Source: Masters Abstracts International, Volume: 49-01, page: 0412.

Adviser: Michelle L. Bell.

Background: Many studies have investigated the relationship between short-term exposure to ambient ozone and respiratory hospital admissions by time-series analysis or case-crossover analysis. Some generated positive results, suggesting a significant adverse impact from ozone on human morbidity. But some studies reported negative or uncertain effect estimates for the relationship, resulting in difficulty in interpreting results. Although these studies all focused on the association between short-term ozone exposure and respiratory hospital admissions, the differences in study characteristics (e.g., age group selection, lag selection, exposure season selection) make the comparison between these studies difficult. Therefore, we conducted a meta-analysis of the association between short-term ambient ozone exposure and respiratory hospital admissions. We used meta-analytical methods to combine the estimates from previous studies taking each estimate's uncertainty into consideration. We also explored the sources of heterogeneity in study characteristics and examined how the factors influenced the relationship. Publication bias was also assessed. We compared estimates from meta-analysis of single-city with results of multi-city studies.

Methods: We included 136 single-city estimates in our analysis. For meta-analysis of the single-city study results, we applied fixed-effect or random-effect models to calculate the summary effect estimates for different hospitalization types (i.e., general hospital admission, emergency hospital admission, and ER/ED visits), disease categories (e.g., chronic obstructive pulmonary disease (COPD), pneumonia), and age groups (e.g., all ages, elderly). We also combined the estimates grouped by lag selection, exposure season of analysis and geographic regions separately to examine the sources of heterogeneity across studies. To examine publication bias, we generated funnel plots and applied the linear regression test, and then generated an adjusted estimated using the "trim and fill" method. Finally, we compared the meta-analysis based on single-city studies with 10 multi-city results.

Results: Although some individual studies did not demonstrate significant results, in the meta-analysis we found significant positive outcomes for most analyses. For instance, risk of emergency, total, or general respiratory hospital admissions increased 1.90% (95% CI: 0.74, 3.07%) per 10ppb increment of daily ozone among all ages and the effect estimate was 4.47% (95% CI: 2.48, 6.50%) among elderly. For cause-specific diseases, COPD all-age hospital admissions increased 5.06% (95% CI: 1.24, 9.05%) for the same ozone increment. Asthma hospital admissions rose 6.64% (95% CI: 2.60, 10.85%) per 10 ppb ozone among all-ages or adults, but for children the association was of 2.83% (95% CI: -3.45, 9.52%). The results indicated an adverse health effect of short-term ozone exposure. The analyses by age groups, lag selection, and the exposure season showed that these factors could affect the relationship; however the effects by comparison groups were not statistically different, suggesting the need of further investigation. Funnel plot and linear regression test provided some evidence of publication bias. After adjusting for the publication bias, effect estimates became smaller and results remained statistically significant. Generally, estimates from the single-city meta-analysis were higher than estimates from multi-city studies did, which is further evidence for the existence of publication bias.

Conclusion: From meta-analysis of single-city results, we found significantly increased risk of hospital admissions from short-term ozone exposure for several disease categories among different age groups. Future investigation on study characteristics' impact on the relationship is needed. This study found evidence of publication bias, which indicated that single-city meta-analysis could overestimate the result.