Results (
Thai) 1:
[Copy]Copied!
Considering the potentially nonlinear and lagged associationsbetween ambient temperature and adverse cardiovascular outcomes,we used a distributed lag nonlinear model (DLNM) to estimatethe effects of temperature on OHCD mortality (Ma et al.,2014). Specifically, to establish a “cross-basis” function of temperaturebased on the DLNM, we used a natural cubic spline with 5 dfto account for the nonlinear effect of temperature, and also used anatural cubic spline with 5 df to account for the lagged effects (lagspace) of temperature (Gasparrini, 2011). Because it is not easy todetermine the maximum lag of the effects, we alternatively usedmultiple lag intervals including days 0e3, 0e7, 0e14 and 0e21. Wethen introduced the “cross-basis” matrix of temperature into theGAM. We further controlled for time trends (7 df per year in naturalfunctions) and day of the week, as well as the same-day airpollutant concentrations (PM2.5 and O3). We first flexibly plottedthe relative risks (RRs) of the temperature-mortality associationcurves. Then, we calculated the RR comparing the 1st percentile oftemperature to the minimum-mortality temperature (MMT) andthe RR comparing the 99th percentile of temperature to the MMT.Further, to quantify these effects per an absolute change (1 C) intemperature, we calculated them as the log-RR divided by therange from the MMT to the corresponding temperature percentiles(Ma et al., 2014). In brief, the cold effect was defined as the percentincrease in daily mortality per 1 C decrease below the MMT, andthe heat effect was defined as the percent increase in daily mortalityper 1 C increase above the MMT.We also used the aforementioned models to analyze the effectsof air pollutants and temperature on IHCD mortality.To explore the potential interactions between air pollution andtemperature on OHCD, we performed an analysis of the effects ofair pollution on OHCD stratified by different temperature levels, i.e.,low temperature (<25th percentile), moderate temperature (between25th and 75th percentiles) and high temperature (>75thpercentile).We also conducted several sensitivity analyses to evaluate therobustness of our results for the association between air pollutionand OHCD. First, we changed the df per year in the smoothnessfunction for time trend from 4 to 10. Second, we controlled forlonger temperature lags using the moving averages of the 0e3, 0e7and 0e14 days. Third, we re-analyzed the data after excluding thedays with extremely high air pollution levels (above 95thpercentile).All statistical tests were two-sided, and values of P < 0.05 wereconsidered statistically significant. All analyses were performedusing R software (version 3.1.2, R Foundation for StatisticalComputing, http://cran.r-project.org/) with the GAM fitted usingthe “mgcv” package and the DLNM using the “dlnm” package.
Being translated, please wait..