Air pollution and risk of stroke: underestimation of effect due to misclassification of time of event onset.

BACKGROUND: Epidemiologic studies linking ambient air pollution to the onset of acute cardiovascular events often rely on date of hospital admission for exposure assessment.
METHODS: We investigated the extent of exposure misclassification resulting from assigning exposure to particulate matter based on (1) date of hospital admission, or (2) time of hospital presentation compared with particulate matter exposure based on time of stroke symptom onset. We performed computer simulations to evaluate the impact of this source of exposure misclassification on estimates of air pollution health effects in the context of a time-stratified case-crossover study.
RESULTS: Among 1101 patients admitted for a confirmed acute ischemic stroke to a Boston area hospital, symptom onset occurred a median of 1 calendar day before hospital admission (range = 0-30 days). The difference between ambient particulate matter exposure based on the calendar day of admission versus time of symptom onset ranged from -47 to 36 microg/m3 (-0.1 +/- 7.1 microg/m3; mean +/- SD). The simulation study indicated that for nonnull associations, exposure assessment based on hospitalization date led to estimates that were biased toward the null by 60%-66%, whereas assessment based on time of hospital presentation yielded estimates that were biased toward the null by 37%-42%.
CONCLUSIONS: Epidemiologic studies of air pollution-related risk of acute cardiovascular events that assess exposure based on date of hospitalization likely underestimate the strength of associations. Using data on time of hospital presentation would marginally attenuate, but not eliminate, this important source of bias.