Short term exposure patterns in time-series studies of air pollution and health

Abstract

Background: Air pollution related risks on health are often reported using percentages. Hence, a 1% increase in mortality risk per 10 unit increase in pollution compares two days say with 2 and 12 units; the days need not be sequential. But a change from 2 to 12 could follow different patterns―increase by 2 units each day for 5 days, all in one go over 1 day, etc. As conventional studies don’t take into account such variation in exposure pattern, our study aims to assess its impact on risk estimates.

Methods: Using data on particulate pollution (PM) from London we defined exposure patterns on each day by counting number of
1-Positive changes in PM between successive days (delta)
2-PM peaks
3-Delta peaks
for the week just before the mortality day. We then investigated the association between PM and mortality accounting for exposure patterns.

Results: For a 10 unit increase in PM, the percentage increases in mortality (with 95% confidence interval) were 1.81(1.21–2.41), 6.46(3.55–9.45) and 2.28(0.8–3.78) for number of positive deltas≥4, PM10 peaks=3 and delta peaks=3 respectively. These were well above conventional estimate, 0.82(0.45–1.18), which ignored exposure patterns.

Conclusion: Allowing for exposure patterns could influence risk estimates.