A simulation study to quantify the impacts of exposure measurement error on air pollution health risk estimates in copollutant time-series models

Abstract

BackgroundExposure measurement error in copollutant epidemiologic models has the potential to introduce bias in relative risk (RR) estimates. A simulation study was conducted using empirical data to quantify the impact of correlated measurement errors in time-series analyses of air pollution and health.MethodsZIP-code level estimates of exposure for six pollutants (CO, NOx, EC, PM2.5, SO4, O3) from 1999 to 2002 in the Atlanta metropolitan area were used to calculate spatial, population (i.e. ambient versus personal), and total exposure measurement error.Empirically determined covariance of pollutant concentration pairs and the associated measurement errors were used to simulate true exposure (exposure without error) from observed exposure. Daily emergency department visits for respiratory diseases were simulated using a Poisson time-series model with a main pollutant RR = 1.05 per interquartile range, and a null association for the copollutant (RR = 1). Monte Carlo experiments were used to evaluate the impacts of correlated exposure errors of different copollutant pairs.ResultsSubstantial attenuation of RRs due to exposure error was evident in nearly all copollutant pairs studied, ranging from 10 to 40% attenuation for spatial error, 3–85% for population error, and 31–85% for total error. When CO, NOx or EC is the main pollutant, we demonstrated the possibility of false positives, specifically identifying significant, positive associations for copollutants based on the estimated type I error rate.ConclusionsThe impact of exposure error must be considered when interpreting results of copollutant epidemiologic models, due to the possibility of attenuation of main pollutant RRs and the increased probability of false positives when measurement error is present.Electronic supplementary materialThe online version of this article (doi:10.1186/s12940-016-0186-0) contains supplementary material, which is available to authorized users.

Generalized Additive Models with Principal Component Analysis: An Application to Time Series of Respiratory Disease and Air Pollution Data

“…Hence, the correlation structure in the data may attenuate the true RR‐estimate, which can lead to a false positive conclusion (this empirical evidence was also discussed in Dionisio et al . () in a different simulation scenario). Thus, if a GAM is fitted to time series variables, without mitigating the temporal correlation structure of the covariates as, for example, by removing this from the data, the RR‐estimate may not correspond to the true relationship between the variables.…”

“…The fit can be affected, for instance, by a wrong choice of the number of degrees of freedom in the smooth component and by the presence of auto‐correlation in the series under study, among others (see for example, Dionisio et al . ()). Some works that aim to solve these problems include Dominici et al .…”

“…Additionally, most of the references in the epidemiological research area related to the study of the association between pollution and adverse health effects usually consider only one pollutant whereas the population under study is exposed to a complex mixture of pollutants; a broad discussion of the effect of correlated measurement errors in time series on the relative risk estimates has recently been given in Dionisio et al . (). The choice of a simple model may be, in general, due to the fact that the pollutants are linearly time‐correlated variables, which implies bias in regression estimates since the presence of multicollinearity (the linear dependence between the covariates) can inflate the variance of the estimators.…”

“…In spite of its widespread use, many researchers claim that care is needed when applying the GAM to time series. The fit can be affected, for instance, by a wrong choice of the number of degrees of freedom in the smooth component and by the presence of auto-correlation in the series under study, among others (see for example, Dionisio et al (2016)). Some works that aim to solve these problems include Dominici et al (2002), who proposed a correction to the degrees of freedom in the smooth component, Dominici et al (2006), Lall et al (2011) and Michelozzi et al (2007), who have used lag-distributed models to relate the response variable to lagged values of a time-dependent predictor, and Figueiras et al (2005) and Ramsey et al (2003), who have proposed some approaches to control the problem of concurvity (the non-linear dependence that can remain among the covariates).…”

Environmental Exposure Mixtures: Questions and Methods to Address Them

“…Several approaches have been developed to formally address exposure measurement error. [40–43] However, with rare exception, [44, 45] these methods have not been applied to studies of exposure mixtures.…”