Table 3 Percentage change in all-cause mortality associated with a 10 μgm−3 increase in black smoke for the cool (October to March) and warm (April to September) periods
 Parameter % Increase (95% CI) Black smoke0 (lag 0) 0.46 (−0.43 to 1.35) Black smoke1 (lag 1–6) 0.03 (−0.22 to 0.28) Black smoke2 (lag 7–12) −0.07 (−0.31 to 0.17) Black smoke3 (lag 13–18) 0.16 (−0.07 to 0.40) Black smoke4 (lag 19–24) −0.12 (−0.35 to 0.11) Black smoke5 (lag 25–30) 0.15 (−0.07 to 0.37) Black smoke0*season −0.29 (−1.20 to 0.62) Black smoke1*season −0.05 (−0.31 to 0.21) Black smoke2*season 0.12 (−0.14 to 0.37) Black smoke3*season −0.04 (−0.29 to 0.20) Black smoke4*season 0.08 (−0.16 to 0.32) Black smoke5*season −0.11 (−0.34 to 0.12)
• All models contain terms for season and other long-term trend, day of week, black smoke (lag 0, 1–6, 7–12, 13–18, 19–24 and 25–30 days), temperature (modelled as a double linear relationship, lagged 0, 1–6, 7–12, 13–18, 19–24 and 25–30 days), and season (taking the value “1” if period is cool period October to March and “0” if period is warm period April to September), plus interaction terms defined as the product of black smoke and season.

• Interpretation of coefficients: eg, black smoke effects at lag 0 would be 0.46% in the warm season and approximately 0.17% (0.46 to 0.29) in the cool season (this calculation would be carried out on the original model coefficient scale). All lag periods were included in the model simultaneously. Coefficients refer to a 1 day increase in black smoke concentration within each lag period.