METHODS

Section:

ChooseTop of pageAbstractMETHODS <<RESULTSDISCUSSIONReferencesCITING ARTICLES

The CHS methods have been published elsewhere. Details about the fire study are provided online. In brief, the CHS consists of repeated annual health assessments to monitor the course of respiratory health. The fire questionnaire study focused on participants of two ongoing CHS cohorts, including one cohort of high-school students (17 to 18 yr old during the fire), originally enrolled in 1996, and a cohort of elementary-school children (aged 6 to 7 yr), recruited in 2002. The older student cohort included high schools from the 12 (9, 10) original CHS communities and the elementary-school cohort involved 13 communities (nine of which were the same) (10). The study protocol was approved by the institutional review board for human studies at the University of Southern California, and written, informed consent was provided by participating students and a parent or legal guardian of minors.

The 2003 Southern California fires peaked between October 20 and November 2. During November–December 2003, the high-school students and the parents of the elementary-school children received the fire questionnaire by mail (see online supplement) and/or during the first 6 mo in 2004 as an annex activity of the ongoing CHS. The reporting period referred to the “two weeks of the October 2003 fire period.” Although the first page of the fire questionnaire asked about health-related problems, the second page referred to exposure to fire smoke and personal measures taken to modify this exposure (including evacuation, wearing of masks, reduction in time spent outdoors, and changes in physical activity). To quantify exposure duration, questionnaire response categories included the following: “not at all,” “1–2 d,”“3–5 d,” “6–10 d,” or “all days” (i.e., up to 2 wk).

Objective smoke measurements (i.e., PM₁₀ [U.S. Environmental Protection Agency–approved Federal Reference Methods to quantify PM₁₀]) were available only on the community rather than on an individual level. PM₁₀ was the strongest marker of fire smoke pollution (1, 2). High concentration periods lasted approximately 5 d; thus, we used the 5-d mean PM₁₀ level to characterize fire smoke. Missing air-quality data required estimation procedures of the 5-d average PM_10. Five-day average PM₁₀ concentrations were estimated for 5 of the 16 communities. San Dimas, Glendora, and Anaheim had all 5 d (October 24–28) estimated, and San Bernardino had 4 d (October 25–28) estimated. Because Alpine was directly affected by the fires from October 26 to 28, PM₁₀ concentrations at Alpine were averaged over the 3 fire days with 2 d estimated (October 27–28). For more details, see text and Table E1 in the online supplement and Reference 1.

RESULTS

Section:

ChooseTop of pageAbstractMETHODSRESULTS <<DISCUSSIONReferencesCITING ARTICLES

Table 1 summarizes the study populations and participation. High-school students' participation rates during the first 8 wk (2003) reached 52.4%, whereas only 39.0% of the parents (younger cohort) returned the mail-in questionnaire. The extended distribution of the fire questionnaire during 2004 strongly improved response rates, ultimately reaching 95.5% in the older and 68.0% in the younger cohort.

Table 2 summarizes the distribution of reported fire exposure and the ambient levels of measured or estimated PM₁₀ (see Methods). Both the subjective and objective measures of fire smoke showed that communities not directly affected by local fires suffered substantial smoke exposure (e.g., Mira Loma, Riverside, and Anaheim).

Table 3 summarizes the prevalence of the reported outcomes, by cohort and asthma status. As expected, prevalence rates were much higher among individuals with asthma. Dry cough, medication, and physician visits were more frequently reported by parents of elementary-school children, whereas high-school students were more likely to report eye symptoms. Home loss due to fire was reported by 35 (0.75%) study participants. In Alpine and Lake Arrowhead, more than 3% of study participants lost their homes (n = 15 and 10, respectively).

The main results are summarized in Table 4. Six or more days of fire smell indoors was significantly associated with all outcomes, and the smaller risk estimates for 1 to 5 d of exposure reached statistical significance in all but two outcomes (asthma attacks and bronchitis). Having fire smoke smell indoors for more than 6 d was associated with more than fourfold higher rates of eye symptoms, approximately threefold increased rates of dry cough and sneezing, and more than twofold higher rates of cold, sore throat, wet cough, medication use, physician visits, and missed school due to symptoms. The three types of wheezing (general, sleep-disturbing, and speech-limiting) occurred 3.5, 4.9, and 5.5 times more often, respectively, among those with 6 or more days of fire smell indoors. Asthma attacks increased 63%. The trend across the different levels of fire smell duration was highly significant for all outcomes except for asthma attacks (p = 0. 12).

The between-community comparisons were analyzed with two different metrics, namely PM₁₀ and the community mean response to the fire smell question. Results are presented for the former only (Table 4) as they were similar for both metrics. The community mean of the reported level of fire smoke indoors and the estimates of the 5-d mean PM₁₀ were highly correlated (r = 0.81) in both high-school and elementary-school students. Comparing highest with lowest community exposures, the between-community results were statistically significant and similar for both metrics in case of dry cough, eye, nose, and throat symptoms, as well as for medication. Eye symptoms were approximately three times as frequent in the communities most affected by fires as compared with lesser-affected communities. The between-community estimate for wheezing was significant only with the mean reported smoke (odds ratio [OR], 1.37 per unit change) but not with PM₁₀ (Table 4). The other outcomes were not significantly associated with either community-level metric.

The model presented in Table 4 also adjusted for asthma status before the 2003 fire period (“physician-diagnosed asthma”). Therefore, the model also estimates the contribution of physician-diagnosed asthma to symptom frequency in the end of October 2003, independent of the fire smoke. We present these effects (ORs) in Table E3 to highlight the much higher symptom rates among children with asthma (see also Table 3). Children with asthma were two to three times more likely to report symptoms than nonasthmatic children. Thus, the effect of having asthma was similar to the effects of fire smoke. In the case of wheezing, asthma status was more strongly related to the symptom (OR = 7.4; see Table E3.) than fire smoke (OR = 3.5; see Table 4).

The effect of fire smoke was, however, not restricted to children with asthma. Results of Table 4, stratified by asthma status, are presented in the online supplement (Tables E4 and E5). In fact, among nonasthmatic children, coefficients were either very similar or stronger (wheezing) than in children with asthma (n = 706) in whom point estimates tended to be smaller and not statistically significant for speech-limiting wheezing, asthma attacks, bronchitis, cold, wet cough, physician visits, and missed school. The between-community estimates followed a similar pattern as in nonasthmatic children, with significant associations among nine questionnaire items.

To evaluate the joint effects of fire and asthma status on reported symptoms, we examined five indicator variables for the combinations of fire smell (none, 1–5 d, ⩾ 6 d) and asthma status (yes/no) using nonasthmatic children without fire exposure as the reference group. Figure 2 presents the effects of fire smoke among children with and without asthma.

Figure 2. Effect of reported smell of fire smoke indoors (during 1–5, > 5 d, respectively) for four symptoms among children with and without asthma. Odds ratios and 95% confidence intervals from models with interaction terms for asthma and fire, adjusted for sex, ethnicity, educational level of parents, and cohort. No asthma (No Ast), no fire = reference.

Download Figure | Download Powerpoint

[More] [Minimize]

DISCUSSION

Section:

ChooseTop of pageAbstractMETHODSRESULTSDISCUSSION <<ReferencesCITING ARTICLES

To our knowledge, this is the largest investigation of acute effects of wildfire smoke on children's health. We confirmed very substantial effects of wildfire smoke exposure on eyes as well as upper and lower respiratory symptoms, in both children with asthma and nonasthmatic children. The study was population based; thus, findings may be generalized more broadly to other comparable populations. Our findings are consistent with other studies conducted after wildfire outbreaks and occupational studies among firefighters, which suggest that wildfire smoke leads to acute exacerbations of respiratory and eye symptoms and increased demand for health services (13). Like ambient urban air pollution, wildfire smoke contains numerous primary and secondary pollutants, including particles, polycyclic aromatic hydrocarbons, carbon monoxide, aldehydes, organic acids, organic compounds, gases, free radicals, and inorganic materials with diverse toxicologic properties (14), which may explain the wide range of acute symptoms observed in our survey (15).

Biases require particular attention in the interpretation of these findings. Because many parents and students completed the fire questionnaire several months after the fire (from 1 to 7 mo later), and because both exposure and outcome are reported by participants, the study may be subject to interrelated reporting, recall, and selection biases. Due to the lack of individual-level PM₁₀ data, we were able to compare effects of objective (PM₁₀) and subjective (reported) markers of exposure in the between-community comparison only. We used the community mean of reported fire smoke as the subjective aggregate exposure.

For reported fire smoke, the estimates for individual and community mean were similar for most outcomes. However, between-community estimates using the mean reported fire smell were not entirely consistent with those based on PM₁₀. The latter showed no clear association with cold, cough, asthma symptoms, physician visits, and missing school. There are several possible reasons for these inconsistencies.

First, the exposure metrics are inherently different and measure different domains of exposure. PM₁₀ estimates the average concentration during the 5 most extreme days. In contrast, the questionnaire-based approach relates to the duration (i.e., number of days of observed smoke) rather than the level of the smoke in the community. Duration may characterize the true contrasts in exposure better than the 5-d average PM₁₀ because some communities experienced fire smoke for longer or shorter periods.

Second, PM₁₀ levels had to be estimated for five fire communities (see Methods and online supplement). The unknown errors in these estimates may lead to under- or overestimation in the between-community effects. Thus, the results based on “objective” measures of community-level exposure are not necessarily unbiased.

Third, PM₁₀ community levels are not sensitive to spatial differences in smoke densities that may have occurred within communities. Therefore, PM₁₀ concentrations at the monitor may not represent the mean of the true, but unknown, home outdoor PM₁₀ levels. We have no objective data to validate the reported diversity on the individual level. Wu and colleagues estimated PM₁₀ distributions all across the Southern California area during the wildfire period, using PM measurements, light extinction, meteorologic data, and smoke information from satellite images (1) (see Figure E1). We used these results to investigate the range of daily mean PM₁₀ concentrations for small areas representing size and location of several CHS communities. For example, the PM₁₀ concentration estimates for a 1 × 1–km grid within a 10-km buffer around San Dimas indicated substantial temporal differences during the fire period, with daily means ranging from 115 μg/m³ (October 28) to 220 μg/m³ (October 26) as well as large spatial gradients across the grid points. For example, on October 25, the point estimates ranged from 54 to 250 μg/m³, and from 90 to 337 μg/m³ the next day, with spatial standard deviations up to 50% of the daily means (see Figure E1). Although these PM₁₀ estimates may be associated with significant uncertainties at the neighborhood scale, they demonstrated substantial spatial heterogeneity, which corroborates the notion that smoke concentrations may vary substantially within communities. The distribution of reported smoke—and thus the community mean of the reported conditions—may reflect these distinct spatial gradients that are influenced by topography and wind patterns (1, 2).

Fourth, the community-level PM₁₀ does not take into account PM₁₀ levels in locations to which the children might have been evacuated, nor does it account for other individual preventive action taken during the fire period. Thus, the monitor PM₁₀ value may again be offset from the true, but unknown, mean PM₁₀ across children.

Fifth, the reported fire smell related to the indoor environment where most of the time was spent, whereas outdoor PM₁₀ levels are not sensitive to differences among children's indoor environments.

The community mean of the reported fire smell was highly correlated with measured PM₁₀. However, others have shown that community mean reported annoyance of ambient air pollution correlates highly with objective measurements, whereas individual scores may poorly correlate with the home outdoor NO₂ measurements (16). Reporting was associated with health status and sex. A recent review also concluded that reported exposure to traffic was poorly associated with objective data (17). It is not clear whether findings for reported ambient air pollution also apply to fire smoke perception. Reporting of fire smoke may be less affected by personal attitudes than reported ambient air pollution, given the strong smell of fire smoke, the visibility of the problem, and the exceptional situation of the fire period. Reporting was also associated with sex. Eye symptoms, cold, medication, and physician visits were significantly more often reported among girls, whereas boys were more likely to report wheezing (data not shown). However, sex did not confound nor modify the main effects of fire smoke.

In conclusion, although it is neither possible to dismiss the possibility of biases nor to quantify their effects on our results, we believe that the arguments outlined above support the questionnaire-based results.