Methods

Study population

Subjects eligible for this study were individuals with a self-reported physician diagnosis of asthma who resided in the SJV region and for whom the CHIS collected health data between November 2000 and September 2001. This restriction to individuals with asthma diagnoses minimises the possibility of misclassification of outcome due to other respiratory diseases. CHIS is a biennial random-digit dial telephone survey of California adults, adolescents and children, and in 2001 was conducted in 55 428 California households. The weighted extended interview completion rate for adults was 64%, and the weighted adult screener completion rate was 59%. The interviews were conducted in English, Spanish or one of four Asian languages to obtain information on demographic characteristics, health-related behaviours, health status and conditions, access to healthcare and insurance coverage. Respondents were also asked to report their residential zip code. Detailed descriptions of CHIS 2001 sampling and survey methods are available elsewhere.6

Interviews were completed for 10 307 individuals residing in the SJV study area. For the following analyses, we selected 1502 (14.6% of interviewed) respondents who reported ever having been diagnosed by a physician as having asthma. The actual study population varied for each model depending on the availability of covariate data and pollutant measurements. The University of California, Los Angeles Institutional Review Board approved this study as being exempt from review.

Exposure assessment

We restricted our study population to individuals residing within zip codes in relatively close proximity (5 miles ≃ 8 km) to a California Air Resources Board or local air-quality management district monitoring station. Using US Census Block (2000) population densities, we located the population-weighted centroid of each zip code in which one or more respondents resided and assigned each of these centroids to the nearest monitoring station measuring a specific pollutant within an 8 km radius. The distance from each population-weighted zip code centroid to each air-monitoring station was assessed using ArcView GIS software (Version 3.3; ESRI, Redlands, California). For each of five pollutants (CO, NO₂, O₃, PM₁₀ and PM_2.5), we selected the nearest station within 8 km with available data for a given pollutant. Consequently, for some zip codes, measurements of specific pollutants may have been taken at different stations. Sensitivity analyses were performed using 3-, 6-, 8- and 16 km distances for linking residential zip code centroids to monitors. The 8 km range was selected to ensure a radius large enough to balance the need for a sufficient sample size against a potential increase in exposure misclassification with increasing residential distances from monitoring stations.7

Annual average concentrations of the five pollutants were estimated for each subject within the 8 km radius using data collected at the assigned stations for the 1-year period prior to the interview date. These averages were based on hourly measurements for the gaseous pollutants (CO, NO₂ and O₃); and 24 h average measurements for PM₁₀ and PM_2.5 (with most stations recording measurements every 6 and 3 days for these pollutants, respectively). In the SJV, 11, 18, 21, 15 and 11 stations provided CO, NO₂, O₃, PM₁₀ and PM_2.5 measurements, respectively.

Statistical methods

We employed logistic regression to evaluate associations between our air pollution metrics and asthma morbidity. Specifically, we examined differences in our exposure metrics for: (1) respondents with asthma reporting daily or weekly symptoms (participants were asked in a single question to report frequency of asthma symptoms, such as coughing, wheezing, shortness of breath, chest tightness and phlegm production) in the previous year versus those reporting less-than-weekly symptoms; and (2) respondents with asthma reporting at least one asthma-related ED visit or hospitalisation in the previous year versus those not reporting such healthcare utilisation. Regression analyses incorporated sampling weights to take unequal probabilities of selection into the CHIS sample into account. In addition to evaluating the pollutants (CO, NO₂, O₃, PM₁₀, PM_2.5) as continuous measures, we also looked at quartiles of their distribution in the study population. Exposure at levels below the 25th percentile was used as the referent category for each pollutant. Pollutant associations were evaluated in single- and multipollutant crude and adjusted models.

We evaluated changes in point and 95% CI estimates when including the potentially confounding risk factors—age, race/ethnicity, poverty level, gender, insurance status, delays in care for asthma, cigarette smoking (adults only) and employment (adults only)—in our models (table 1). Poverty level was used as an aggregated indicator of socio-economic status, while insurance status served as an indicator of access to care. Delays in care for asthma, cigarette smoking and employment were not included in the final models because they did not change the air-pollution-effect estimates by more than 10%. Consequently, the final adjusted models presented here include age, gender, race/ethnicity, poverty level and health insurance status.

Results

Overall, 25.7% of respondents with asthma in the SJV reported experiencing daily or weekly symptoms in the past 12 months (table 1). The prevalence of daily or weekly symptoms increased with age, with older people (65 years and older) being most likely to report having frequent symptoms. Those reporting being currently uninsured were more likely to report daily or weekly symptoms than those who were insured. In line with this, respondents who experienced delays in care for their asthma were also much more likely to report daily/weekly symptoms than those who never experienced delays in care. Adults who were unemployed, currently smoking or previous smokers were also more likely to report daily or weekly symptoms.

We observed an overall prevalence of 9.2% for asthma-related ED visits or hospitalisations in the past year among SJV respondents who reported a previous diagnosis of asthma, with the highest prevalence in children (≤17 years of age) (17.3%). Latino, Asian/Other and African–American respondents reported a higher prevalence of ED visits or hospitalisations than Caucasians, with the rate for Latinos being approximately double that of Caucasians. The prevalence of ED visits or hospitalisations was also higher in lower-income groups. Respondents who reported delays in care for their asthma were almost three times more likely to have visited the ED or to have been hospitalised than those who did not report delays in care. Adults who reported that they previously smoked and those who were currently unemployed were also more likely to visit the ED or be hospitalised for their asthma.

Interquartile ranges and correlation coefficients for respondents' annual average pollutant measures are shown in table 2. The annual average O₃ was positively correlated with both PM₁₀ and PM_2.5, and negatively correlated with NO₂ and CO in the SJV during our study period.

Overall, we did not observe any associations between annual average concentrations of NO₂ or CO and the outcomes of interest. Thus, we limit our discussion of results to O₃, PM₁₀ and PM_2.5. Crude and adjusted effect estimates (ORs and 95% CIs) for each pollutant, as continuous and categorical measures from single pollutant models, are shown in table 3. ORs based on multipollutant models for O₃ combined with PM₁₀ or PM_2.5 did not differ substantially from single-pollutant estimates (results not shown).

Based on the adjusted models, we observed a 23% increase in the odds of daily or weekly symptoms per 10 ppb increase in annual average O₃ (OR 1.23, 95% CI 0.94 to 1.60), a 29% increase per 10 μg/m³ increase in annual average PM₁₀ (OR 1.29, 95% CI 1.05 to 1.57) and an 82% increase per 10 μg/m³ increase in PM_2.5 (OR 1.82; 95% CI 1.11 to 2.98). Based on quartiles of annual averages of the pollutants, we observed positive exposure-response trends for PM₁₀ and PM_2.5.

We observed a 49% increase in prevalence of asthma-related ED visits or hospitalisations per 10 ppb increase in annual average O₃ (OR 1.49, 95% CI 1.05 to 2.11) and a 29% increase in odds per 10 μg/m³ in PM₁₀ (OR 1.29, 95% CI 0.99 to 1.69) after adjusting for age, gender, race/ethnicity, poverty and insurance status. Although annual average PM_2.5 exposures above the level of the reference group (≥25th percentile, 17.9 μg/m³) appeared to double the risk for hospitalisations or ED visits, the CIs around these estimates were quite wide, perhaps due to the smaller sample size available for this outcome.

Age-stratified results are shown in table 4. Among children, there was a 63% increase in ED visits and hospitalisations per 10 ppb increase in O₃ (OR 1.63, 95% CI 0.95 to 2.81). No associations between daily or weekly symptoms and exposure to O₃, PM₁₀ and PM_2.5 were observed for children. In contrast, the odds of daily or weekly symptoms were increased for all three pollutants among adults, with a 40% increase in this outcome per 10 ppb increase in O₃ (OR 1.40, 95% CI 1.02 to 1.91), a 43% increase per 10 μg/m³ PM₁₀ (OR 1.43, 95% CI 1.13 to 1.82) and an almost threefold increase in the odds of frequent symptoms per 10 μg/m³ change in PM_2.5 (OR 2.96, 95% CI 1.60 to 5.50). The odds of adult ED visits for asthma increased 30–60% per 10 ppb or μg/m³ increase in O₃ and PM, respectively, but 95% CIs for all estimates included null values.

Discussion

Few studies have examined associations between outdoor air pollution and uncontrolled asthma in the SJV, an area with some of the highest pollutant levels in the USA. Our results indicate that both O₃ and PM in the SJV have adverse health effects on individuals with asthma. We observed associations between annual average concentrations of O₃, PM₁₀ or PM_2.5 and frequent asthma symptoms or asthma-related ED visits and hospitalisations while adjusting for insurance status, poverty level, race/ethnicity, gender and age.

Comparison with previous studies

Our results agree with existing evidence that O₃ can impact asthma morbidity.8 9 Studies have linked short-term O₃ exposures with increased asthma-related hospitalisations and ED visits among children.10–12 Fewer studies have looked at the effect of O₃ on adults with asthma, but several have shown an association with increased asthma exacerbations and ED visits.13 14

Our results agree with previous studies linking PM₁₀ with asthma symptoms in adults.15 Using 5-day averages and controlling for aeroallergens, one study showed an association with increased ED visits and both PM₁₀ and PM_2.5 in individuals with asthma.16 Our study population represents mainly urban residents (including suburban) in SJV (94–97% urban depending on pollutant) because we restricted our study population to respondents residing in zip codes where the population centroid fell within 5 miles (≃ 8 km) of a monitoring station. Since urban areas in SJV are moderately sized and surrounded by agricultural lands, sources of ambient particles differ from other metropolitan areas. The sources of PM₁₀ in Fresno and Bakersfield are, in descending order, dust from roads and agricultural activities, secondary ammonium nitrate formed from mobile and stationary combustion sources, wood burning and secondary ammonium sulfate, while the major sources of PM_2.5 are ammonium nitrate, organic and elemental carbon from combustion, and ammonium sulfate.17 Limited studies have been carried out on the health effects of this unique mix of particles. An experimental study showed that concentrated ambient fine and ultrafine particles from Fresno air, a mainly urban area (including suburbs), consisting primarily of ammonium nitrate, organic and elemental carbon, and metals, caused inflammatory changes in the airway lining fluid of healthy adult rats.18 Another ongoing study in the SJV, the Fresno Asthmatic Children's Environment Study aims to determine the short- and long-term health effects of PM and to identify which components are responsible for the exacerbation of symptoms in children with asthma in Fresno, California. Some of their preliminary data suggest that traffic may be an important contributor to decreased lung function in children with asthma.19 If the traffic-related fraction of PM_2.5 is the most important contributor to asthma symptoms in some of our study population in SJV, there is likely exposure misclassification in our estimates, since ambient monitoring data from stations deliberately sited away from major roadways do not adequately represent exposures to some traffic-related pollutants (eg, ultrafine particles).20 21 This misclassification may also explain why we did not observe any effect for NO₂, an indicator for traffic-related pollution, in our analysis.

Age-specific effects

Our analyses stratified by age (1–17 years vs 18+ years) suggest that exposures to O₃ and PM₁₀ are associated with ED visits and hospitalisation in children and adults with asthma, though associations seem stronger for children, and association estimates are imprecise due to small sample sizes for ED visits and hospitalisations. On the other hand, associations for daily or weekly symptoms were only observed for adults in our study. For children, the absence of associations for daily or weekly symptoms may be in part due to outcome misclassification, since symptoms were parent-reported for the 12 months prior to the interview for children younger than 12 years old.22 For adults, ED visits and hospitalisations may be less common because they are more likely to self-treat for an attack at home with medication, while worried parents may be more likely to take their children to the ED at the first sign of an attack.23

EPHT

The goal of EPHT is to develop a tracking system that integrates data about environmental exposures with data about diseases that are possibly linked to the environment.5 The results of this study enhance our confidence in using survey data, such as the CHIS, linked with air-quality monitoring data for EPHT at the regional level. Our findings also suggest a few lessons about the use of population-based surveillance data for EPHT. First, it is very important to ensure a sufficient sample size at the geographic resolution of interest, for example air basin or county, to detect regional variations in associations between environmental exposures and health effects. Second, zip code information for the study population can be used to examine associations between health effects and air pollutants, especially those relatively homogeneously distributed within communities (eg, O₃). Third, since many chronic diseases have multiple causes and are influenced by many factors, it is important to control for confounding factors related to socio-economic status, access to healthcare and health risk behaviours. In this regard, CHIS has advantages over many administrative data sources such as vital statistics, hospital discharge data or claims data which do not have information on these confounders.7

The use of CHIS 2001 data for EPHT does have some limitations. We did not have any information on duration of residence in the same house and neighbourhood for the study population to evaluate whether exposures preceded the outcome events, though subsequent CHIS will have the information. We also did not have any information on how much time respondents spent at or near their home, so some misclassification would result for respondents who worked, went to school or otherwise spent a considerable amount of time in areas away from home that are characterised by different air pollution levels. In addition, the growing popularity of cell phones is a concern for landline-telephone surveys. According to the 2001 Current Population Survey data for the year 2001, only 2.2% of California households and only 5.2% of households below the poverty level did not have a telephone line. SJV has one of the highest poverty levels in the state (CHIS 2001). If poor people, who are more likely to be exposed to higher levels of air pollution and also more likely to suffer from uncontrolled asthma, are less likely to be sampled, the associations we report here for air pollution are probably underestimates.

Study outcomes and a prior asthma diagnosis were self-reported by respondents and not verified by objective measures. Because clinical measurements of airway responsiveness appear to reflect the activity and severity of asthma at the time of measurement only, it is generally acceptable to collect data on long-term prevalence of symptoms and exacerbations by questionnaires.24 However, due to variations in physician practices and a lack of access to care, especially in uninsured and low-income populations, underdiagnosis of asthma may be a concern.25 26 In addition, we recognise that the outcome of ED visits or hospitalisations is problematic, given that some of the CHIS respondents may have visited the ED for asthma care that was not emergent. Such visits are likely motivated by asthma symptoms, and thus relate to asthma care needs. Finally, while including only CHIS respondents with currently active asthma would have been desirable for the study of associations between air pollutants and control of asthma, only information on ever diagnosis of asthma was available from the CHIS 2001 survey. The inclusion of respondents with inactive asthma, therefore, may have led to an underestimation of the strength of pollutant–asthma outcome associations.

Conclusions

In summary, we observed that both O₃ and particulate matter are associated with frequent asthma symptoms and asthma-related ED visits or hospitalisations in the SJV of California. The results of this study also enhance our confidence in using California Health Interview Survey data linked with air monitoring data to track the impact of air pollution or policy changes designed to decrease air pollution in specific regions of California. Further study is needed to determine how changes in the contributions of various sources of pollution will affect the health of residents, particularly in vulnerable subpopulations.