Estimating mortality displacement during and after heat waves

Abstract

The proportion of excess deaths occurring as a result of hot weather that are brought forward by only a short time ("displaced") is important but not easy to estimate. A recent proposal by Saha et al. (Am J Epidemiol. 2014;179(4):467-474) was to estimate this using a "displacement ratio" equal to the sum of deficits of daily deaths below an expected baseline divided by the sum of excesses over all days during and up to 15 days after a heat wave. Unfortunately, this method results in important artifacts due to natural Poisson variation in deaths by which deficits, and hence displacement ratios above zero, will occur even when there is no real short-term displacement. Simulations confirm this and further show spurious patterns, such as the displacement ratio diminishing with more severe waves. This displacement ratio cannot be relied upon for interpretation. Quantifying mortality displacement remains an incompletely resolved problem.