,
A3,
Row
­
5th
and
95th;
Listed
in
order
of
ICRSSL,
ICR,
Low,
and
High)
[
A],
[
B],
[
G],
and
[
H]
­
Sum
of
columns
E
and
F
[
C],
[
D],
[
I],
and
[
J]
­
Sum
of
columns
K
and
L.
State
data
from
Exhibits
O.
5d
and
O.
5e.

6.3
Rule
Implementation
Costs
This
section
presents
the
estimated
costs
for
PWSs
and
States/
Primacy
Agencies
to
perform
administrative
activities
associated
with
the
LT2ESWTR.
These
cost
estimates
are
the
same
for
all
regulatory
alternatives.

6.3.1
PWSs
All
nonpurchased
surface
water
and
GWUDI
systems
subject
to
the
LT2ESWTR
(
including
filtered
and
unfiltered
systems)
will
incur
one­
time
costs
that
include
time
for
staff
to
read
the
rule
and
become
familiar
with
its
provisions
and
for
training
employees
on
rule
requirements.
The
technical
and
managerial
labor
rates,
as
presented
in
section
6.1.1,
were
used
along
with
estimates
of
labor
hours
to
generate
rule
implementation
costs
for
all
systems.
Labor
rates
used
to
estimate
implementation
costs
vary
by
activity
and
system
size.
Costs
for
systems
serving
up
to
1,000
people
are
based
on
only
the
Economic
Analysis
for
the
LT2ESWTR
Proposal
6­
12
June
2003
System
Size
(
population
served)
3%
7%
A
B
£
10,000
1,155,166
$
1,070,413
$
>
10,000
332,404
$
332,404
$
Total
1,487,570
$
1,402,817
$
technical
rate
($
24.96
per
hour).
For
those
systems
serving
at
least
1,000
people,
costs
are
based
on
an
assumed
80/
20
split
between
technical
and
managerial
labor
rates
(
the
respective
rates
are
$
24.96
per
hour
and
$
44.91
per
hour).
Labor
hour
estimates
are
based
on
EPA's
experience
with
previous
rules.

Exhibit
6.5
summarizes
the
estimated
implementation
costs
for
PWSs
to
comply
with
the
LT2ESWTR.
Implementation
costs
are
the
same
for
all
rule
options.
In
Appendix
D,
Exhibit
D.
10
provides
detailed
estimates
of
hours
and
calculations
for
system
implementation
costs
by
system
size
(
Appendix
D
costs
are
undiscounted;
discounted
costs
are
shown
in
Appendix
O).

Exhibit
6.5
Present
Value
of
System
Implementation
Total
Costs
by
System
Size,
Preferred
Alternative
(
2000$)

Note:
Detail
may
not
add
to
totals
due
to
independent
rounding.
Large
systems'
implementation
costs
entirely
occur
in
year
0,
therefore,
no
discount
is
applied.
Sources:
[
A]
Appendix
O,
Exhibit
O.
3b,
Rows
­
A3,
ICR
and
ICRSSL;
Columns
A
and
G.
[
B]
Appendix
O,
Exhibit
O.
3c,
Rows
­
A3,
ICR
and
ICRSSL;
Columns
A
and
G.

6.3.2
States
and
Other
Primacy
Agencies
State
and
other
Primacy
Agency
implementation
activities
include:

°
Adopting
the
regulation
and
developing
the
program
°
Training
State
or
other
Primacy
Agency
staff
°
Training
PWS
staff
and
providing
technical
assistance
°
Updating
the
data
management
system
To
estimate
implementation
costs
to
States/
Primacy
Agencies,
the
number
of
full­
time
equivalent
employees
(
FTEs)
per
activity
is
multiplied
by
the
number
of
labor
hours
per
FTE,
the
cost
per
labor
hour,
and
the
number
of
States/
Primacy
Agencies.

EPA
estimates
the
number
of
FTEs
required
per
activity
based
on
previous
experience
with
other
rules.
In
estimating
State/
Primacy
Agency
costs,
a
labor
rate
of
$
28.89
is
assumed
(
see
section
6.1.1).
The
number
of
States
and
territories
is
estimated
as
the
sum
of
the
50
states,
six
territories
(
American
Samoa,
Commonwealth
of
the
Northern
Marianas,
Guam,
Palau,
Puerto
Rico,
and
the
Virgin
Islands),
and
the
Navajo
Nation.
Economic
Analysis
for
the
LT2ESWTR
Proposal
6­
13
June
2003
Appendix
D
(
Exhibit
D.
11)
presents
estimates
of
hours
and
calculations
of
implementation
costs
by
activity.
Based
on
the
labor
rate
and
FTE
estimates,
total
State/
Primacy
Agency
present
value
costs
at
3
percent
for
rule
implementation
are
$
6.6
million
(
Exhibit
O.
5b,
column
A).

6.4
Source
Water
Monitoring
for
Initial
Bin
Classification
Costs
6.4.1
PWSs
Source
water
monitoring
costs
are
estimated
on
a
per­
plant
basis.
Purchased
plants
are
assumed
not
to
treat
source
water
and
not
to
have
any
monitoring
costs,
as
with
rule
implementation
activities.
There
are
three
types
of
monitoring
that
plants
may
be
required
to
conduct
 
turbidity,
E.
coli,
and
Cryptosporidium.
Source
water
turbidity
is
a
water
quality
parameter
that
most
plants
measure
frequently
for
operational
control.
Also,
to
meet
the
SWTR,
IESWTR,
and
LT1ESWTR
requirements,
most
water
systems
have
turbidity
analysis
equipment
in­
house
and
their
operators
are
experienced
in
its
use.
Thus,
EPA
assumes
that
the
incremental
turbidity
monitoring
burden
associated
with
the
LT2ESWTR
is
negligible.

For
large
and
medium
systems,
all
are
required
to
conduct
monthly
E.
coli
and
Cryptosporidium
monitoring
for
2
years
to
determine
initial
bin
classification.
If
systems
achieve
5.5
log
Cryptosporidium
removal/
inactivation
prior
to
rule
promulgation,
then
they
are
not
required
to
conduct
source
water
monitoring.
(
See
Exhibits
4.5
and
4.11
for
the
baseline
number
of
unfiltered
and
filtered
plants
conducting
monitoring,
respectively.)

Small
systems,
except
those
achieving
at
least
5.5
log
Cryptosporidium
removal/
inactivation,
must
conduct
bi­
weekly
E.
coli
monitoring
for
1
year.
Systems
exceeding
the
following
levels
must
monitor
semi­
monthly
for
Cryptosporidium.

°
Annual
mean
>
10
E.
coli/
100
mL
for
lakes
and
reservoirs
°
Annual
mean
>
50
E.
coli/
100
mL
for
flowing
streams
To
estimate
source
water
monitoring
costs
for
small
systems,
this
EA
assumes
all
systems
(
except
those
achieving
5.5
log
treatment)
will
conduct
E.
coli
monitoring
and
only
those
predicted
to
require
additional
treatment
(
from
the
binning
prediction
discussed
in
section
4.5.4)
will
monitor
Cryptosporidium
in
their
source
water
(
see
Appendix
D,
Exhibit
D.
4
for
a
calculation
of
plants
conducting
E.
coli
and
Cryptosporidium
monitoring).
EPA
will
continue
to
investigate
the
use
of
a
surrogate
for
determining
source
water
microbial
quality,
in
order
to
reduce
the
cost
burden
on
these
systems.
If
a
reliable
indicator
is
identified,
EPA
will
issue
guidance
for
system
monitoring.
The
costs
in
this
chapter
may,
therefore,
be
an
overestimate
of
actual
costs.

From
the
modeled
Cryptosporidium
occurrence
distributions,
EPA
estimated
the
percentage
of
plants
that
would
fall
into
treatment
bins
for
each
rule
option.
Exhibit
6.6
presents
the
annualized
monitoring
costs
based
on
the
modeled
Cryptosporidium
occurrence
distributions.
Appendix
D
provides
an
explanation
of
how
these
costs
are
developed.
Economic
Analysis
for
the
LT2ESWTR
Proposal
6­
14
June
2003
System
Size
ICR
(
3%)
ICR
(
7%)
ICRSSL
(
3%)
ICRSSL
(
7%)
ICRSSM
(
3%)
ICRSSM
(
7%)
A
B
C
D
E
F
£
 
10K
34.6
$
29.7
$
25.7
$
22.2
$
29.2
$
25.1
$
>
10k
25.7
$
24.3
$
25.7
$
24.3
$
25.7
$
24.3
$
Total
60.3
$
54.0
$
51.4
$
46.5
$
54.9
$
49.4
$
£
 
10K
4.5
$
3.8
$
4.5
$
3.8
$
4.5
$
3.8
$
>
10k
­
$
­
$
­
$
­
$
­
$
­
$
Total
4.5
$
3.8
$
4.5
$
3.8
$
4.5
$
3.8
$
Systems
State/
Primacy
Agency
6.4.2
State
and
Other
Primacy
Agency
Costs
Because
EPA
will
manage
the
data
collection
for
large
and
medium
initial
source
water
monitoring
directly,
States/
Primacy
Agencies
are
not
predicted
to
incur
any
costs
for
these
activities.
They
will,
however,
incur
costs
from
the
small
system
initial
monitoring
requirement.
The
delayed
start
of
small
system
monitoring
will
allow
some
States
to
assume
Primacy
for
small
system
monitoring.
To
estimate
State/
Primacy
Agency
costs,
the
number
of
FTEs
required
per
activity
is
multiplied
by
the
number
of
labor
hours
per
FTE,
the
State/
Primacy
Agency
labor
hour
cost,
and
the
number
of
States
or
Primacy
Agencies.
Exhibit
6.6
presents
the
estimated
total
cost
incurred
by
States/
Primacy
Agency
for
initial
source
water
monitoring
(
see
Appendix
D,
Exhibit
D.
17,
for
the
derivation
of
costs).

Exhibit
6.6
Initial
Source
Water
Monitoring
Present
Value
Costs
at
3
and
7
Percent,
Preferred
Alternative,
by
System
Size
($
Millions,
2000$)

Notes:
Detail
may
not
add
to
totals
due
to
independent
rounding.
Includes
laboratory
costs,
labor
costs,
and
reporting
costs.
There
are
no
State
costs
for
large
systems
(
EPA
assumes
no
State
will
obtain
Primacy
before
or
during
large
system
source
water
monitoring).
Sources:
[
A]
and
[
C]
Systems
data
from
Appendix
O,
Exhibit
O.
3b;
Rows
­
A3,
ICR
and
ICRSSL;
Columns
B,
C,
and
H.
State
data
from
Appendix
O,
Exhibit
O.
5b;
Rows
­
A3,
ICR
and
ICRSSL;
Columns
B
and
C.
[
B]
and
[
D]
Systems
data
from
Appendix
O,
Exhibit
O.
3c;
Rows
­
A3,
ICR
and
ICRSSL;
Columns
B,
C,
and
H.
State
data
from
Appendix
O,
Exhibit
O.
5b;
Rows
­
A3,
ICR
and
ICRSSL;
Columns
B
and
C.

6.5
Treatment
Costs
As
shown
in
Chapter
4,
filtered
plants
make
up
the
majority
of
surface
water
and
GWUDI
plants
subject
to
the
LT2ESWTR.
Costs
of
treatment
associated
with
these
plants
make
up
the
majority
of
the
costs
associated
with
the
LT2ESWTR.
This
section
reviews
the
cost
methodology
and
total
costs
for
the
Preferred
Regulatory
Alternative
as
follows:

6.5.1
Discusses
the
technologies
and
describes
how
plant
unit
costs
($/
plant)
are
derived.

6.5.2
Presents
the
compliance
and
technology
selection
forecasts.

6.5.3
Describes
how
total
capital
investment
and
annualized
costs
are
estimated
for
each
size
category
and
rule
option.

Treatment
costs
are
calculated
by
estimating
the
number
of
plants
that
will
be
adding
treatment
technologies,
then
multiplying
these
estimates
by
the
unit
costs
($/
plant)
of
the
selected
technologies.
Although
individual
information
is
not
available
on
every
plant,
assumptions
are
made
on
the
percentages
Economic
Analysis
for
the
LT2ESWTR
Proposal
6­
15
June
2003
of
plants
nationwide
that
will
select
each
technology.
Capital
improvement
costs
are
converted
to
present
values
and
are
then
annualized
using
the
discount
rates
presented
in
section
6.1.1.
Exhibit
6.7
provides
an
overview
of
the
methodology
used
to
generate
national
treatment
costs.
Economic
Analysis
for
the
LT2ESWTR
Proposal
6­
16
June
2003
1
Baseline
information
is
discussed
in
Chapter
4.
2
Capital
unit
costs,
annual
O&
M
costs,
and
number
of
plants
selecting
technologies
are
also
used
in
the
derivation
of
household
cost
distributions.
See
section
6.9
and
Appendix
J
for
further
discussion.
Average
Daily
Flows
per
Plant
1
Technology
Cost
Equations
(
O&
M)
Point
Estimates
of
Annual
O&
M
Costs
($/
plant)
2
Calculations
discussed
in
section
6.5.1
Cost
Uncertainty
of
+/­
15%
by
Technology
Total
Annual
O&
M
Costs
Total
Capital
Costs
Present
Value
of
Costs
Annualized
Value
of
Costs
Calculations
discussed
in
section
6.5.3
Number
of
Plants
Selecting
Technologies
2
Compliance
Treatment
Forecasts
Baseline
Number
of
Plants
1
Design
Flows
per
Plant
1
Technology
Cost
Equations
(
Capital)
Point
Estimates
of
Capital
Unit
Costs
($/
plant)
2
Calculations
discussed
in
section
6.5.1
Calculations
discussed
in
section
6.5.2
Cost
Uncertainty
of
+/­
30%
by
Technology
Technology
Selections
Compliance
Schedule
Discount
Rates
Shading
scheme:

Inputs
Data
produced
from
inputs
Results
Exhibit
6.7
Methodology
for
Estimating
Treatment
Costs
Economic
Analysis
for
the
LT2ESWTR
Proposal
6­
17
June
2003
6.5.1
Toolbox
Options
and
Unit
Costs
Under
the
LT2ESWTR,
systems
will
have
a
"
toolbox"
of
treatment
and
management
options
to
select
from
to
meet
their
bin
requirements.
This
section
describes
those
options
that
plants
will
most
likely
select
to
meet
the
LT2ESWTR
requirements
and
the
typical
conditions
under
which
the
selected
technologies
may
be
installed.
The
derivation
of
capital
unit
costs
(
cost
per
plant)
and
annual
O&
M
unit
costs
(
annual
cost
per
plant)
for
each
technology
is
also
discussed.
Unit
cost
data
for
each
technology
for
CWS
plants
are
presented
in
Appendix
E.

Exhibit
6.8
summarizes
the
technologies
considered
for
estimating
benefits
and
costs
of
the
LT2ESWTR.
EPA
recognizes
that,
to
meet
all
or
part
of
the
LT2ESWTR
requirements,
some
plants
may
select
toolbox
options
that
are
not
listed
in
this
exhibit.
The
following
section
briefly
describes
each
of
the
options
not
included
and
provides
reasons
why
they
were
not
considered.
Section
6.5.1.2
discusses
the
technologies
that
were
considered
for
this
EA.

6.5.1.1
Toolbox
Options
Not
Considered
Changing
Sources
Plants
may
avoid
adding
treatment
by
connecting
to
a
nearby
water
system
or
changing
to
a
source
of
water
that
has
lower
Cryptosporidium
concentrations.
EPA
estimates
that
approximately
22
percent
of
small
systems
are
located
in
metropolitan
areas
where
distances
between
potential
connecting
water
systems
may
allow
interconnection
at
reasonable
costs
(
USEPA
2000f).
Changing
source
water
may
also
be
possible,
although
EPA
has
no
information
on
the
number
of
systems
that
may
have
less
contaminated
sources
available
to
them.
Even
if
EPA
could
estimate
how
many
systems
might
choose
these
options,
the
costs
for
both
are
highly
variable.
Costs
for
connecting
to
another
system
involve
costs
for
new
pipes
and
facilities,
and
for
those
consolidating
with
another
water
system,
costs
associated
with
merging
management
capabilities
as
well.
Costs
for
changing
to
another
source
water
would
depend
on
the
new
source
and
the
system.
Costs
could
include
such
items
as
drilling
a
new
well,
installation
of
new
piping,
and
new
treatment
facilities
or
adaptation
of
existing
ones
to
handle
different
source
quality.
The
costs
for
both
of
these
options
would
be
highly
system­
dependent
and
are
difficult
to
predict.
For
these
reasons,
these
toolbox
options
were
not
considered
in
these
cost
estimates.

Intake
Changes
Relocating
the
intake
and
managing
the
timing
or
level
of
withdrawal
are
all
toolbox
options.
The
purpose
of
these
options
is
to
change
the
location
or
timing
of
withdrawal
of
water
from
the
source
in
order
to
draw
from
those
parts
of
the
source
and
at
those
times
when
Cryptosporidium
concentrations
are
lower.
These
options
may
cost
little
compared
to
adding
treatment,
especially
for
systems
drawing
from
reservoirs.
The
costs
would
depend
on
the
existing
intake
structures
and
the
nature
of
the
source.
It
is
unknown
how
many
systems
could
likely
take
advantage
of
such
strategies
and
how
much
reduction
they
might
achieve.
Because
of
this
uncertainty,
these
options
were
not
considered
in
this
LT2ESWTR
analysis.

Performance
Studies
Some
plants
may
have
filtration
processes
that
achieve
a
higher
level
of
Cryptosporidium
removal
across
their
treatment
train
than
assumed
by
the
LT2ESWTR.
Plants
may
conduct
performance
Economic
Analysis
for
the
LT2ESWTR
Proposal
6­
18
June
2003
studies
of
their
overall
treatment
process
to
demonstrate
to
the
State
that
the
required
level
of
Cryptosporidium
treatment
is
being
achieved.
It
is
unknown,
but
likely
very
low,
as
to
the
number
of
systems
that
could
demonstrate
a
higher
level
of
treatment
on
a
consistent
basis.
Also
the
cost
of
such
studies
could
be
higher
than
implementing
another
toolbox
option.

The
toolbox
options
that
were
omitted
from
the
costs
analysis
may
be
less
expensive
than
the
technologies
considered
and,
therefore,
the
costs
presented
here
may
be
overestimated.
Cost
uncertainties
for
this
analysis
are
summarized
in
section
6.12.

6.5.1.2
Technologies
Considered
for
the
LT2ESWTR
Cost
Analysis
Exhibit
6.8
shows
technologies
that
were
included
in
the
cost
analysis
for
the
LT2ESWTR.
The
second
column
summarizes
the
condition(
s)
under
which
the
technology
use
is
constrained
in
this
EA.
Plants
may
be
constrained
from
installing
a
technology
for
various
reasons.
During
the
Small
Surface
Water
Delphi
process,
industry
experts
and
the
Technical
Work
Group
(
TWG)
identified
limitations
in
the
use
of
several
technologies.
A
more
extensive
explanation
of
these
groups
and
their
conclusions
can
be
found
in
the
Stage
2
DBPR
Economic
Analysis
(
USEPA
2002b).

The
third
column
in
Exhibit
6.8
identifies
the
conditions
for
which
cost
estimates
are
developed.
To
capture
the
range
of
costs,
many
technologies
are
evaluated
over
a
range
of
possible
influent
water
qualities
and
operating
conditions.
For
the
purposes
of
estimating
the
costs
of
the
LT2ESWTR,
the
TWG
selected
the
water
quality
and
operating
parameters
to
capture
the
"
normal"
circumstances
under
which
plants
may
use
the
technology.
EPA
does
not
assume
that
all
systems
would
operate
under
these
conditions,
but
that
the
cost
equations
generate
capital
and
O&
M
costs
that
are
typical
for
the
range
of
system
types
and
sizes.
While
these
assumptions
simplify
the
true
variety
of
operating
conditions,
EPA
believes
they
capture
reasonable
estimates
of
national
costs.

EPA
recognizes
that
similar
systems
may
experience
different
capital
or
O&
M
costs
due
to
site­
specific
factors.
Inputs
to
unit
costs
such
as
water
quality
conditions,
labor
rates,
and
land
costs
can
be
highly
variable
and
increase
the
system­
to­
system
variability
in
unit
costs.
In
developing
the
unit
cost
estimates,
there
is
insufficient
information
to
fully
characterize
what
the
distribution
of
this
variability
will
be
on
a
national
scale
for
all
of
the
treatments
and
all
possible
conditions.

Instead,
the
unit
costs
are
developed
as
average
or
representative
estimates
of
what
these
unit
costs
will
be
nationally.
That
is,
in
developing
unit
costs,
design
criteria
for
the
technologies
are
selected
to
represent
typical,
or
average,
conditions
for
the
universe
of
systems.
As
a
result,
there
is
uncertainty
inherent
in
these
unit
cost
estimates
reflecting
the
fact
that
they
are
based
on
independent
assumptions
with
supporting
data
and
vendor
quotes,
where
available,
rather
than
on
an
detailed
aggregation
of
State,
regional,
or
local
estimates
based
on
actual
field
conditions.
In
this
EA,
uncertainty
in
these
national
average
unit
costs
factors
is
characterized
as
follows:

°
Capital
costs:
+/­
30%

°
O&
M
costs:
+/­
15%

These
estimates
were
developed
by
EPA
and
reflect
information
presented
by
the
National
Drinking
Water
Advisory
Council
(
2001)
in
their
review
of
the
national
cost
estimation
methodology
for
Economic
Analysis
for
the
LT2ESWTR
Proposal
6­
19
June
2003
the
Arsenic
Rule.
EPA
believes
that
the
uncertainties
in
capital
and
O&
M
costs
for
a
given
technology
are
independent
of
one
another,
and
that
uncertainties
across
all
technologies
are
independent.

The
final
column
in
Exhibit
6.8
provides
the
source
for
the
unit
costs.
Many
unit
costs
are
derived
from
equations
and
other
information
in
the
Technology
and
Cost
document
(
USEPA
2003a).
For
other
technologies,
unit
costs
were
provided
by
the
FACA
committee
and
its
TWG
in
June
2000
as
part
of
the
Stage
2
M­
DBP
negotiation
process
or
are
estimated
from
earlier
EPA
rules.

In
most
cases,
technology
unit
costs
are
specified
for
single
average
daily
and
design
flows
for
nine
system
size
categories.
In
reality,
there
will
be
a
range
of
unit
costs
for
a
category,
relative
to
the
range
of
flows.
EPA
believes,
however,
that
using
a
mean
unit
cost
($/
plant)
per
category,
derived
from
mean
flow
data,
provides
an
accurate
representation
of
total
national
costs.
For
technologies
that
are
a
combination
of
two
or
more
unit
processes
(
e.
g.,
cartridge
filters
with
ozone),
the
technology
unit
costs
are
simply
assumed
to
be
the
sum
of
the
costs
for
each
unit
process.
This
may
result
in
the
overestimation
of
certain
unit
costs
since
some
economies
of
scale
are
expected
when
such
"
combined"
technologies
are
implemented.
Economic
Analysis
for
the
LT2ESWTR
Proposal
6­
20
June
2003
Exhibit
6.8
Advanced
Technologies
Considered
for
the
LT2ESWTR
Technology
Constraints
[
1]
Water
Quality
and
Operational
Parameters
Source
of
Unit
Cost
[
2]

Bag
Filtration
Not
practical
for
systems
serving
more
than
10,000
people
Bag
replacement
four
times
per
year
Figures
D­
19
and
D­
20
from
T&
C
Document
Cartridge
Filtration
Not
practical
for
systems
serving
more
than
10,000
people
Cartridge
replacement
twice
per
year
Figures
D­
21
and
D­
22
from
T&
C
Document
Chlorine
Dioxide
(
ClO2)
Not
practical
for
systems
serving
500
people
or
fewer
Concentration
=
1.25
mg/
l
No
additional
contact
basins
Figures
D­
5
and
D­
6
from
T&
C
Document
Combined
Filter
Performance
None
None
Figures
D­
30
and
D­
31
from
T&
C
document
Individual
Filter
Performance
None
None
Not
Costed
Bank
Filtration
(
In­
Bank)
Not
practical
for
systems
serving
fewer
than
10,000
people
None
Figure
D­
23
from
T&
C
document
Microfiltration/
Ultrafiltration
(
MF/
UF)
None
0.3
NTU,
10
°
C,
disposal
of
reject
stream
to
sewer
Figures
D­
17
and
D­
18
from
T&
C
Document
Ozone
(
O3)
Not
practical
for
systems
serving
500
people
or
fewer
Concentration
and
contact
time
to
achieve
0.5,
1.0,
and
2.0
log
of
Cryptosporidium
inactivation
based
on
3­
dimensional
regressions
of
concentration,
flow,
and
cost.
Concentration
is
the
mean
concentration
from
SWAT
analysis.
Figure
D­
11
through
Figure
D­
16
from
T&
C
Document
Secondary
Filter
(
SF)
Not
practical
for
systems
serving
500
people
or
fewer
None
Figures
D­
24
and
D­
25
from
T&
C
document
Economic
Analysis
for
the
LT2ESWTR
Proposal
6­
21
June
2003
Exhibit
6.8
(
continued)
Advanced
Technologies
Considered
for
the
LT2ESWTR
Technology
Constraints
[
1]
Water
Quality
and
Operational
Parameters
Source
of
Unit
Cost
[
2]

UV
[
3]
None
Median
water
quality
parameters:
UV
254
=
0.051cm­
1,
turbidity
=
0.1
NTU,
alkalinity
=
60mg/
L
as
CaCO3,
hardness
=
100mg/
L
as
CaCO3
Figures
D­
7
and
D­
8
from
T&
C
Document
Watershed
Control
(
WC)
Not
practical
for
systems
serving
10,000
people
or
fewer
None
Figures
D­
28
and
D­
29
from
T&
C
document
[
1]
Constraints
identified
by
the
TWG
and
the
Small
Surface
Water
Delphi
Group.
[
2]
Unit
costs
($/
plant)
for
CWSs
provided
in
Appendix
E
for
each
technology
shown.
[
3]
Patent
fee
of
$
0.015/
1,000gallons
included
for
household
costs,
see
section
6.10
for
rationale.
Economic
Analysis
for
the
LT2ESWTR
Proposal
6­
22
June
2003
6.5.2
Compliance
Forecast
and
Technology
Selection
Three
key
inputs
are
required
to
estimate
the
technologies
plants
will
add
to
comply
with
LT2ESWTR:

1.
The
percent
of
plants
that
must
make
a
treatment
change
to
meet
the
LT2ESWTR
requirements.

2.
The
treatment
technologies
these
plants
have
in
place
prior
to
implementation
of
the
rule.

3.
The
treatment
technologies
these
plants
are
predicted
to
select
to
comply
with
the
rule.

These
inputs,
coupled
with
baseline
data
presented
in
Chapter
4,
provide
an
estimate
of
the
total
number
of
filtered
plants
using
specific
technologies
to
meet
the
requirements
of
the
LT2ESWTR.
Input
1
above
is
largely
a
result
of
system
"
binning,"
the
treatment
bin
into
which
the
system
is
classified
based
on
initial
source
water
Cryptosporidium
monitoring.
The
ICR,
ICRSSL,
and
ICRSSM
modeled
Cryptosporidium
occurrence
distributions
are
used
to
determine
the
binning
for
all
size
systems,
resulting
in
three
possible
binning
scenarios
for
each
rule
alternative.

The
second
input
accounts
for
plants
that
already
have
some
of
the
toolbox
technologies
in
place
and
will
be
able
to
obtain
credit
for
Cryptosporidium
treatment
without
any
additional
costs
or
obtain
a
portion
of
their
required
treatment
credit.
Therefore,
the
existing
technologies
cannot
be
considered
as
a
possible
technology
selection
for
these
plants.
To
accommodate
the
two
treatment
baseline
scenarios,
two
separate
technology
selection
forecasts
are
used
for
plants
with
existing
toolbox
technologies
and
those
without.
These
two
forecasts
follow
the
same
methodology
as
described
below,
with
the
former
omitting
the
existing
technologies
from
the
selection
process.

The
overall
methodology
used
to
develop
the
technology
selection
forecast
in
the
third
input
is
based
on
"
least­
cost."
EPA
assumes
that
the
cost
of
the
rule
is
best
estimated
by
assuming
that
drinking
water
plants
will
select
the
least
expensive
technology
or
combination
of
technologies
available
to
meet
the
treatment
requirements
of
a
given
action
bin.
The
technology
selections
are
constrained
by
maximum
use
percentages,
which
recognize
that
all
plants
may
not
be
able
to
implement
certain
technologies
because
of
site­
specific
conditions.
Appendix
F
details
the
methodology
for
developing
the
technology
selection
forecast
and
the
assumptions
involved.

Technology
selection
results
(
i.
e.,
the
number
of
plants
adding
specific
technologies)
are
derived
using
mean
distributions
drawn
from
models
of
the
ICR,
ICRSSL,
and
the
ICRSSM
Cryptosporidium
occurrence
data
sets.
Exhibit
6.9
shows
the
selections
for
the
Preferred
Alternative,
derived
from
the
ICR
(
high)
and
ICRSSL
(
low)
occurrence
distributions.
Appendix
G
provides
results
for
all
regulatory
alternatives
and
sensitivity
analyses.
Many
plants
end
up
selecting
technologies
that
achieve
greater
removal
than
required
by
their
bin
assignments.
In
many
cases,
the
least
costly
technology
results
in
higher
levels
of
Cryptosporidium
inactivation
or
removal
than
required
for
that
bin
(
this
is
always
the
case
when
UV
is
selected).
Although
direct
filtration
plants
have
0.5
log
higher
bin
requirements
than
conventional
and
other
filtration
plants,
no
additional
treatment
is
estimated
for
them.
The
0.5
log
higher
requirement
for
those
plants
are
adequately
addressed
by
the
higher
levels
of
Cryptosporidium
treatment
achieved
by
the
selected
technologies.
Economic
Analysis
for
the
LT2ESWTR
Proposal
6­
23
June
2003
ICR
ICRSSL
ICRSSM
ICR
ICRSSL
ICRSSM
Bag
Filter
1.0
Log
1,545
1,236
1,441
O3
0.5
Log
26
17
21
Cartridge
Filter
2.0
Log
190
17
52
O3
1.0
Log
24
18
21
CLO2
0.5
Log
77
60
70
O3
2.0
Log
9
1
2
Combined
Filter
Performance
0.5
Log
16
12
14
Secondary
Filter
1.0
Log
0
0
0
In­
bank
Filtration
1.0
Log
5
3
4
UV
2.5
Log
998
490
632
MF/
UF
2.5
Log
10
3
5
WS
Control
0.5
Log
0
0
0
Total
Plants
Selecting
Technologies
[
2]
2,893
1,852
2,255
Technology
Selections
[
1]
Data
Set
Data
Set
Technology
Selections
[
1]
Exhibit
6.9
Technology
Selection
Forecast
for
Filtered
Plants
Notes:
[
1]
Selection
includes
non­
purchased
plants
in
CWSs,
NTNCWSs,
and
TNCWSs
adding
treatment
and
purchased
plants
that
could
not
be
linked
with
their
sellers.
[
2]
Some
plants
select
more
than
one
technology
to
meet
the
bin
requirements:
therefore,
total
technologies
selected
is
more
than
the
number
of
plants
adding
treatment.
Source:
Appendix
G.
ICR
Data
from
Exhibits
G.
37­
39.
ICRSSL
Data
from
Exhibits
G.
43­
45.
ICRSSM
Data
from
Exhibits
G.
49­
51.

6.5.3
Capital
and
Annual
Costs
To
estimate
the
treatment
costs
for
filtered
plants,
the
technology
unit
costs
(
capital
and
O&
M)
are
multiplied
by
the
number
of
plants
predicted
for
each
technology
and
size
category.
The
O&
M
costs
are
costs
that
systems
incur
yearly
to
maintain
system
performance.
The
capital
costs
are
adjusted
to
account
for
the
life­
span
of
the
capital
equipment
(
assumed
to
be
20
years)
that
is
not
included
in
the
time
period
of
this
economic
analysis
(
years
2003
through
2027).
Appendix
O,
Exhibit
O.
2
shows
the
schedule
of
implementation
used
to
determine
when
systems
will
install
the
equipment.
For
example,
10
percent
of
large
systems
are
estimated
to
begin
incurring
capital
costs
for
treatment
in
year
2007.
For
these
systems
the
capital
costs
were
reduced
by
a
factor
of
4/
20.

The
treatment
costs
for
CWS,
NTNCWS,
and
TNCWS
plants
are
calculated
separately
since
each
type
has
different
population
per
system
averages,
producing
different
mean
flows
(
see
Exhibit
4.4a)
and,
thus,
different
mean
plant
unit
costs.
These
costs
are
summed
across
technologies
and
size
categories
to
estimate
the
total
treatment
costs
for
LT2ESWTR.
They
are
incorporated
into
the
rule
schedule
(
Appendix
O),
converted
to
2000
dollars,
and
annualized
using
discount
rates
of
3
and
7
percent.
Exhibit
6.10
shows
the
results
for
the
Preferred
Alternative
based
on
the
ICR,
ICRSSL,
and
ICRSSM
occurrence
distributions.
Economic
Analysis
for
the
LT2ESWTR
Proposal
6­
24
June
2003
Data
Set
System
Size
(
population
served)
Present
Value
Capital
Costs
at
3%
Present
Value
Capital
Costs
at
7%
Annualized
O&
M
Costs
at
3%
Annualized
O&
M
Costs
at
7%
Total
Annualized
Costs
at
3%
Total
Annualized
Costs
at
7%
A
B
C
D
E
F
ICR
£
10,000
$
76.1
$
56.0
$
5.2
$
4.3
$
9.6
$
9.1
>
10,000
$
1,092.4
$
868.0
$
26.1
$
22.7
$
88.8
$
97.1
TOTAL
$
1,168.5
$
924.0
$
31.3
$
26.9
$
98.4
$
106.2
ICRSSL
£
10,000
$
42.8
$
31.5
$
2.9
$
2.4
$
5.3
$
5.1
>
10,000
$
707.1
$
561.8
$
16.2
$
14.0
$
56.8
$
62.3
TOTAL
$
749.8
$
593.3
$
19.0
$
16.4
$
62.1
$
67.3
ICRSSM
£
10,000
$
52.6
$
38.7
$
3.5
$
2.9
$
6.6
$
6.2
>
10,000
$
842.4
$
669.3
$
19.4
$
16.9
$
67.8
$
74.3
TOTAL
$
894.9
$
708.0
$
23.0
$
19.8
$
74.4
$
80.6
Exhibit
6.10
Treatment
Costs
for
Filtered
and
Unfiltered
Plants,
Preferred
Alternative
($
Millions,
2000$)

Note:
Detail
may
not
add
to
totals
due
to
independent
rounding.
Sources:
Appendix
O.
[
A]
Exhibit
O.
4b;
Rows
­
Mean,
A3,
ICR,
ICRSSL,
and
ICRSSM;
Columns
A,
C,
G,
and
I.
[
B]
Exhibit
O.
4c;
Rows
­
Mean,
A3,
ICR,
ICRSSL,
and
ICRSSM;
Columns
A,
C,
G,
and
I.
[
C]
Exhibit
O.
4d;
Rows
­
Mean,
A3,
ICR,
ICRSSL,
and
ICRSSM;
Columns
B,
D,
H,
and
J.
[
D]
Exhibit
O.
4e,
Rows
­
Mean,
A3,
ICR,
ICRSSL,
and
ICRSSM;
Columns
B,
D,
H,
and
J.
[
E]
Exhibit
O.
4d;
Rows
­
Mean,
A3,
ICR,
ICRSSL,
and
ICRSSM;
Columns
A,
B,
C,
D,
G,
H,
I,
and
J.
[
F]
Exhibit
O.
4e;
Rows
­
Mean,
A3,
ICR,
ICRSSL,
and
ICRSSM;
Columns
A,
B,
C,
D,
G,
H,
I,
and
J.

6.6
Costs
of
Treatment
for
Unfiltered
Plants
As
summarized
in
Chapter
1,
the
LT2ESWTR
requires
all
unfiltered
plants
to
achieve
2
log
Cryptosporidium
inactivation
if
their
source
water
concentration
is
less
than
or
equal
to
1
oocyst
per
100
liters,
and
3
log
inactivation
if
it
is
greater
than
1
oocyst
per
100
liters.
UV
is
the
least
expensive
technology
that
can
achieve
the
required
log
inactivation
of
Cryptosporidium
and,
therefore,
most
likely
to
be
installed.
However,
as
with
filtered
systems,
EPA
estimated
that
a
small
percentage
of
plants
would
elect
to
install
a
technology
more
expensive
than
UV
due
to
the
configuration
of
existing
equipment
or
other
factors.
Ozone
is
the
next
cheapest
technology
that
will
meet
the
inactivation
requirements,
and
therefore,
is
estimated
to
be
used
by
plants
that
cannot
use
UV.
Due
to
the
high
concentrations
of
ozone
necessary
to
inactivate
Cryptosporidium,
EPA
estimated
that
ozone
can
achieve
only
2.0
log
inactivation
and
all
systems
that
are
required
to
obtain
3.0
log
inactivation
are
assumed
to
use
UV.
Although
the
toolbox
lists
UV
as
achieving
2.5
log
inactivation,
studies
have
shown
that
it
can
achieve
much
greater
inactivation
at
the
standard
drinking
water
dose
of
40
mJ/
cm2
used
in
the
cost
analysis
(
Clancy
et
al.
2000;
Craik
et
al.
2001).
MF/
UF
was
considered
as
a
substitute
for
UV
in
this
analysis,
but
is
much
more
expensive
than
these
technologies,
especially
for
larger
systems.

Unit
costs
for
UV
and
ozone
and
the
conditions
under
which
they
can
be
used
are
the
same
as
for
the
filtered
plants
(
section
6.5).
Ozone
unit
costs
are
for
an
ozone
concentration
that
provides
2.0
log
of
Cryptosporidium
inactivation.
Economic
Analysis
for
the
LT2ESWTR
Proposal
6­
25
June
2003
All
unfiltered
plants
must
meet
the
requirements
of
the
LT2ESWTR;
therefore,
the
percent
of
such
plants
adding
technology
is
100
percent.
The
following
assumptions
were
used
in
developing
the
technology
selection
forecast
for
plants
needing
2.0
log
Cryptosporidium
inactivation:

°
100
percent
of
very
small
systems
will
use
UV.

°
90
percent
of
other
plants
will
be
able
to
install
UV
(
the
least
expensive
of
the
two
technologies).

°
10
percent
of
other
plants
will
add
ozone.

Consistent
with
assumptions
for
filtered
systems,
very
small
plants
were
also
assumed
to
be
unable
to
use
ozone.
This
was
because
of
the
high
level
of
operator
attention
and
training
required
to
operate
and
maintain
an
ozone
system.
Because
of
the
high
costs
for
alternatives
like
MF/
UF
for
very
small
plants,
the
analysis
assumes
that
all
these
plants
would
find
a
way
to
use
UV.
EPA
believes
this
to
be
a
reasonable
assumption
because
small
plants
have
less
existing
infrastructure
that
might
limit
their
technology
selection.
Exhibit
6.10
summarizes
the
present
value
and
annualized
treatment
costs
at
3
and
7
percent
discount
rates.

6.7
Costs
for
Benchmarking
and
Technology
Reporting
Requirements
The
benchmarking
requirement
was
introduced
in
the
IESWTR
and
LT1ESWTR.
For
the
LT2ESWTR,
it
requires
any
system
proposing
a
change
to
its
disinfection
process
to
complete
an
evaluation
of
the
existing
process
and
consult
with
the
State/
Primacy
Agency
about
how
the
proposed
change
will
affect
the
current
disinfection
performance.
Estimated
costs
for
both
systems
and
States
reflect
the
time
to
conduct
an
evaluation
and
consult
with
the
State.
This
EA
assumes
benchmarking
costs
will
be
incurred
for
only
those
systems
estimated
to
make
changes
in
treatment
and
by
the
States/
Primacy
Agencies.
The
present
value
of
cost
at
a
3
percent
discount
rate
using
the
modeled
ICR
occurrence
distribution
is
estimated
to
be
$
59,000
and
$
115,000,
for
States
and
systems,
respectively
(
Appendix
O,
Exhibit
O.
3b
 
systems
and
Exhibit
O.
5b
 
States).
At
a
3
percent
discount
rate
using
the
modeled
ICRSSL
occurrence
distribution
the
values
are
$
32,000
and
$
64,000
for
States
and
systems,
respectively.
Appendix
D,
Exhibits
D.
25
 
D.
27
for
systems
and
Exhibits
D.
28
for
States/
Primacy
Agencies,
show
the
derivation
of
system
and
States/
Primacy
Agency
costs
for
benchmarking.

Systems
required
to
meet
additional
Cryptosporidium
treatment
will
also
have
to
report
compliance
monitoring
data,
depending
on
the
toolbox
option
implemented.
Appendix
D,
Exhibits
D.
21
 
D.
24
for
systems
and
Exhibits
D.
18
 
D.
20
for
States/
Primacy
Agencies,
show
the
derivation
of
system
and
States/
Primacy
Agency
costs
and
total
for
technology
compliance
reporting.

6.8
Costs
of
Treatment
for
Uncovered
Finished
Water
Reservoirs
As
part
of
the
LT2ESWTR,
systems
with
uncovered
finished
water
reservoirs
have
the
options
to
cover
the
reservoirs,
to
provide
disinfection
downstream
of
the
reservoir,
or
to
implement
a
Stateapproved
risk
management
plan.
Disinfection
alternatives
must
achieve
at
least
4
log
of
virus
inactivation.
To
develop
national
cost
estimates
for
systems
to
comply
with
this
provision
of
the
LT2ESWTR,
unit
costs
for
each
treatment
alternative
and
the
percentage
of
systems
selecting
each
alternative
are
Economic
Analysis
for
the
LT2ESWTR
Proposal
6­
26
June
2003
estimated
for
the
inventory
of
systems
having
uncovered
finished
water
reservoirs
(
presented
in
Chapter
4).
This
section
describes
the
methodology
for
developing
the
unit
costs
of
reservoir
covers
and
disinfection.
The
basis
for
estimating
the
number
of
systems
that
select
each
alternative
is
then
discussed.
A
full
description
of
the
unit
costs,
along
with
assumptions,
is
presented
in
Appendix
I.

6.8.1
Unit
Costs
There
are
two
types
of
reservoir
covers
 
fixed
and
floating.
The
unit
costs
for
both
types
are
estimated
from
information
provided
in
the
Uncovered
Finished
Water
Reservoirs
Guidance
Manual
(
USEPA
1999c).

Fixed
covers
are
commonly
constructed
of
concrete,
steel,
or
aluminum.
Floating
covers
are
flexible
membrane
structures
generally
made
of
polypropylene
or
similar
material.
Unit
costs
are
based
on
the
estimated
surface
area
of
each
reservoir
and
the
average
cost
of
materials.
Appendix
I
summarizes
the
unit
costs.

Systems
have
the
option
to
add
a
disinfectant
after
the
reservoir
instead
of
installing
a
cover.
A
review
of
available
disinfection
options
showed
that
gas
chlorine
is
the
least­
cost
disinfectant
that
can
achieve
at
least
4
log
of
virus
inactivation
(
USEPA
2001d).
Unit
costs
of
gas
chlorination
presented
in
Appendix
I
are
a
function
of
flow
and
include
the
costs
of
typical
process
equipment
and
the
chemical
building.
Note
that
unit
costs
for
the
smallest
size
category
are
high
due
to
fixed
(
size­
independent)
equipment
costs.

6.8.2
Compliance
Forecast
and
Technology
Selection
EPA
conservatively
assumed
that
every
system
having
an
uncovered
finished
water
reservoir
will
install
a
cover
or
treat
its
discharge.
Some
systems
with
uncovered
finished
water
reservoirs
may
be
allowed
to
implement
a
risk
management
plan
instead
of
treating
or
covering.
It
is
quite
likely
that
many
systems,
perhaps
even
most,
would
be
able
to
obtain
State
approval
of
a
risk
management
plan.
If
so,
estimated
costs
could
be
well
below
actual
costs
or
even
negligible
if
this
approach
is
widely
used.
EPA
had
no
basis,
however,
for
making
an
assumption
about
how
many
States
could
or
would
take
this
approach.

The
technology
selection
methodology
for
the
uncovered
finished
water
reservoirs
also
uses
a
least­
cost
approach.
For
systems
with
reservoir
capacities
of
5
million
gallons
(
MG)
or
less,
covering
is
the
least
expensive
alternative.
Although
chlorination
is
the
least
expensive
approach
for
the
remaining
systems,
the
ability
of
a
system
to
use
booster
chlorination
depends
on
its
current
residual
disinfectant
type.
Approximately
50
percent
of
all
surface
water
systems
are
predicted
to
use
chloramination
following
implementation
of
the
Stage
2
DBPR.
Adding
chlorine
to
water
treated
with
chloramines
can
cause
quality
problems;
therefore,
a
maximum
of
50
percent
of
systems
were
assumed
to
add
booster
chlorination
after
the
reservoir.
UV
was
not
considered
a
practical
option
because
the
dose
requirements
for
virus
inactivation
are
high,
making
the
capital
and
O&
M
costs
higher
than
for
chlorine.
The
technology
selection
for
uncovered
finished
water
reservoirs
is
presented
in
Exhibit
6.11.
Economic
Analysis
for
the
LT2ESWTR
Proposal
6­
27
June
2003
Exhibit
6.11
Technology
Selection
for
Uncovered
Finished
Water
Reservoirs
Size
Category
(
MG)
Number
of
Uncovered
Reservoirs
Floating
Cover
(%)
Booster
Chlorination
(%)

0­
0.1
25
100
­

0.1­
1
7
100
­

>
1­
5
44
100
­

>
5­
10
12
100
­

>
10­
20
10
100
­

>
20­
40
9
50
50
>
40­
60
4
50
50
>
60­
80
4
50
50
>
80­
100
6
50
50
>
100­
150
6
50
50
>
150­
200
2
50
50
>
200­
250
4
50
50
>
250­
1,000
4
50
50
>
1,000
1
50
50
Source:
Appendix
I,
Exhibit
I.
7.

Because
the
technology
selection
is
based
on
least
costs,
and
fixed­
cover
costs
are
the
most
expensive
treatment
option
considered,
no
systems
were
assumed
to
install
fixed
covers.
EPA
recognizes
that
some
systems
may
select
fixed
covers
for
other
reasons,
but
these
incremental
costs
are
not
attributable
to
this
rule.

6.8.3
Total
Annual
Treatment
Costs
Total
annual
treatment
costs
are
calculated
by
multiplying
the
number
of
reservoirs
in
a
category
by
the
percent
selecting
a
technology
and
the
unit
cost
of
that
technology.
Exhibit
6.12
summarizes
costs
for
all
reservoirs
for
each
size
category.
Economic
Analysis
for
the
LT2ESWTR
Proposal
6­
28
June
2003
Capital
O&
M
Total
Capital
O&
M
Total
 
£
10,000
$
3,520
$
1,649
$
5,169
$
4,713
$
1,552
$
6,264
>
10,000
$
3,349,320
$
2,046,425
$
5,395,745
$
4,483,927
$
1,925,203
$
6,409,129
Total
$
3,352,840
$
2,048,074
$
5,400,915
$
4,488,639
$
1,926,754
$
6,415,393
Annualized
Cost
at
3%
Annualized
Cost
at
7%
System
Size
(
Population
Served)

System
Size
ICR
(
3%)
ICR
(
7%)
ICRSSL
(
3%)
ICRSSL
(
7%)
ICRSSM
(
3%)
ICRSSM
(
7%)
A
B
C
D
E
F
£
 
10K
23.5
$
14.3
$
18.4
$
11.3
$
20.7
$
12.6
$
>
10k
14.4
$
9.8
$
16.4
$
11.2
$
15.6
$
10.7
$
Total
37.8
$
24.1
$
34.8
$
22.5
$
36.3
$
23.3
$
Exhibit
6.12
Total
Annualized
Cost
for
Uncovered
Finished
Water
Reservoirs
(
2000$)

Note:
Detail
may
not
add
to
totals
due
to
independent
rounding.
Source:
Appendix
O.
Annualized
Cost
at
3%
from
Exhibit
O.
4d;
Row
­
Mean,
A3,
ICR;
Columns
E,
F,
K
and
L.
Annualized
Cost
at
7
%
from
Exhibit
O.
4e;
Row
­
Mean,
A3,
ICR;
Columns
E,
F,
K
and
L.

6.9
Future
Source
Water
Monitoring
Six
years
after
initial
bin
classification,
filtered
plants
will
be
required
to
conduct
a
second
round
of
monitoring
to
reassess
source
water
conditions
for
bin
assignments.
EPA
will
evaluate
new
analytical
methods
and
surrogate
indicators
of
Cryptosporidium
in
the
interim.
While
the
costs
of
monitoring
are
likely
to
change
in
the
6
years
following
rule
promulgation,
it
is
difficult
to
predict
how
they
will
change.
In
the
absence
of
any
other
information,
it
was
assumed
that
the
laboratory
costs
would
be
the
same
as
for
the
initial
monitoring.
All
plants
that
conducted
initial
monitoring
were
assumed
to
conduct
future
monitoring
as
well,
except
for
those
systems
that
achieve
5.5
log
Cryptosporidium
treatment
credit.
Exhibit
6.13
shows
the
costs
for
future
monitoring.
Costs
vary
among
the
Cryptosporidium
occurrence
data
sets
because
the
numbers
of
plants
that
add
technologies
and
achieve
5.5
log
treatment
credit
differ.
Also,
different
numbers
of
small
plants
are
triggered
into
monitoring
for
Cryptosporidium.
Appendix
D,
Exhibits
D.
31­
D.
34
show
the
calculations
for
the
cost
estimates.

Exhibit
6.13
Present
Value
Costs
for
Future
Monitoring,
Preferred
Alternative
Note:
Detail
may
not
add
to
totals
due
to
independent
rounding.
Sources:
Appendix
O.
[
A]
and
[
C]
Exhibit
O.
3b;
Rows
­
A3,
ICR
and
ICRSSL;
Columns
D
and
E,
and
I.
[
B]
and
[
D]
Exhibit
O.
3c;
Rows
­
A3,
ICR
and
ICRSSL;
Columns
D
and
E,
and
I.
Economic
Analysis
for
the
LT2ESWTR
Proposal
6­
29
June
2003
6.10
Household
Costs
EPA
assumes
that
systems
will
pass
some
or
all
of
the
costs
of
a
new
regulation
on
to
their
customers
in
the
form
of
rate
increases.
Household
costs,
which
are
in
units
of
$
per
household
per
year,
are
estimated
in
this
chapter
to
provide
a
measure
of
the
increase
in
water
bills
that
may
result
from
the
LT2ESWTR.
Exhibit
6.14
presents
the
mean
expected
increases
in
yearly
household
costs
by
system
size,
system
type,
and
occurrence
data
set,
for
those
systems
subject
to
the
rule.
(
Appendix
J,
Exhibit
J.
4
presents
household
cost
estimates
for
those
systems
predicted
to
make
treatment
changes.)

These
cost
increases
incorporate
the
costs
of
rule
implementation
(
e.
g.,
reading
and
understanding
the
rule),
initial
and
future
monitoring
for
bin
classification,
fixing
uncovered
finished
water
reservoirs,
treatment
changes,
benchmarking,
and
compliance
reporting.
A
detailed
description
of
the
derivation
of
per­
household
costs
is
in
Appendix
J.
Per­
household
costs
for
uncovered
finished
water
reservoirs
are
determined
by
taking
the
costs
for
fixing
the
reservoirs
from
section
6.8
and
assigning
them
to
systems
as
described
in
section
4.6.

To
annualize
capital
costs
when
determining
the
costs
to
households,
EPA
uses
different
discount
rates
for
private
and
public
systems
and
for
systems
of
different
sizes.
The
rate
differences
between
systems
represent
the
different
borrowing
sources
each
type
of
system
has
available
to
it,
differences
in
risk
and
expectations
regarding
inflation.
The
rates
vary
from
5.20
to
6.27
percent
depending
on
system
size
and
ownership,
and
are
summarized
in
Appendix
J,
Exhibit
J.
1.
Per­
household
costs
also
include
costs
for
royalty
payments
on
the
use
of
UV
light
(
described
below).

The
unit
costs
for
treatment
in
dollars
per
thousand
gallons
is
then
multiplied
by
the
annual
perhousehold
usage
rate
to
obtain
their
contribution
to
per­
household
costs.
Although
rule
implementation
and
monitoring
represent
relatively
small,
one­
time
costs,
they
have
been
included
in
the
analysis
to
provide
a
complete
distribution
of
the
potential
per­
household
cost
increase.

EPA
has
learned
that
Calgon
Carbon
Corporation
holds
a
patent
(
No.
6,129,893)
for
"
A
method
for
prevention
of
Cryptosporidium
oocysts
and
similar
organisms
in
water
by
irradiating
the
water
with
ultraviolet
light
in
a
range
of
200
to
300
nm
in
concentrations
of
about
10
mJ/
cm2
to
about
175
mJ/
cm2."
It
is
EPA's
understanding
that
this
patent
would
apply
to
systems
using
medium­
pressure
mercury
vapor
lamps
to
inactivate
Cryptosporidium.
EPA
also
understands
that
Calgon
has
applied
for
a
continuation
in
part
to
extend
coverage
of
the
patent
to
low­
pressure
mercury
vapor
lamps
and
to
lower
UV
concentrations.
Calgon
has
informed
EPA
that
the
company
intends
to
charge
a
license
fee
of
$
0.015/
1,000
gallons
treated
to
water
producers
using
UV
under
conditions
covered
by
the
patent.
This
cost
was
added
to
the
unit
cost
of
UV
in
the
per­
household
cost
calculations.
It
was
not
used
in
the
national
cost
estimates
because
it
represents
a
transfer
payment
involving
no
use
of
resources.

For
purchased
systems
that
are
linked
to
larger,
nonpurchased
systems,
the
per­
household
costs
are
calculated
using
the
unit
costs
of
the
larger
system,
however
they
are
reported
within
the
size
category
distributions
for
the
purchased
system.
Household
costs
for
these
purchased
systems
are
based
on
the
per­
household
usage
rates
appropriate
for
the
retail
system
and
not
the
system
selling
the
water.
This
reflects
the
fact
that
although
purchased
systems
will
not
face
increased
costs
from
adding
their
own
treatment,
whatever
costs
the
wholesale
utility
incurs
would
likely
be
passed
on
as
higher
water
costs.
Economic
Analysis
for
the
LT2ESWTR
Proposal
6­
30
June
2003
System
Type/
Size
Households
Mean
Median
90th
Percentile
95th
Percentile
Percent
of
Systems
with
Household
Cost
Increase
<
$
12
Percent
of
Systems
with
Household
Cost
Increase
<
$
120
All
CWS
65,816,979
$
1.68
$
0.13
$
4.06
$
7.57
98.37%
99.99%
All
CWS
 
£
10,000
3,318,012
$
4.61
$
1.34
$
13.04
$
14.92
87.88%
99.88%

All
CWS
65,816,979
$
1.07
$
0.03
$
3.24
$
5.43
99.31%
100.00%
All
CWS
 
£
10,000
3,318,012
$
2.68
$
0.80
$
6.10
$
9.39
95.71%
99.95%

All
CWS
65,816,979
$
1.28
$
0.03
$
3.48
$
6.47
99.07%
100.00%
All
CWS
 
£
10,000
3,318,012
$
3.27
$
0.80
$
6.62
$
13.04
93.90%
99.93%

All
CWS
65,816,978
$
1.83
$
0.14
$
4.51
$
7.58
98.25%
99.99%
All
CWS
 
£
10,000
3,318,012
$
4.98
$
1.43
$
13.50
$
15.78
87.15%
99.87%

All
CWS
65,816,979
$
0.73
$
0.03
$
2.81
$
3.84
99.60%
100.00%
All
CWS
 
£
10,000
3,318,012
$
1.76
$
0.46
$
5.37
$
6.10
97.51%
99.98%
All
Systems
­
Low
All
Systems
­
ICR
All
Systems
­
ICRSSL
All
Systems
­
ICRSSM
All
Systems
­
High
Exhibit
6.14
Summary
of
Annual
Per­
Household
Cost[
1]
Increases,
Preferred
Alternative
($/
Year)

Note:
[
1]
Annualized
at
discount
rates
varied
by
system
size
and
ownership
(
see
Appendix
J,
Exhibit
J.
2).
[
2]
Households
served
by
systems
subject
to
the
LT2ESWTR.
Source:
Appendix
J,
Exhibit
J.
3.

EPA
estimates
that
all
households
served
by
surface
and
GWUDI
sources
will
face
some
increase
in
costs
due
to
implementation
of
the
LT2ESWTR
(
except
for
those
few
who
have
already
installed
5.5
log
of
treatment
for
Cryptosporidium;
see
Chapter
4
for
a
summary
of
households
served
by
systems
subject
to
various
LT2ESWTR
provisions).
Of
all
the
households
subject
to
the
rule,
24
to
35
percent
are
projected
to
incur
costs
for
adding
treatment,
depending
on
the
Cryptosporidium
occurrence
data
set
used.
Approximately
95
percent
of
the
households
potentially
subject
to
the
rule
are
connected
to
systems
serving
at
least
10,000
people;
these
systems
experience
the
lowest
increases
in
costs
due
to
economies
of
scale.
Over
90
percent
of
all
households
will
face
an
annual
cost
increase
of
less
than
5
dollars.

6.11
Unquantified
Costs
EPA
has
quantified
all
the
major
costs
for
this
rule
and
has
provided
uncertainty
analyses
to
bound
the
over­
or
underestimates
in
the
costs.
Some
costs
are
unquantifiable,
however,
because
of
a
lack
of
information.
One
cost
not
quantified
is
the
effect
of
systems
needing
to
comply
with
several
rules
in
a
similar
time
period.
This
analysis
took
into
consideration
compliance
with
the
Stage
2
DBPR,
the
LT1ESWTR,
and
the
IESWTR.
It
did
not,
however
take
into
account
other
rules
that
are
or
will
be
promulgated
before
this
rule.
These
include
the
Arsenic
Rule,
the
Ground
Water
Rule,
the
Radon
Rule,
and
the
Filter
Backwash
Rule.
Although
most
of
these
will
not
affect
surface
water
sources,
they
may
limit
the
use
of
alternative
sources.
The
rules
affecting
ground
water
could
affect
the
GWUDI
systems
in
the
LT2ESWTR.
There
could
be
lower
costs
for
these
systems
if
technologies
installed
for
this
rule
also
achieved
reduction
in
other
contaminants.
There
could
be
unquantifiable
savings
in
monitoring
and
Economic
Analysis
for
the
LT2ESWTR
Proposal
6­
31
June
2003
implementation
costs
for
complying
with
several
rules
at
once.
There
are
also
unquantifiable
costs
savings
associated
with
some
of
the
treatment
and
management
strategies
listed
in
section
6.5.1.1
that
were
not
included
in
this
analysis,
but
that
may
be
less
expensive
than
the
treatment
technologies
which
were
evaluated.

Another
cost
not
quantified
is
that
of
systems
merging
to
comply
with
this
rule.
Although
mergers
could
make
compliance
easier
or
treatment
costs
lower
(
due
to
economies
of
scale)
for
many
small
systems,
it
is
difficult
to
tell
how
many
would
result
from
this
rule
and
how
many
would
be
due
to
other
factors.
Costs
would
also
be
difficult
to
quantify.
There
could
be
savings
because
of
economies
of
scale,
but
there
could
also
be
increases
because
of
additional
capital
costs
to
connect
the
systems.

Toolbox
options
that
were
not
quantified
included
intake
management,
performance
studies,
and
peer
review.
These
measures
may
prove
cheaper,
so
their
inclusion
would
result
in
lower
costs.
The
cost
savings
are
difficult
to
quantify,
however,
because
the
effectiveness
and
applicability
of
these
options
is
unknown.

6.12
Summary
of
Uncertainties
and
Sensitivity
Analyses
As
described
in
section
6.1,
there
is
uncertainty
in
these
cost
estimates
that
could
result
in
either
an
overestimate
or
underestimate
of
the
costs
as
presented
in
this
chapter.
Exhibit
6.15
below
presents
a
summary
of
these
issues,
references
the
section
or
appendix
where
the
issue
is
introduced,
and
estimates
the
effects
that
each
may
have
on
national
costs.
Economic
Analysis
for
the
LT2ESWTR
Proposal
6­
32
June
2003
Exhibit
6.15
Summary
of
Uncertainties
Affecting
LT2ESWTR
Cost
Estimates
Uncertainty
Section
with
Full
Discussion
of
Uncertainty
Effect
on
Cost
Estimates
Underestimate
Overestimate
Under
or
Over
Estimate
Occurrence
data
used
to
predict
plant
binning
4.5.3
Appendix
O
X
All
systems
are
charged
the
same
laboratory
fee
for
Cryptosporidium
monitoring
6.1.1
X
Single
flow
used
to
evaluate
unit
costs
within
each
of
9
size
categories
6.5.1
X
Potentially
lower
cost
treatment
or
toolbox
options
not
considered
6.5.1
X
Typical
water
quality
and
operating
parameters
used
to
estimate
unit
costs
6.5.1
X
Economies
of
scale
not
considered
for
combined
treatment
technologies
6.5.1
X
Inability
to
link
all
purchased
systems
4.3.2
X
Number
of
systems
achieving
credit
for
technologies
in
place
4.5.1,
Appendix
A
X
Assuming
no
uncovered
reservoirs
obtain
State
waiver
6.8,
Appendix
I
X
Note:
Source
water
quality
conditions
may
inhibit
the
age
of
a
technology
or
cause
decreased
or
increased
capital
and
O&
M
costs.

Of
the
uncertainties
in
Exhibit
6.15,
EPA
has
identified
two
areas
that
could
have
an
effect
on
the
LT2ESWTR
cost
estimates:

°
Differences
in
predicted
Cryptosporidium
distributions
among
the
ICR
and
ICRSS
data
sets
and
subsequent
predictions
of
filtered
plant
binning
°
Use
of
typical
source
water
bromide
levels
(
affects
the
ability
to
use
ozone)

Sensitivity
analyses
were
prepared
to
quantify
the
potential
effects
of
these
uncertainties
in
an
attempt
to
"
bound
the
range"
of
probable
costs
of
the
LT2ESWTR.
Appendix
G
presents
multiple
technology
selection
forecasts
reflecting
the
three
occurrence
data
sets
and
an
increased
source
water
bromide
concentration
for
all
regulatory
alternatives.
Economic
Analysis
for
the
LT2ESWTR
Proposal
6­
33
June
2003
6.12.1
Cryptosporidium
Occurrence
Data
Sets
The
modeled
results
of
Cryptosporidium
occurrence
distributions
have
a
large
influence
on
the
national
costs
of
this
rule
and
have
much
uncertainty
associated
with
them.
For
this
reason,
separate
analyses
for
all
three
data
sets
were
carried
through
all
cost
analyses.
Exhibit
6.10
showed
a
significant
range
of
costs
between
the
occurrence
data
sets.
In
addition
to
the
uncertainty
between
data
sets,
the
distributions
generated
within
each
also
have
uncertainties
that
can
be
bounded.

To
further
bound
potential
costs,
distributions
were
created
to
represent
the
range
of
costs
given
the
credible
outside
bounds
of
the
occurrence
distributions.
A
distribution
that
approximates
the
credible
lower
bound
(
5th
percentile)
of
the
ICRSSL
occurrence
data
set
is
used
as
the
lowest
occurrence
data
set
likely
to
occur.
A
distribution
that
approximates
the
credible
upper
bound
(
95th
percentile)
of
the
ICR
occurrence
data
set
is
used
as
the
highest
occurrence
data
set
likely
to
occur.
All
other
factors
are
kept
the
same
as
the
assumptions
used
in
the
analyses
in
sections
6.3
to
6.5.
Exhibit
6.16
compares
the
predicted
binning
and
subsequent
treatment
costs
for
filtered
plants
for
the
approximated
"
high"
and
"
low"
occurrence
data
sets.
The
cost
estimates
presented
in
Exhibit
6.16
are
the
5th
percentile
estimate
of
the
low
occurrence
estimate
and
the
95th
percentile
of
the
high
occurrence
estimate.
Assuming
the
three
occurrence
data
sets
well
represent
the
full
range
of
possible
distributions,
and
that
other
cost
components
well
represent
the
costs
of
the
rule,
then
this
bounding
analysis
indicates
that
there
is
a
90
percent
or
greater
probability
that
the
true
mean
cost
will
be
within
this
range.

Exhibit
6.16
Sensitivity
of
Treatment
Cost
to
Selection
of
Cryptosporidium
Occurrence
Distribution
Bins
2,
3,
and
4
Low
High
Percent
of
total
plants
in
bins
14.4%
38.5%

Total
Annual
Treatment
Cost
($
Millions,
2000$)

3
percent
discount
rate
$
27
$
106
7
percent
discount
rate
$
29
$
115
Sources:
Percent
of
Total
Plants
in
Bins
from
Appendix
B,
Exhibit
B.
3,
1.0
log
removal
bin,
2.0
log
removal
bin,
2.5
log
removal
bin.
Total
Annual
Cost,
from
Appendix
O.
4d
(
3%)
and
O.
4e
(
7%);
Row
­
All
System
Sizes,
Low;
sum
of
Columns
A,
B,
G,
and
H.

6.12.2
Sensitivity
Analysis
of
Influent
Bromide
Levels
on
Technology
Selection
for
Filtered
Plants
In
the
LT2ESWTR
least
cost­
modeling
approach,
ozone
is
selected
after
UV
disinfection
and
chlorine
dioxide,
but
before
MF/
UF
 
one
of
the
most
expensive
treatment
technologies
of
the
toolbox
options.
The
concentrations
of
bromide
in
the
plant
influent
can
limit
ozone
use.
Economic
Analysis
for
the
LT2ESWTR
Proposal
6­
34
June
2003
ICR
ICRSSL
ICRSSM
ICR
ICRSSL
ICRSSM
A
B
C
D
E
F
Number
of
Plants
Converting
to
Ozone
58
36
44
41
31
37
3
Percent
$
87
$
50
$
63
$
97
$
54
$
68
7
Percent
$
93
$
55
$
68
$
105
$
58
$
73
Standard
Condition
(
Influent
Bromide
from
ICR)
Increased
Influent
Bromide
Total
Annual
Treament
Cost
($
Millions,
2000$)
The
ICR
database
includes
plant
influent
bromide
concentrations
for
July
1997
through
December
1998.
Bromide
levels
vary
from
year
to
year
and
are
highest
during
drought
periods.
There
is
concern
that
ICR
bromide
data
were
not
collected
during
a
drought
period
and,
thus,
do
not
accurately
reflect
maximum
influent
levels
around
which
plants
would
design
their
ozonation
systems.
For
the
standard
conditions,
maximum
use
percentages
for
ozone
reflect
the
SWAT
analysis
using
influent
bromide
equivalent
to
the
values
reported
in
the
ICR
(
USEPA
2003b).
EPA
conducted
another
SWAT
analysis
where
the
influent
bromide
concentrations
for
each
plant
were
increased
by
50
parts
per
billion
(
ppb).
Those
results
provided
a
maximum
use
percent
of
ozone
for
the
LT2ESWTR
decision
tree,
under
high
influent
bromide
levels.
Exhibit
6.17
compares
the
number
of
plants
selecting
UV
and
ozone
and
the
filtered
plant
treatment
cost
estimates
for
each
technology
selection
(
standard
and
influent
bromide
of
50
ppb).
Technology
selection
constraints
on
ozone
use
have
little
impact
on
annual
costs.

Exhibit
6.17
Sensitivity
of
Technology
Selection
to
Influent
Bromide
Concentration
for
Filtered
Plants
[
1],[
2]

Notes:
[
1]
Evaluated
under
conditions
for
the
Preferred
Regulatory
Alternative.
[
2]
Nonpurchased
plants.
Sources:
"
Number
of
Plants
Converting
to
Ozone":
A
ppendix
G,
Exhibits
G.
37­
48;
Row
­
Total
Plants;
Columns
H­
J.
"
Treatment
Cost
(
Annual)"
[
A],[
B],
and
[
C]:
Appendix
O,
Exhibit
O.
4d
(
3%)
and
O.
4e
(
7%);
Row
­
A3,
ICR
[
A],
ICRSSL
[
B],
ICRSSM
[
C];
Columns
A,
B,
G,
and
H.
"
Treatment
Cost
(
Annual)"
[
D],[
E],
and
[
F]:
Appendix
O,
Exhibit
O.
4d
(
3%)
and
O.
4e
(
7%);
Row
­
A3
UV90­
10B,
ICR
[
D],
ICRSSL
[
E],
ICRSSM
[
F];
Columns
A,
B,
G,
and
H.

6.13
Comparison
of
Regulatory
Alternatives
Exhibits
6.18a­
b
provide
a
summary
of
the
annualized
present
value
of
filtered
plant
costs
for
each
regulatory
alternative,
for
each
data
set,
using
3
and
7
percent
discount
rates,
based
on
a
25­
year
period
of
analysis.
Appendix
V
provides
a
more
detailed
breakout
of
costs
for
each
regulatory
alternative.
These
costs
do
not
include
unfiltered
plants
because
their
regulatory
requirements
do
not
change
between
alternatives.
Economic
Analysis
for
the
LT2ESWTR
Proposal
6­
35
June
2003
Distribution
System
Type
Preferred
Alternative
Alternative
1
Alternative
2
Alternative
4
A
B
C
D
£
 
10K
13.0
$
40.6
$
21.9
$
8.2
$
>
10K
96.5
$
319.0
$
111.0
$
50.2
$
Total
109.5
$
359.6
$
132.9
$
58.5
$
£
 
10K
7.9
$
40.6
$
17.6
$
5.0
$
>
10K
64.6
$
319.0
$
81.1
$
30.8
$
Total
72.5
$
359.6
$
98.7
$
35.8
$
£
 
10K
9.5
$
40.6
$
19.0
$
6.1
$
>
10K
75.6
$
319.0
$
92.5
$
36.6
$
Total
85.1
$
359.6
$
111.5
$
42.7
$
ICR
ICRSSL
ICRSSM
Distribution
System
Type
Preferred
Alternative
Alternative
1
Alternative
2
Alternative
4
A
B
C
D
£
 
10K
13.0
$
38.9
$
22.2
$
8.4
$
>
10K
106.5
$
347.3
$
121.9
$
55.3
$
Total
119.5
$
386.2
$
144.1
$
63.8
$
£
 
10K
8.0
$
38.9
$
18.1
$
5.2
$
>
10K
71.7
$
347.3
$
89.1
$
34.2
$
Total
79.8
$
386.2
$
107.2
$
39.4
$
£
 
10K
9.6
$
38.9
$
19.5
$
6.3
$
>
10K
83.7
$
347.3
$
101.7
$
40.4
$
Total
93.3
$
386.2
$
121.2
$
46.7
$
ICR
ICRSSL
ICRSSM
Exhibit
6.18a
Comparison
by
Regulatory
Alternative
of
Total
Costs,
Annualized
at
3
Percent
for
Filtered
Plants
($
Millions,
2000$)

Exhibit
6.18b
Comparison
by
Regulatory
Alternative
of
Total
Costs,
Annualized
at
7
Percent
for
Filtered
Plants
($
Millions,
2000$)

Sources:
Appendix
O.
Exhibit
6.18a:
[
A]
Exhibit
O.
4d;
Rows
­
Mean,
A3,
ICR
and
ICRSSL
+
Exhibit
O.
3d;
Rows
­
A3,
ICR,
ICRSSL,
and
ICRSSM.
[
B]
Exhibit
O.
4d;
Rows
­
Mean,
A1,
ICR
and
ICRSSL
+
Exhibit
O.
3d;
Rows
­
A1,
ICR,
ICRSSL,
and
ICRSSM.
[
C]
Exhibit
O.
4d;
Rows
­
Mean,
A2,
ICR
and
ICRSSL
+
Exhibit
O.
3d;
Rows
­
A2,
ICR,
ICRSSL,
and
ICRSSM.
[
D]
Exhibit
O.
4d;
Rows
­
Mean,
A4,
ICR
and
ICRSSL
+
Exhibit
O.
3d;
Rows
­
A4,
ICR,
ICRSSL,
and
ICRSSM.
Exhibit
6.18b:
[
A]
Exhibit
O.
4e;
Rows
­
Mean,
A3,
ICR
and
ICRSSL
+
Exhibit
O.
3e;
Rows
­
A3,
ICR,
ICRSSL,
and
ICRSSM.
[
B]
Exhibit
O.
4e;
Rows
­
Mean,
A1,
ICR
and
ICRSSL
+
Exhibit
O.
3e;
Rows
­
A1,
ICR,
ICRSSL,
and
ICRSSM.
[
C]
Exhibit
O.
4e;
Rows
­
Mean,
A2,
ICR
and
ICRSSL
+
Exhibit
O.
3e;
Rows
­
A2,
ICR,
ICRSSL,
and
ICRSSM.
[
D]
Exhibit
O.
4e;
Rows
­
Mean,
A4,
ICR
and
ICRSSL
+
Exhibit
O.
3e;
Rows
­
A4,
ICR,
ICRSSL,
and
ICRSSM.
