2716

Associations between exposure to air pollution and health outcomes are usually investigated through three types of epidemiologic study designs: time-series designs, case-crossover designs, and cohort designs. The first two types can be applied with case-only data, which has great potential when cohort data is limited. Associations between short-term exposure and health outcomes are investigated via these two types. However, associations between long-term exposure and health outcomes are hardly investigated. One of the reasons is that when controlling for seasonality and long-term time trend, estimates of coefficients of lagged variables or moving averages that exceed roughly two months can become unusual, making it difficult to make inferences. Our preliminary results show that this problem is at least partially related to model specifications of adjusting for seasonality and time-trend. We argue that time-series designs may be used to estimate linear associations between long-term exposure and health outcomes, although exposure measurement error remains an issue to solve.