                                  MEMORANDUM

To:
Bryan Hubbell  -  Leader, Risk and Benefits Group OAQPS
From:
Stephen Graham  -  Physical Scientist, Risk and Benefits Group OAQPS
Date:
July 15, 2012
Re:
Comprehensive review of published averting behavior studies and available technical documents.



Overview
The following review of the averting behavior literature draws upon key results from several published studies (including the ISA evaluation of these studies [ISA, section 4.1.1]), though also extending the literature reviewed to include several available technical reports.  The discussion that follows is separated into two broadly identified components of averting behavior: (1) the prevalence of air quality system awareness and (2) the common behavior modifications performed in response to either the air quality alert information or high ambient pollutant concentrations.  We attempt to distinguish the various characteristics influencing each of these two components, though recognize that neither exists solely in isolation.  In addition, the implications of averting behavior relevant to our ozone exposure and risk assessment are then briefly summarized, also separated into two categories: (1) the impact to epidemiologic concentration-response functions, (2) the effect of using time location activity pattern data for persons that may be averting when simulating population ozone exposures.    At the end of this document is a table that summarizes the relevant findings from each publication reviewed.
   1. Air Pollution Alert Awareness
There are three components of a fully functioning local-scale air quality system that can influence a personal behavioral response to an air quality alert: (1) awareness of local ambient air quality conditions, (2) understanding the relationship between air pollutant exposure and adverse health effects, and (3) knowledge of local air pollution management programs designed to reduce pollutant concentrations and/or personal exposure.  The magnitude of impact the air quality system information can have will be directly influenced by the type and amount of communication offered by both National and local health and environmental protection agencies.  A number of the studies reviewed here collected relevant information on these three factors, commonly in the form of recall or panel surveys designed to gauge the particular degree of awareness in survey participants.  As the types of survey questions and study participants used were wide ranging and dependent on the particular study objectives, the responses and participation rates were also highly variable and dependent on a variety of factors.  The objective in this current evaluation is to generalize the findings regarding people's knowledge of air quality systems and to recognize the important variables influencing awareness.
In considering the two commuter panel studies reviewed, awareness of AQ systems in urban areas can be quite prevalent (>90%; Blanken et al, 2001) with general air pollution awareness raised by factors such as time (day-of-week, month-of-year), residential proximity to urban core, and personal demographics (i.e., those individuals of higher income-level, obtaining higher education, or are self-described as a non-minority) (Blanken et al., 2001; Henry and Gordon, 2002).  In several other larger-scale population-based random surveys, I found a large portion of study participants were also aware of air quality alert information (e.g., 42.7%, KS DOH, 2006; 33%, Mansfield and Corey, 2003; 61% Mansfield et. al, 2006; 91% Mansfield et al., 2009; 82-89%; Semenza et al., 2008) and those with greater awareness had similar characteristics as described above for commuters (e.g., higher income level, older age, reside in an urban location).  
Perhaps more importantly though, awareness was also found consistently heightened in persons having a compromised health condition relative to those without (e.g., 46% of asthmatics vs. 42.5% of non-asthmatic, KS DOH, 2006; 95% of asthmatic children versus 80% of non-asthmatic children, McDermott et al., 2006).  The likelihood of checking for air quality alerts is also influenced by health status, as 25% of parents with asthmatic children check the forecast daily, compared with only 9% of parents of non-asthmatic children (Mansfield et al., 2009).  Further, Houtven et al. (2003) also reports increased awareness among asthmatic children compared to non-asthmatics, noting also the degree of disease severity as a statistically significant explanatory variable.  
There were a few studies that collected information on how persons felt about the relationship between air pollution and health effects, and when considered, the resulting fraction of persons making the connection between the two factors was generally small.  For example, it was estimated that only 6% of the Kansas population thought that an illness or symptom they had was caused by an outdoor air pollution event (KS DOH, 2006).  Though still, the ability to make a connection between air pollution and health effects remains a significant factor increasing overall awareness of air quality alerts (Henry and Gordon, 2002).  It is possible that the small percent of persons able to make this connection is the result of their having limited communication with their health care providers about this issue.  For example, it was estimated that only 4% of Kansans surveyed were advised by a health professional to reduce outdoor activity level when air quality is poor (KS DOH, 2006).  Further, it is possible that only a few surveyed persons actually experienced a high enough exposure to elicit a health response, allowing one to make a firm connection between air pollution and health effects, though none of the studies I reviewed evaluated this factor.
That said, it would be misleading to suggest that all studies indicated persons were unable to make connection between exposure and health risk.  As an example of an upper bound estimate, 91% of parents from a convenience survey of parents in Salt Lake City, Utah linked poor air quality with asthma and other illness, with little difference observed between asthmatic and non-asthmatic cohorts (McDermott, 2006).  It is possible that the observed variability in awareness indicates that personal awareness and perception regarding air pollution and health is influenced by geography, beyond simple urban versus rural classifications.  For example, six states (Colorado, Florida, Indiana, Kansas, Massachusetts, and Wisconsin) reported AQI awareness and outdoor activity level information for the 2005 Behavioral Risk Factor Surveillance System (BRFSS) (Wen et al., 2009).  In total, 49% of asthmatics and 46% of non-asthmatics were aware of media alerts on air quality.  Data for one of the states, Kansas, reported a lower percent for each cohort, i.e., 46% and 42.5%, respectively (KS DOH, 2006).  Similarly, small but statistically significant differences in alert system awareness were observed in surveyed populations in Portland (82.3%) and Houston (89.1%) (Semenza et al., 2008).
To summarize, the prevalence of awareness was variable.  I estimate about 50% to 90% of survey study participants acknowledge or were familiar with air quality systems, with asthmatics or parents of asthmatic children having a consistently greater degree of awareness (approximately a few to 15 percentage points) when compared to that of non-asthmatics.  Advice from a health care provider also increased awareness, and while the portion of the population receiving this communication is small, it most likely correlated with persons having a health condition.  Some of the personal variables found associated with increasing awareness, generally in order of greatest importance, include health status (particularly asthmatics), urban proximity, and a suite of commonly correlated demographic attributes (i.e., education, income level, non-minority, and age). 
I suggest that one influential factor not adequately explored is city- or region-specific differences in awareness.  Most studies took place in a single area and few if any that did take place in more than one study area had the proper data to evaluate such differences while controlling for all potential influential variables.  Another important factor not analyzed in the reviewed literature that may remain un-testable for some time, is changes in awareness within a location over time.  It is highly likely that public awareness of air quality systems is increasing across the US.  For example, there are now communications via the internet and television (both government and news media sources), availability of mobile phone applications, drives for participation in AQI school flag programs, among several others sources, of which, most have only been developed within the past decade or few recent years.  This increased communication and heightened public awareness is likely to yield greater a rate of averting behavior in response to air quality alerts, supporting further the importance of our investigations into understanding the impact averting may have on exposure and risk assessments.  
   2. Averting Behaviors
There can be a wide range of behavior modifications performed in response to environmental pollutants.  For example, one of the earliest studies I reviewed took place in an area having historical groundwater contamination and analyzed factors that influenced local residents' choice of drinking water source and their attendance at public meetings (Smith and Desvousges, 1986).  While perhaps informative in stimulating generalized questions regarding how and the degree of which persons might react in response to environmental pollutants, for our purposes, I focused my review on those studies primarily designed to evaluate air pollution averting behaviors.  I characterized the averting behaviors of interest as either activities performed that would reduce pollutant emissions or actions taken to reduce personal exposures to ambient pollutant concentrations.
   0.1 Averting behavior to reduce pollutant emissions
As part of an air quality management plan to reduce pollutant emissions, local environmental and health agencies may strongly encourage daily commuters to use alternative modes of transportation rather than use single occupied passenger vehicles to get to their workplace.  Even though many commuters are keenly aware of air quality alerts and requests for pollutant emission reductions, many of the early studies showed that adult commuting patterns are not dramatically affected by air quality alerts (Blanken et al., 2001; Henry and Gordon, 2002).  For example, while the total commuter miles driven may be reduced on alert days (e.g., by about 15%, Henry and Gordon, 2002), rates for this averting behavior tend to range from low (e.g., 12% of commuters said they changed pattern on alert day, however 76% do not, Blanken et al., 2001; 3-3.5% reduction in traffic volume on alert days, Cutter and Neidell, 2009) to at most, a moderate extent (e.g., 28% of adults said they adjusted their driving pattern on alert days, Mansfield and Corey, 2003).  Further, there was one instance where commuter miles were not reduced but increased in response to smog alerts (Noonan, 2010, 2011).  One recent study conducted in Salt Lake City, Utah reported a large percent of parents (71%) say they reduce fossil fuel use when air quality is poor (McDermott et al., 2006), possibly indicating that air quality awareness may be increasing over time and potentially influencing the rate of averting as suggested above in the discussion regarding awareness.  However, a study conducted by Semenza et al. (2008) may suggest otherwise for Portland, OR and Houston, TX where participants indicated they neither reduced driving nor postponed refueling events on air pollution advisory days.  As was discussed regarding awareness, perhaps the study location (or particular geography) is the prevailing influential variable, though, in the absence of multi-city data from the McDermott et al. (2006) study, this is purely speculation.
There were a few factors identified in the reviewed studies that appear to influence whether or not someone participates in an averting behavior to reduce local pollution emissions from motor vehicles.  First, the perceived inconvenience of mass transportation and limited employer support may be a significant driver behind the low response rate observed.  For example, public employees exhibited a greater reduction in both number of trips and miles travelled when compared with their private counterparts, suggesting the likelihood of a public policy or supportive position regarding commuting to these workplaces (Henry and Gordon, 2002).  In addition, those persons more educated along with having higher income tend to drive more than their counterparts, implying a strong regularity in their commuting pattern and an inherent non-conformance to air quality alerts (Henry and Gordon, 2002; Mansfield and Corey, 2003).  However, the perceived efficacy of the behavior may also contribute to low rate of averting activities as very few commuters feel that reducing their mobile source emissions has an impact on local air quality (Blanken et al., 2001).  As was described above for air quality alert awareness, parents of asthmatic children were more likely to say they reduced pollutant emissions on poor air quality days compared with parents of non-asthmatic children (79% vs. 61%, respectively; McDermott et al., 2006), again highlighting the importance of health condition regarding averting decisions made.  And finally, time-of-day had a significant influence on observed traffic volume reductions, particularly during the morning rush hour compared with other times in the day (Cutter and Neidell, 2009).
One other averting activity noted to reduce pollutant emissions and its associated ambient air concentrations is the placing of a restriction on wood burning.  It is entirely possible that having mandatory restrictions could positively bias estimated participation, however the strong response rate (i.e., 86%, Blanken et al., 2001) suggests a strong perceived link between the emissions and the local ambient air pollution.  Another activity, adjusting the time of vehicle refueling events, was only briefly mentioned in one study and found to have negligible participation (Semenza et al., 2008).
To summarize, there was limited participation in activities that might reduce pollutant emissions, therefore there are limits to the number of variables identified as significantly influential to these types of averting behaviors.  The ability to participate in and/or having active support for such voluntary emission reduction programs (i.e., car-pooling, ride-share), as well as being vested in the ultimate response (e.g., you or your child has a health condition known to be affected by pollutant emissions) appear to be the most influential variables.


   0.2 Averting behavior to reduce exposure
Many of the reviewed studies were designed to investigate averting behavior in response to an air quality alert.  Commonly, the air pollutant responsible for most alert notifications is ozone hence, a variable of interest used to evaluate averting behavior is how a person's outdoor activities are changed on alert days compared to non-alert days.  The exertion level of the person (e.g., performing less vigorous activities outdoors) as well as modified time expenditure in microenvironments (e.g., increasing time indoors or less time outdoors) are typical averting behaviors that may potentially reduce exposure concentrations to ambient air pollutants, particularly ozone, because ozone does not persist in indoor environments.  These two features, exertion level and time expenditure, are discussed below.   For the most part, we attempt to distinguish between the two features, although at times for convenience and flow, the overall discussion theme may be driven largely by the influential variable (e.g., day-of-week).  We begin the discussion with an evaluation of the literature that reported changes in outdoor activity level.
Studies have identified three important variables that influence outdoor activity level, particularly considering those persons most susceptible to the effects of air pollution.  These variables are 1) alert awareness, 2) understanding the link between air quality and health responses, and 3) personal health status.  As discussed above regarding awareness, those individuals potentially at greatest risk are not only more aware of their local air quality system but also more likely to have been informed of health risks by a health professional (KS DOH, 2006).  Asthmatic adults, either aware of the alert system or able to link poor air quality with adverse health responses, are estimated as having a two-fold likelihood of reducing their outdoor exertion level on alert days than those adults without asthma (KS DOH, 2006; Wen et al., 2009).  Those advised by a health professional had an even greater likelihood of adjusting outdoor activity level, regardless of whether they were asthmatic or not (KS DOH, 2006; Wen et al., 2009).  Noonan (2010, 2011) also reports potential reduction in outdoor event participation on smog alert days, particularly notable for those observed exercising (i.e., a 25% reduction).  While awareness was not directly measured, participation in outdoor activities on smog alert days was also reduced for three other study groups: a cohort comprised of all study participants, the second representing children, and the third constituting elderly persons, though only the elderly were found to be statistically significant (Noonan, 2010, 2011). 
While this link between professional advice and attention paid to the air quality alerts is strong, there may be a number of reasons for the limited rate of participation in averting activities.  It may be that too few persons are getting this health communication (as discussed above regarding awareness), perhaps persons are overstating what they might do in response to air pollution alerts in response to the survey questionnaire, or it is possible that they are simply inconvenienced by a rigid daily (and weekly) schedule.  For example, a recent study indicated that 55% of parents stated they restrict children's outdoor play on bad air quality days (McDermott et al., 2006).  In an area that had, on average, 31 pollution advisories per year, most parents (74%) indicated that they would only restrict their child's outdoor activity for fewer than 6 days per year (McDermott et al., 2006).  While asthmatic children were more likely than non-asthmatics to restrict their outdoor activities 3 to 5 days per year (34% vs. 17%, respectively), there was no difference between the two cohorts regarding a greater degree of compliance with air quality alerts.  Simply put, there appears to be a limit to how many days per ozone season one might adhere to air quality alerts and associated recommendations, even for persons most at risk of adverse health effects.  
Zivin and Neidell (2009) evaluated such an effect, i.e., changes in the averting response across three consecutive days, using outdoor facility attendance records in Los Angeles.  Significant reductions in attendance were observed on the first alert day (i.e., about 15% lower at the Los Angeles Zoo, 8% at the Griffith Observatory) however, on the two consecutive days that followed with smog alerts, there were no statistically significant reductions in attendance at either facility.  These analyses suggest that even the notably significant, albeit overall limited degree of participation in averting eventually wanes in response to multiple (and consecutive) alerts.  Interestingly though, this apparent lack of attention paid to the alert appears short-lived and averting "rebounds" when evaluating a slightly different 3-day consecutive period: the first an alert day, followed by no alert, and ending the period with an alert day.  Zivin and Neidell (2009) reported a reduction in attendance on the 3[rd] day (an alert day) similar to that observed on the 1[st] day (also an alert day).  Similarly, Sexton (2011) evaluated time spent engaged in vigorous outdoor activities and the effect of three consecutive days air quality alerts, though compensation in performing outdoor activities appeared to be delayed to the third day (that is, averting was noted on days 1 and 2).  There were no studies indicating longer term effects of time and alert status on averting behavior related to outdoor time, though Henry and Gordon (2002) do find awareness significantly increasing over summer months, peaking in August then decreasing in September, suggesting a similar pattern may exist for interest in air quality alerts and active participation in averting behavior.
One other potentially influential variable not thoroughly evaluated was the potential for regional differences in averting activities.  While there were no significant differences in observed averting behavior rates, there were significant differences in evaluating whether air quality was poor or not in Houston, TX when compared with the rate in Portland, OR (Semenza et al., 2008).  As stated above, the perception of how poor air quality is linked with poor health is an important driver for reducing outdoor activity level (KS DOH, 2006; Wen et. al, 2009).  Geographic differences may explain why slightly fewer Kansans (21.8% of asthmatics, 7.7% of non-asthmatics) changed their outdoor activity level compared with that estimated for six states (25.6% of asthmatics, 12.0% of non-asthmatics).   A smaller though not similarly patterned difference was observed between Kansans (32.6% of asthmatics, 13.7% of non-asthmatics) and in six States (31.1% of asthmatics, 16.1% of non-asthmatics) considering those who adjusted their outdoor activity base on air quality alerts.  Mansfield et al. (2009) and Houtven et al. (2003) report variable influence of location on outdoor time, though mainly not statistically significant, in three broadly defined regions (Northeast, Southeast, and West Coast).
Three key features affecting outdoor time reduction appear to be the magnitude of the alert or ozone concentrations, the particular time-of-day, along with health condition severity.   As far as changes in time expenditure, Mansfield et al. (2006) estimates that asthmatic children reduce their time spent outdoors by 30 minutes on code-red alert days relative to green, yellow, or orange days.  In a related unpublished paper, Mansfield et al. (2009) observed that in air pollution sensitive asthmatics, the reduction in time spent outdoors can occur during the afternoon hours when ozone concentrations are highest.  Neidell (2010) also found that averting during the daytime hours can be greatest when the highest alert level is issued, rather than when lower alert thresholds are exceeded or when an outdoor event can take place during evening hours.   Sexton (2011) showed that reduction in time spent engaged in vigorous outdoor activities for the general population was on average about 21 minutes, occurring more so during the afternoon when compared to morning hours of an alert day (and 18 minute versus 10 minute reduction, respectively).  One significantly affected population subgroup from the same study, elderly persons, showed greatest reductions in outdoor time (approximately 65 minutes) in response to AQI values ranging from 175-199.  Outdoor time expenditure appears to be influenced by high ozone concentrations, as Bresnahan et al., (1997) estimates people may spend about 40 minutes less per day when 1-hour daily maximum ozone concentration exceeds 120 ppb.  A similar ozone effect was not shown in analyses performed by Yen et al. (2004) and Eiswerth et al., (2005) using either the daily maximum 1-hour ozone concentrations or ozone as a binary variable (exceed standard or not) though it was reported the severity of study participants asthma significantly influenced overall time expenditure (either indoors and outdoors) and their associated activity levels.
Furthermore, as described above for commuting behavior, the limited rate of averting by adjusting time spent outdoors for much of the population may also be influenced by personal circumstance or convenience, particularly noteworthy for employed adults.   For example, Neidell (2009) found a significantly greater reduction in attendance at an outdoor facility on pollution alert days for children (22-25%) and the elderly (19%) when compared to the total attending population (13-15%).  Sexton (2011) evaluated participation in vigorous outdoor activities among persons aged 15 and greater, showing the majority of outdoor time and activity reduction contributed by elderly persons.  Children and elderly population groups are less likely to have rigid daily summer schedules and thus are more likely to avoid high pollution days when compared to working adults.
To summarize, averting to reduce exposure, whether through limiting outdoor time or exertion level, is evident, albeit not overly prevalent, among the study participants.  Influential factors are similar to those identified above for awareness, that is, persons having a respiratory health condition and its severity (e.g., asthma) and whether the subject is aware of alert notifications and are able to connect poor air quality to a potential adverse health response.  I estimate that perhaps 30% of asthmatics reduce their outdoor activity level on alert days, with the percent for the general population about half that (i.e., 15%).  Children and the elderly are more likely to avert, while non-elderly adults appear to demonstrate very limited to no averting.  Alert level also appears to be a factor, the greater the severity of the notification, the greater the likelihood of an averting response.  When an averting activity is performed, I estimate that outdoor time/exertion, particularly during afternoon hours may be reduced by about 20-40 minutes in response to an air quality alert notification.  There may be limits to the number of averting events one might engage in over short periods of time, and potentially longer term patterns influenced by the month of the year and the number and accuracy of alerts issued.  
   1. Impact of Averting Behavior
Based on the above discussion, the increased awareness of air quality alerts and corresponding averting behavior is beneficial for preventing adverse health outcomes in persons most susceptible to air pollutants, that is, individuals having asthma and/or other comprised respiratory condition or at a select life-stage.  From an exposure and risk modeling perspective, the more prevalent public awareness becomes and the likelihood of engaging in averting activities increases, the greater the challenge in accurately estimating of epidemiologic-based concentration response functions and population exposure distributions if ignoring the influence of these factors.  The general impact of each of these is described below.
   0.1 Epidemiologic-related
Individual level exposure to ambient air pollutants is neither modeled nor measured in most epidemiologic studies, though inhalation exposure has increasingly become an important consideration in such studies, particularly when short-term health effects are of interest and when exposure concentrations are expected to have greater spatial and temporal variability than ambient monitor concentrations.  When ambient monitor concentration variability alone does not fully explain observed heterogeneity in health effect responses in epidemiologic studies (e.g., Bell et al., 2004), factors not accounted for such as variability in time expenditure or outdoor exertion level (examples of so-called exposure or measurement errors) are identified as potential explanatory variables (ISA section 4.1.2).  Simply put, the strong correlation between personal ozone exposure and ambient concentrations (0.3-0.8, ISA section 4.3.3) will be influenced by averting behavior, such that as averting behavior increases on high ozone concentration days, the estimated correlation between the two will fall.  It follows that, when ignoring averting behavior which reduces ozone exposure (and is predicted to increase over time with improvements to air quality systems), existing and newly estimated ambient concentration-health response functions used to estimate risk could be biased low (Bateson et al., 2007), perhaps even approaching a null relationship (ISA section 4.12). 
The ISA exemplifies a recent epidemiologic study supporting this effect (Neidell and Kinney, 2010; Neidell, 2009), whereas the numbers of asthma-related hospitalizations for children and elderly persons were estimated to be reduced by a factor of 2.6 and 1.4, respectively when not accounting for averting behavior.  Ultimately, this influential factor is one of the identified uncertainties in our health risk estimates.  This potential underestimation bias in estimated concentration-response functions would vary based on the expected magnitude of averting that may occur by the types of persons within a particular epidemiologic study.  Over time however, averting behavior will become increasingly important to account for when estimating these epidemiologic concentration health response functions as more persons may become aware and thus actively participate in averting to reduce their exposure.
   0.2 Exposure-Related
Critical to any inhalation exposure calculation are the elements of duration, pattern, frequency, and magnitude of exposure, each of which are driven by the location of pollutant contact and the activities persons perform.  When one ignores the spatial and temporal patterns in pollutant concentrations and human-time-location activity patterns, upper estimates of exposure (commonly the persons of greatest interest in an exposure or risk assessment) will always be biased low.  That is why when modeling human population inhalation exposure, the best approach uses real time-location-activity diaries to more accurately simulate contact with pollutants in their immediate breathing environment.
As described in the ozone REA (section 5.3.2), the diaries used in our APEX modeling were obtained from several different studies, data gradually collected over the past few decades and structured specifically for use in our time-series exposure modeling approach.  The original studies themselves had their own set of objectives, some of which were exposure-related, while others were not, though most of the studies were not designed to estimate or directly capture averting behavior per se.  This is not to suggest that there are not diaries where persons may have exhibited a particular averting behavior.  All of the earlier studies in the Consolidated Human Activity Database (CHAD; many are from 1980's-2000) could contain diary data from individuals that may have performed averting to some extent, either influenced by elevated ozone concentrations occurring at the time the diary record was created or perhaps resulting from the individual's knowledge of the AQI or other local alert system (albeit a likely limited fraction of persons).  It is possible that there are a greater number of individuals that engaged in averting behaviors considering the most recent study data that are now incorporated into CHAD compared with some of the historical CHAD data.  The NSAS study is one such recent study that, in addition to perhaps capturing a relatively greater number of persons aware of air quality systems and alert notifications, was also designed to potentially capture averting on alert days (i.e., there was an over-sampling of data from ozone alert days).  However, even considering this somewhat non-random study design, the highest alert level (i.e., code red) occurred only on 8.6% of diary days.  Therefore, given all the CHAD diary days available (>40,000 person-days) and considering the rate of averting behavior, it is likely that only a small percentage of the diaries would represent persons exhibiting averting behavior.  It is also important to add that at this time, none of the diary days used by APEX have been identified as representing days where a person did or did not perform an averting behavior to reduce their exposure.
When using all of these diaries to represent what people might do and where they might go within our exposure modeling domain, and have them applied somewhat randomly regarding averting behavior, the anticipated effect on exposure estimates is somewhat complicated and dependent on how and how often the particular diaries are used in constructing daily activity pattern profiles.  To further elaborate, in modeling multi-day exposures we use an approach that links CHAD diaries together based on a user-selected, important exposure variable, e.g., time spent outdoors.  The longitudinal diary profile generated for each individual is neither perfectly correlated among all days nor completely random, though each simulated person will have a diary profile that maintains some degree of correlation across a multi-day period (i.e., simulated persons who spend more time outdoors than others do so consistently over the duration of the model run).  Therefore, even though the diaries that may have expressed averting behavior due to ozone alerts are not directly identified, it is possible that a simulated person who typically spends significantly more time outdoors than most other persons, will have a diary day (or more, depending on the duration of the model run) where time spent outdoors is limited.
The daily diary chosen by the model is not directly matched with ozone concentrations using this longitudinal approach however, we do use temperature as a selection variable to best match the weather conditions of the day, an important variable influencing time spent outdoors.  As the weather conditions are generally linked to ozone concentrations, the diary days will, some of the time, be correlated with ozone concentrations.  I state "some of the time" because neither of the two relationships, outdoor time-to-temperature and temperature-to-ozone, are entirely linear.  For example, it is possible that some simulated individuals that normally spend a lot of time outdoors will appear to `avert' because a diary was selected that had reduced outdoor time in response to a high temperature (rather than high ozone, if occurring on that day).  
Regardless, I could speculate any number of scenarios that could occur in our exposure model simulations, some simulated persons will appear to exhibit averting to high ozone on some days, while others will not.  Averting may be taken into account in our exposure modeling, albeit to an unknown degree, though definitely generating exposure estimates that would be biased low if one were interested in estimating exposures that would occur in the complete absence of averting behavior.  
The principal issue is that neither the simulated nor correct number/fraction of persons averting is a known quantity for any study area.  We could attempt to better characterize the affect of this influential variable on exposure estimates by perhaps, constructing a shorter-term exposure scenario (e.g., 2 weeks) using an unbalanced (but proportionally known) selection of diaries designated as "representative" of averting behavior and compare exposure results with the standard diary selection approach.  However, we expect that, given the duration of the standard exposure simulations (i.e., an entire ozone season) , understanding variability in human time-location-activity patterns, and the potential for decreasing likelihood in multi-day averting for "real" people, it is likely that even if the proper proportion of the population (including asthmatics) exhibiting averting behaviors was known, the most significant impact would be to estimates of multi-day exposures above high exposure levels occurring in a year.  Based my review, in areas having several or more days at or above the highest alert levels, I estimate that accounting for averting would likely have little to no effect on our estimated number of persons with at least one exposure at or above a selected upper exposure level and likely little effect on the number of persons with a few (e.g., two to three) multi-day exposures in a year at or above the same upper exposure level.  Further, in areas having the only a few ozone alert days in a year, accounting for averting could potentially affect the number of estimated persons at or above select exposure levels for both single- and multi-day exposure in a year. 
   1. References
Bateson TF, Coull BA, Hubbell B, Ito K, Jerrett M, Lumley T, Thomas D, Vedal S, Ross M.  (2007).  Panel discussion review: session three  --  issues involved in interpretation of epidemiologic analyses  --  statistical modeling.  J Expos Sci Environ Epidemiol.  17:S90 - S96.
Bell ML, McDermott A, Zeger SL, Samet JM, Dominici F.  (2004).  Ozone and short-term mortality in 95 us urban communities, 1987-2000.  JAMA.  292(19):2372-2378.
Blanken PD, Dillon J, Wismann G.  (2001).  The impact of an air quality advisory program on voluntary mobile source air pollution reduction.  Atmos Environ.  35:2417-2421.
Blomquist GC.  (2004).  Self-protection and averting behavior, values of statistical lives, and benefit cost analysis of environmental policy.  Review of Economics of the Household.  2:89-110.
Bresnahan BW, Dickie M, Gerking S.  (1997).  Averting behavior and air pollution.  Land Econ.  73:340-357.
Bresnahan BW and Dickie M.  (1995).  Averting behavior and policy evaluation.  J Environ Econ Manage.  29:378-392.
Cheng W-L, Chen Y-S.  (2007).  Comparison of the revised air quality index with the PSI and AQI indices.  Sci Total Environ.  382:191-198.
Courant PN and Porter RC.  (1981). Averting expenditure and the cost of pollution.  J Environ Econ Manage.  8:321-329.
Cutter WB, Neidell M.  (2009).  Voluntary information programs and environmental regulation: evidence from `Spare the Air'.  Journal of Environmental Economics and Management.  58:253-265.
Eiswerth ME, Shaw WD, Yen ST.  (2005).  Impacts of ozone on the activities of asthmatics: Revisiting the data.  Journal of Environmental Management.  77(1):56 - 63.
Henry GT and Gordon CS.  (2003).  Driving less for better air: impacts of a public information campaign.  Journal of Policy Analysis and Management.  22(1):45-63
Houtven GV, Johnson FR, Mansfiled C, Yang J-C, Pyles A.  (2003).  Parental Averting Behavior With Respect to Ozone Alerts.  For presentation at AEA/ASSA Meeting, San Diego CA, January 2004.
KS DOH.  (2006).  Environmental Factors, Outdoor Air Quality, and Activity Level.  Results from 2005 Kansas Behavioral Risk Factor Surveillance System.  Office of Health Promotion, Kansas Department of Health and Environment.  Available at: http://www.kdheks.gov/brfss/PDF/cste_report_final.pdf.
Mansfield CA and Corey C.  (2003).  Task 4: Analysis of Survey Data on Ozone Alert Days.  Research Triangle Institute (RTI) International.  Project Number 7647.004.399.
Mansfield C, Johnson FR, Van Houtven G.  (2006).  The missing piece: averting behavior for children's ozone exposures.  Resource Energy Econ.  28:215-228.
Mansfield C, Houtven GV, Johnson FR, Yang J-C.  (2009).  Environmental Risks and Behavior: Do children spend less time outdoors when ozone pollution is high?  ASSA annual meeting, January 5, 2009.  Update of Houtven et al. (2003), using the OAB CHAD data set, related to Mansfield et al. (2006).
McDermott M, Srivastava R, Croskell S.  (2006).  Awareness of and compliance with air pollution advisories: A comparison of parents of asthmatics with other parents.  J Asthma.  43:235-239.
Neidell MJ.  (2004).  Air pollution, health, and socio-economic status: the effect of outdoor air quality on childhood asthma.  Journal of Health Economics.  23:1209-1236.
Neidell M.  (2005).   Incorporating Behavior into Biological Models: Understanding the Health Effects of Ozone.  Chicago: University of Chicago (Tech. Rept. #25). 
Neidell M.  (2006).  Public Information and Avoidance Behavior: Do People Respond to Smog Alerts?  Available at: http://www.uh.edu/academics/sos/econ/documents/abid_neidell.pdf.
Neidell M.  (2009).  Information, avoidance behavior and health.  J Human Res.  44:450-478.
Neidell M.  (2010). Air quality warnings and outdoor activities: evidence from Southern California using a regression discontinuity approach design.  J Epidemiol Community Health.  64:921-926.
Neidell M, Kinney PL. (2010).  Estimates of the association between ozone and asthma hospitalizations that account for behavioral responses to air quality information.  Environ. Sci. Policy.  13:97-103.  
Noonan D.  (2010).  Abatement or averting: the effects of ozone alerts on driving and outdoor recreation behavior in Atlanta.  Environmental Policy Instruments: Voluntary Mechanisms I.  Fourth World Congress of Environmental and Resource Economists, Montreal, Canada.  Available at: http://www.webmeets.com/WCERE/2010/prog/default.asp?pid=1236.
Noonan D.  (2011).  Smoggy with a chance of altruism: Using air quality forecasts to drive behavioral change.  AEI Working Paper #2011-08.  Available at: http://www.aei.org/files/2011/12/14/-smoggy-with-a-chance-of-altruism-using-air-quality-forecasts-to-drive-behavioral-change_140952566926.pdf.
Semenza JC, Wilson DJ, Parra J, Bontempo BD, Hart M, Sailor DJ, George LA. (2008).  Public perception and behavior change in relationship to hot weather and air pollution.  Environ Res.  107:401-411.
Sexton AL.  (2011).  Responses to Air Quality Alerts: Do Americans Spend Less Time Outdoors?  Available at: http://www.apec.umn.edu/prod/groups/cfans/@pub/@cfans/@apec/documents/asset/cfans_asset_365645.pdf.
Shofer S, Chen T-M, Gokhale J, Kuschner WG.  (2007).  Outdoor air pollution: counseling and exposure risk reduction.  Am J Med Sci.  333(4):257-260.
Smith VK and Desvousges WH.  (1986).  Averting behavior: does it exist?  Economics Letters.  20:291-296.  
Wen X-J, Balluz L, Mokdad A.  (2009).  Association between media alerts of Air Quality Index and change of outdoor activity among adult asthma in six states, BRFSS, 2005.  J Comm Health.  34:40-46.
Whitehead JC.  (2005).  Environmental risk and averting behavior: predictive validity of jointly estimated revealed and stated behavior data.  Environmental and Resource Economics.  32:301-316.
Wilson C.  (2005).  Exposure to pesticides, ill-health and averting behaviour: costs and determining the relationships.  International Journal of Social Economics.  32(12):1024-1034.
Yen ST, Shaw WD, Eiswerth ME.  (2004).  Asthma patients' activities and air pollution: a semiparametric censored regression analysis.  Review of Economics of the Household. 2:73 - 88.
Zivin JG, Neidell M.  (2009).  Days of haze: Environmental information disclosure and intertemporal avoidance behavior.  J Environment Econ Manag.  58:119-128.
Zivin JG, Neidell MJ.  (2011).  The Impact of Pollution on Worker Productivity.  National Bureau of Economic Research (NBER) Working Paper 17004. Available at: http://www.nber.org/papers/w17004.
Zujić AM, Radak BB, Filipović AJ, Marković DA.  (2007).  Extending the use of air quality indices to reflect effective population exposure.  Environ Monit Assess.  156(1-4):539-549.









Table 1.  Raw notes extracted from all publications collected and reviewed that may have data or information relevant to the discussion of ozone averting behavior. 
1[st] Author
Year pub.
Source
Cohort
Relevant
AQ Alert Awareness or Perception
Averting Behavior
Influential Variables
My Notes and Comments
Blanken
2001
Atmospheric Environment
281 commuters (aged 18-65) in Boulder/Westminster, CO.  Nov 1999 
yes
AWARE: 94% of persons aware and 93% know what it means.
Commuting: 76% of car commuters said never change commute mode, 12 % adjusted commute on red advisory days when convenient.
 
Self administered mail survey to 1,000 commuters.  76% of commuters were single passenger cars.  For the mandatory restriction of wood burning, participation may be biased (lawful desirability) a bit high.






Wood Combustion: 86% of those who burn wood said did not burn on red alert days (it is a mandatory restriction).
 

Blomquist
2004
Review of Economics of the Household
 
no
 
 
 
Mainly about VSL and WTP, theory and such, nothing relevant for our project that I could discern.
Bresnahan
1997
Land Economics
226 COPD (presumed adults mean age 48, mostly males) subjects over 2-5 days in a year (928 total days) in Los Angeles (July 1985-86), containing a "disproportionate number of individuals with compromised respiratory function"
yes
 
Leisure Activity: 39% said altered when "smoggy"
Experiencing symptoms was significant for activity and indoor (Table 3).  Having chronic respiratory impairments (asthma, hay fever, other) was not significant.  Education (HS diploma) was weakly significant (and positive) with respect to altered activity.   Their "composite averting variable" also significant with symptoms (Table 4).  
The author defined "smog-related symptoms" are used to indicate sensitivity to air pollution.   The symptoms (eye irritation, headache, chest tightness) are broadly applicable (e.g., allergens) though the inquiry includes the experience is aligned with "smoggy conditions" (page 345).  The authors do not define "altered", "more", and "smoggy".  In my opinion, HS diploma vs no HS diploma are a weak representation of "schooling" or education, even considering 1985.






Indoor Time: 40% said increased when "smoggy"








Ran A/C: 20% said when "smoggy"








Outdoor Time: reduced by few variables (see Table 5).   Outdoor time could be reduced by 40 min/day on bad ozone days for sensitive individuals.  Note also a reduction in the high temp by 4 degrees F would reduce outdoor time similarly to that of 0.1 ppm increase in ozone above 0.12 ppm (about 6 minutes per day).  
Significant: #workdays (-, most significant), hayfever (-), other chronic resp conditions (+), temperature (+), high O3 concentrations (> 0.12 ppm) .  Not significant: Asthma (-), symptoms (-+), humidity (-), pollutants (+-).   The variable symptoms is now virtually negligible, though when linked by interaction with pollutant concentrations, collectively are significant (though individually not). 
I wonder why "workdays" was not considered in the first set of analyses where "symptoms" were driving much of the averting response.  Respiratory conditions did not significantly influence the averting behaviors reported (table 3 and 4), though having hay fever reduced outdoor time by nearly an hour per day while having other chronic pulmonary conditions actually increased outdoor time by about (3/4) hour per day (Table 5).  This attribute was not well discussed.  The temperature variable (daily max) is not likely linear, as there is a range of values that will be comfortable  -  deviations will occur outside the range (too cold or too hot).
Bresahan
1995
Journal of Environmental Economics and Management
 
no
 
 
 
There is no data on averting, just theory and models.
Cheng
2007
Science of the Total Environment
AQ analysis at 5 monitoring stations in Taiwan using the 6 criteria pollutants (6 for AQI/RAQI, 5 for PSI  -  no PM)
no
 
Compared old PSI with standard AQI a revised AQI (RAQI).  The RAQI and AQI indicate more unhealthy AQ days (though mistermed "accuracy" in the paper)
 
There are inadequate comparisons of AQI to RAQI.  Conclusions are drawn regarding the "accuracy" of the RAQI and its ability to distinguish local sources from sandstorms, however, similar treatment is not given to the AQI.  Data in the text (pg 194) for one station indicate AQI more sensitive than RAQI.








Also, of course the AQI and RAQI are sensitive to PM2.5 than the PSI, they both use the PM2.5 data.
Courant
1981
J Environ Econ Manage
 
no
 
 
 
Logical discussion regarding willingness to pay for air quality and averting.  Conclusions are inconclusive and thus uninformative to the discussion.
Cutter
2009
J Env Econ Manage
Traffic counts and Mass Transit usage in San Francisco and LA CA, jun-oct 2001-2004
yes
 
Commuting: Auto traffic volumes decrease by 3-3.5% on STA days.  Rail use increases 1% (though not significant)
morning time of day significantly depressed volumes
STA is spare the air in SF.  LA does not have a corresponding program.
Henry
2002
Journal of Policy Analysis and Management
2,935 Adults (head of house) in 13 county metro Atlanta, GA.  Single-day rolling population survey over May-Sep 1998, 703 persons on alert days.
yes
AWARE: of O3?

Significant (+): Knowledge of alert, age, income, non-minority, day of week, month, urban location, read front page newspaper, commuting pattern.
Support for program appears to be an important variable, much like in the averting, recommendations from a health care professional are highly influential in modifying behavior.






Commuting: Significantly less miles driven on alert days (15.5% reduction) compared with non-alert day.  Number of trips was less on alert day but not significant.
No diff in gov't employees drive miles (and # of trips) (reinforced/supported SIP goals) vs non-gov. but on alert days, gov miles (and # of trips) were significantly reduced.

Kansas Department of Health and Environment
2006
Technical Report
~4,300 KS adults (BRFSS 2005 data, see Wen for all BRFSS)
yes
AWARE: 42.7% Total Population; "urban" Kansas City MSA (70%) compared with Wichita and Topeka (37%).
 
Health: asthmatic 46%, nonasthmatic 42.5% among other conditions.  Income (+), age (+), urban (+).
 



Asthma cohort defined by `Yes' to both "have you EVER been told by a doctor, nurse, or other health professional that you had asthma"? And 'STILL' have asthma?


Outdoor Activity Level and AQI: Total Pop: 15.2% of those aware of AQI changed outdoor activity level.
Asthmatic: 32.6%, Nonasthmatic: 13.7%; Urban (19.4%) non-urban (<= 10%); income (weakly -) 
In comparison with Wen (2009) data above, most of these percentages are on the lower end of the range indicating potential for regional differences in percentages, though overall may not be a very influential attribute.  The one small difference was asthmatics aware of AQI, Wen reported 31% reduced activity level (compared with 32.6% here for KS).





PERCEPTION: 6% of population felt poor AQ affects health
Outdoor Activity Level and Perception: Total Pop: 8.7% of population reduced outdoor activity level because they felt AQ was bad.  64% of those who perceived AQ as bad modified behavior based on AQI (only 8% of those who did not perceive AQ as bad and were aware of AQI modified behavior)
Asthmatic: 21.8%; Nonasthmatic: 7.7%; diability 14.0% nondisability 7.5%; urban 10.7%, non urban <8.2%;  income (weakly -), age (weakly +)
 






Outdoor Activity Level and Advice: Very few persons (4%-total, 15.7% asthmatics, 3.1% non)) received advice to reduce activity level when AQ poor
~41% of those advised did alter activity level at least once because they felt AQ was bad, 46% of those advised did based on AQI
The link between professional advice and attention to AQI is strong, though the word is not getting out regarding the AQI to the health professionals.
Mansfield
2003
Technical Report: Task 4 RTI report for BHubbell.
National survey, 6,106 persons aged 18+ years at county level, year 2000
yes
AWARE: 33% of total population aware of AQI 
 
age (+), gender (M-), race (blk+,w+), health (+), education (+, BA/BS graduates), employed (+),  income (+); increased to 37% when selecting county that had alerts currently/previously (+)
Table 4-1 is lacking some of the appropriate N's to reproduce the reported percentages.






Outdoor Time: Of those persons reporting high AQI accurately, 57% reported spending less time outdoors on ozone alert days (orange or greater).
male (-), white (-), high income (-), health (-), whether county had alerts currently/previously (+)
Note only 422 persons ID'd the high alert day and adjusted outdoor activities (that's 26%), suggesting again that most people that are aware still do not really adjust behavior.  It's probably because of the other vars: employed, income, thus time restricted already





 
Commuting: Of those persons reporting high AQI accurately, 56% reported driving less, (combined - 48%) on ozone alert days (orange or greater).
male (-), white (-), high income (-), health (-),whether county had alerts currently/previously (+)
About 28% of those aware of overall system (not those who were correct) modified behavior
Mansfield/Houtven
2009/2004
ASSA Meeting
469 parents of children 2-12 years old - OAB CHAD
yes
AWARE: 91%

asthma (+), number of children in family (+), white (+)
 






Outdoor Time: Stated behavior - 68% said would reduce time spent outdoors, 39% said would reduce activity level
time: asthma (-), red alert (-), number of children (+), temperature (-); level: asthma (-), red alert (-), temperature (+)
 



762 parents of children 2-12 years old - OAB CHAD


Outdoor Time: Diary Data - Indicate 20 minutes less time spent outdoors by air pollution sensitive asthmatics on code red days compared with others
Note Figure 6 shows in air pollution sensitive indivduals, the reduction is greatest during the afternoon hours
The general findings in Houtven (2003) are consistent with Alion WA4-29 memo written by graham glen, an analysis of the OAB data in CHAD (that is, no difference).  Reanalysis by Mansfield (2009) considered more personal attributes of the asthmatics
Mansfield
2006
Resource and Energy Economics -- see also Mansfield (2009)/Houtven (2004)
Sub sample of 231 parents of children 2-12 years, 45% asthmatic, from 35 MSA's with most AQ alerts, 2002 ozone season (OAB-CHAD study)
yes
AWARE: 61% said aware of AQI
 
 
Paper is mostly about WTP.  A working paper is cited for the second result, "averting behavior with respect to ozone alerts: do parents respond?"  I do not have this paper though it is found in Mansfield (2009) as 20 minutes.  This used some of the "OAB" diary data we have in CHAD.  





 
Outdoor Time: persons with asthma reduced time by 30 minutes on Red AQ days relative to Green, Yellow, or Orange days.
 
 
McDermott
2006
Journal of Asthma
Parents of nonasthmatic (98) and asthmatic (110) children > 4 years old in Salt Lake city, UT. Convenience survey.  Jan-Sep 2003
yes
AWARE: 88% aware of AQ alerts
 
95% asthmatics vs 80% non
The asthma cohort was defined as "parents who reported that their children had asthma", note 25% of asthmatics were moderate/severe, 72% believe are air pollution sensitive, 88% controlled.  No tests were done to discern any relationships of asthma severity with survey responses.





PERCEPTION: 91% felt poor AQ affects health

no sig diff in asthma vs nonasthma
 






Fossil Fuel use: 71% say reduce when AQ is poor 
asthmatics 79% vs non-asthmatics 61%
 






Outdoor play: 55% say restrict when AQ poor
64% asthma vs 45% non-asthmatics.
Restriction is largely limited to only a few days per year for both cohorts (between 0-5 days), though greater percentage of asthmatics (31%) restrict 3-5 days vs 17% for non asthmatics








Unclear whether metric is outdoor activity level and/or time.
Neidell
2009
The Journal of Human Resources -- Note: associated with Zivin (2009), Neidell (2010), Neidell and Kinney (2010) and Neidell (2005) Working Paper TR25.  
Attendees of two outdoor public arenas in Los Angeles for years 1989-1997, the LA Zoo (1,949) and Griffith Park Observatory (1,770).  
yes
 
Outdoor time surrogate: reduction in attendance at both public arenas on ozone `smog' alert days (Zoo 13-15%, Observatory 3-6%).
Children (22-25%), elderly (19%), and locals (GLAZA club members, 19%) were more likely to avoid outdoor zoo event in response to an alert than others.
1) GLAZA member cohort may be biased regarding age distribution (possibly having more children and elderly) compared to others. 








2) Not implicit as to whether outdoor time reduced, just attendance at these outdoor events.  Given increases (not statistically significant) in attendance at Dodger/Angel baseball games (control) for alert days: 1) there may be an active decision in response to the alert (avoid zoo/observatory and attend game instead) OR 2) having plans for baseball game (some coinciding with O3 alert day) precluded attendance at zoo/observatory.








3) In Neidell (2006), Similarly modeled attendance reductions for an Arboretum (10%) are similar to the Zoo (11%).






Asthma hospitalizations: Effects estimates (hospital admissions) are reduced by a factor of 2.6 in children and by a factor of 1.4 in the elderly when not accounting for alert info.  Not significantly different C-R function in adults.  Increase in hospitalizations per 10 ppb ozone: 0.037% children, 0.022% adults, 0.031 elderly.
 
4) While they conclude that the averting response for adults is insufficient to affect the C-R relationship, is it also possible that adults are not responding to ozone as would a susceptible population like children or elderly?
Neidell
2010
J Epidemiol Community Health -- Note: associated with Zivin (2009) and Neidell (2009).  
Attendees of two outdoor public arenas in Los Angeles for years 1989-1997, the LA Zoo (1,949) and Griffith Park Observatory (1,770).  
yes
 
Outdoor time surrogate: greatest reduction in attendance on ozone `smog' alert days (15%), compared with health advisory days (7%) and 'unhealthful' advisories (2%, not sig).  Reduction at observatory (3%) but no difference across advisory days.
Patterns among children, elderly and GLAZA were significantly higher than overall population as observed in Neidell (2009)
Not implicit as to whether outdoor time reduced, just attendance at these outdoor events.  







Time of day could be a significant influential factor, as the observatory is mostly an evening hour event
 
Neidell
2004
Journal of Health Economics
California children (0-18) asthma hospital admissions, 1992-1998, monthly data.
no
 
Asthma hospitalizations: O3 negatively correlated with hospital admissions, though if accounting for smog alerts, the effect became less. 
SES was determined significant factor for estimated ambient pollution levels for cohort as well as asthma hospitalization admission rate.
It seems unusual that CO is significantly correlated with asthma admissions, while other pollutants (with far better health effects relationships like O3 and NO2) are not.
Noonan
2010/2011
Fourth World Congress of Environmental and Resource Economists/AEI Working Paper 20011-08
Commuting: 8,069 Atlanta, GA households (21,323 persons), April 2001-02, no July.  Outdoor Park: 4,258 groups in summer 2005.  Also, ATUS data analysis 12,810 observations (2004-06) 
yes
 
Commuting: vehicle miles traveled (VMT) are 26% higher on smog alert days (>0.084 ppm) .  ATUS daytime (6AM-3PM) travel is reduced on smog alert days. 
income (+), education (+), weekday (+)
Noonan (2011) reports range of 3-18% higher VMT on smog alert days and not statistically significant, apparently using 2001 data alone.







Note overly influential were data for two long distance travelers (page 22)
Should have controlled for long distance travelers as they are likely doing planned trips, whether vacation or business related, would not likely be affected by alert.






Outdoor time surrogate: Park attendance: Elderly and exercisers appear less inclined to perform outdoor recreation on smog alert days.  Outdoor afternoon sports time not significantly affected on smog alert days.
 
Admittedly the Park generated data are potentially non-scientific, with graduate student judgments made regarding ages and activity levels.
Semenza
2008
Environmental Research
"Older" Portland (1,254) and Houston (708) residents (random), summer 2005-06, Mean ages 52 and 48, respectively.
yes
AWARE: Houston (89%) and Portland (82%) heard of ozone "action" or "watch" day
 
 
I could not reproduce some of the numbers reported in Table 4; I submitted an inquiry to 3 authors to explain; no response to date.  Unclear as to what health conditions were included, termed "extreme health".  Other health responses of "skin", "fever" and "muscle" and "nausea" were not significant wrt heat index and are also poorly defined.  Portland appears to respond to the heat questions which may be correlated with their answers regarding AQ.






Changed Behavior: 10.5% of Portland and 9.7% Houston residents changed behavior on AQ alert days, though largely in response to perceived poor air quality.

Behavior was said to be changed, not sure what constitutes a change.   Results comparing influential factors in the two cities certainly indicate regional differences in perception, though limited differences in "averting".


 
 
 
 
Commuting and Refueling: No change observed between advisory or control days in either city.
 
data not shown in paper, just mentioned.
Sexton
2011
Technical Report
~54,000 ATUS diaries (2003-10), 15-85 years in age
yes
 
Outdoor Time/Vigorous Activity:  Total population reduced vigorous outdoor activity time by about 21 minutes, participation was about 3%.  Elderly: 59% or 65 minute reduction, 15% participation, driving much of what is observed for total population.
Elderly respond greatest to higher alert levels (AQI 175-199).  Time of day: 31% or 18 minute reduction in afternoon, 79% or 10 minute reduction in morning. 
There are some issues with the ATUS data, mostly that the outdoor time at home is unknown, though some of the activity codes used could compensate (e.g., playing soccer "at home or yard" is likely outdoors).






Indoor Time/Vigorous Activity: No significant increase on alert days

For both of these, the total population was evaluated and not the elderly subgroup that appeared to be the only portion of the population averting.





 
Multiday Response: Not transitory. On two consecutive alert days, people appear to continue to avert
 

Shofer
2007
Am J Med Sci
 
no
 
 
 
Generally an informative plea for health care professionals to advise persons at risk of effects from bad air quality.  Article actively promotes AIRNOW, AQI, NAAQS, and EPA.  Creates a new acronym for use by clinicians, AIR: Ask, Inform, React 
Smith
1986
Economics Letters
425 households in Acton and Boston MA.  Acton had several incidents of hazwaste water contamination.
no
 
 
Install Water filters: positive (+) relationship with respect to individual perception of harm from hazwaste in water.
Older paper but serves to reasonably capture influences on one's risk perception, though deck is a bit stacked:  The influence of the media and study location are the strongest variables positively affecting the selected averting behavior, next to smoking which correlated with a negative adjustment to behavior.







Drink bottled water: age (+), hazwaste perception (+), confidence in water supply (-), smoker (-)








Attend public meetings: resident of Acton (+) and reader of hazwaste news articles on town (+)

Wen
2009
Journal of Community Health --(see also KS DOH, 2006)
2005 BRFSS data: About 24,000 nonasthmatic, 4,000 asthmatic persons in CO, FL, IN, KS, MA, WI MSAs.  Asthma cohort defined by `Yes" to "have you 'EVER' been told by a doctor, nurse, or other health professional that you had asthma"?
 
AWARE: 46-49% were aware of AQI or AQ alerts. 
 
Asthmatic (49%) and non-asthmatic (46%) aware
 






Outdoor Activity Level and AQI: no totals, just asthma cohorts
Asthmatics (31%) and non-asthmatics (16%) reduced outdoor activity level. Gender (Females less), Disability (yes, less)
Curious as to the correlation between number of black/non-Hispanic women age>35 BMI > 30 with a disability and under advisement of a doctor.  This could be an important cohort to explain differences between the asthma and non-asthma cohort?






Outdoor Activity Level and Perception:  no totals, just asthma cohorts
Asthmatic (26%) and non asthmatic (12%) persons say changed outdoor activity level because they felt AQ was poor.  Combined AQI awareness and perception of AQ as bad was influential in reducing outdoor activity level (75%-asthmatics, 68%-non-asthmatics).
Unclear what constitutes outdoor activity level, how much the change was, what is it relative too (are they very active persons or sedentary), and was the change also related to time spent outdoors.  No measure of whether it was as often as the reported AQ alerts or whether they did in fact do it on a day when AQ was poor.  Severity of asthma or medication usage is not indicated as measured or influential.  Also, region/state was not evaluated as influential.




 
 
Outdoor Activity Level and Advice: no totals, just asthma cohorts
Professional recommendation to reduce outdoor activity level when AQ is bad was a highly influential variable for both asthmatics (49%) and non asthmatics (42%)
The media alerts and perception of AQ as bad are important, however it appears that they are not linked very well, even with the professional advice being one of the most important variables. 
Whitehead
2005
Environmental and Resource Economics
895 persons, coastal NC.  Follow up:  490 same persons
no
 
Significant finding on relationship between what is said is done and what is actually done regarding averting (an assessment of predictive validity of averting modeling).
 
The magnitude of adverse consequences and the degree of awareness will surely drive the level of predictive validity of the model.  People are very likely to believe in the effects of a hurricane given the wide/thorough media coverage exhibiting the devastation associated with them.  This is may be why the rate averting air pollution is not high; it is likely that most people are not aware of the warnings or are confused regarding the effects.
Wilson
2005
International Journal of Social Economics
203 Farmers in Sri Lanka
no
 
Averting Related to Pesticide Exposure
Defensive exposure measures is significantly related to education level, number of crops grown, freq of pesticide usage, number of pesticides used.
They should have evaluated the classes of pesticides separately (or insecticides alone) as the mammalian health effects vary widely.








Somewhat bothered by the relationship of cost incurred = amount of safety. "Most pesticide users do not even wear shoes!"  However, it could be that variable protective expenditure serves as an indicator of associated training/education offered to prevent exposure.  This is may be the best way to effectively reduce exposure in these areas (not simply an 8[th] grade education), rather than the seemingly wanton use of protective gear that many may not find effective or culturally acceptable to prevent health issues from occurring.
Yen/Eiswerth
2004/2005
Rev Econ Hous
1,779 diary days of 64 adult asthmatics, Glendora, CA, Fall 1983.
yes
 
Indoor/Outdoor/Work Active/Inactive Time: No significant effect noted except for time spent 'inactive outdoor leisure' (decrease with O3 > 120 ppb 1-h) and 'indoor chores' (decrease with max daily 1-h O3).
Asthma severity was very important for decreasing outdoor time and increasing indoor time (Eiswerth, 2005)
The data are perhaps too old to observe an averting response (less likely people were aware and perceived air pollution as bad for health).
Zivin
2009
J Environment Econ Manag.  Note: associated with Neidell (2009) and (2010).  
Attendees of two outdoor public arenas in Los Angeles, the Zoo (1,878) and Observatory (1,554) from 1989-1978.  Sample size is slightly smaller than Neidell (2009 and 2010).
yes
 
Outdoor Time Surrogate:  Significant reduction in attendance at both arenas on ozone `smog' alert days (Zoo 15%, Observatory 8%).
Precipitation (68% reduction) and temperature are by far the greatest influential variables compared with the AQ alerts.
Interestingly, more people are concerned about getting wet than potential health effects associated with AQ.




 
 
Multiday Response: Transitory.  On consecutive alert days, fewer to no people avert (2nd day: 5% Zoo, and 0% Observatory, 3rd day: 0% Zoo, 0% Observatory, none significant).
Children 2-12 exhibit the greatest avoidance on day 1 (~20%) while other age groups are barely distinguishable from the general population (~15%).  However, children (<2 and 2-12) and elderly have less of a transitory response, they still are averting during consecutive alert days more so than typical adults.
GLAZA members may be comprised of a different population regarding age distribution (possibly having more children and elderly), possibly affecting conclusions (local vs not local). 
Zivin
2011
NBER Working Paper 17004
~1,600 CA Farmworkers over 155 days, 2009-10
yes
 
Outdoor Time Surrogate: farm worker schedule is not adjusted in response to ozone levels.
 
Although farmers in CA are only one type, this could have implications for outdoor workers in general.
Zujic
2009
Environ Monit Assess
Belgrade, Serbia
no
 
Used population weighted (and street canyon adjusted) air quality indices calculated from pollutants (SO2, BS, NO2) measured at available monitoring locations to better compare individual pollution contribution to poor air quality in an urban area. 
Not surprisingly, "black smoke" was the driver for most of the "poor", "critical" and "bad" AQ days.
Interesting article, though telling of environmental monitoring and research in other countries.  No ozone, a big weak spot.

