Effect of the shape of the exposure-response function on estimated hospital costs in a study on non-elective pneumonia hospitalizations related to particulate matter.

Abstract

Objective: We used log-linear and log-log exposure-response (E-R) functions to model the association between PM2.5 exposure and non-elective hospitalizations for pneumonia, and estimated the attributable hospital costs by using the effect estimates obtained from both functions.

Methods: We used hospital discharge data on 3519 non-elective pneumonia admissions from UZ Brussels between 2007 and 2012 and we combined a case-crossover design with distributed lag models. The annual averted pneumonia hospitalization costs for a reduction in PM2.5 exposure from the mean (21.4 mu g/m(3)) to the WHO guideline for annual mean PM2.5 (10 mu g/m(3)) were estimated and extrapolated for Belgium.

Results: Non-elective hospitalizations for pneumonia were significantly associated with PM2.5 exposure in both models. Using a log-linear E-R function, the estimated risk reduction for pneumonia hospitalization associated with a decrease in mean PM2.5 exposure to 10 mu g/m(3) was 4.9%. The corresponding estimate for the log-log model was 10.7%. These estimates translate to an annual pneumonia hospital cost saving in Belgium of (sic)15.5 million and almost (sic)34 million for the log-linear and log-log E-R function, respectively.

Discussion: Although further research is required to assess the shape of the association between PM2.5 exposure. and pneumonia hospitalizations, we demonstrated that estimates for health effects and associated costs heavily depend on the assumed E-R function. These results are important for policy making, as supra-linear E-R associations imply that significant health benefits may still be obtained from additional pollution control measures in areas where PM levels have already been reduced.