One limitation of this study is that we could not link the death data to the previous medical records of the decedents, so that we were unable to abstract the pre-existing comorbidities for each subject. Although the subjects died of cardiorespiratory diseases may probably have the pre-existing chronic cardiorespiratory diseases, modification of the specific primary causes of death identified in this study may still not exactly represent the modification of the pre-existing comorbidities. Secondary, air pollution concentrations averaged from general monitoring stations were assigned to all the decedents. This is the inherent ecological error and may introduce some misclassification of the pollution exposure, fortunately, such bias is non-differential.Another limitation came from the case-only study design, which could not estimate the main effect of the pollution but only the effect modification by certain characteristics. However, the main effects of air pollution on mortality have been well studied in Hong Kong ) and documented in the literature, identifying the modification is the main purpose of this study.