was statistically insignificant when using longer lag periods. Significant effects of cold exposure were observed for all lag periods examined. For example, a 1 C decrease in temperature below the MMT corresponded to a 3.81% (95%CI: 1.48%, 6.14%) increase in OHCD mortality when using lags of 0e21 days. The effects of hot temperature were restricted within the first 3 lag days. A 1 C increase in temperature above the MMT lead to a 4.61% (95%CI: 0.88%, 8.34%) increase in OHCD mortality when using lags of 0e3 days. We did not observe any significant effects of temperature changes on IHCD mortality, regardless of lag length used.
Table 4 shows that the effects of air pollution on OHCD mortality varied considerably at different temperature levels, suggesting that there may be interactions between air pollutants and temperature. For all air pollutants, the strongest effects occurred when the daily temperature was low. For PM10, PM2.5 and O3, the effects were larger at high temperatures than at moderate temperatures.
Results from sensitivity analyses of air pollution and OHCD mortality (data not shown) demonstrated that the effect estimate decreased appreciably when the df used were increased from 4 df