Do associations between airborne particles and daily mortality in Mexico City differ by measurement method, region, or modeling strategy?

Abstract

We evaluated whether associations between PM10 and daily mortality in Mexico City differ by the PM10 measurement device or by regional differences in particle composition. Additionally, we reanalyzed previously collected data in light of recent insights about flaws in commonly used time series analysis techniques. We examined daily associations between mortality and four indicators of ambient PM10 using Poisson regression, controlling for temperature and time trends with cubic natural splines. Associations were calculated for five subregions corresponding to five monitoring sites and pooled for the entire metropolitan area. PM10 was measured with three methods: Tapered Element Oscillating Microbalance (TEOM), Sierra-Anderson High Volume (Hi-Vol) and Harvard Impactor (HI), the latter only at one site. In addition, predicted values of daily PM10 were developed using the Hi-Vol measurements, which were taken every sixth day, and weather, visibility and other pollutant data. We assigned deaths to the exposure from the monitor nearest to their residence. We also re-evaluated the HI PM2.5 and mortality association in southwest Mexico City, which was estimated previously using nonparametric statistical models. Slight decreases in effect estimates were observed (a 1.45% increase (95% CI: 0.09%, 2.83%) in total mortality per 10 microg/m(3) increment of PM2.5 at lag 0) compared to a 1.68% change (95% CI: 0.45%, 2.93%) using the previously employed nonparametric approach. Using data pooled over all the regions, PM10 measured by the TEOM and the predicted PM10 values showed little association with mortality at any of the lags examined. The pooled estimates for Hi-Vol PM10 (using one sixth of the data) were positive across all lags examined and significant for lags 3 and 5. No consistent patterns of differing associations were seen across regions that would correspond with particle toxicity or composition. Particulate air pollution, measured with gravimetric methods, is associated with daily mortality and presents a risk to health in Mexico City. The reanalysis suggests that previous research is robust to statistical method and likely to yield the same overall conclusions about the short-term effects of airborne particles on mortality.