|
Abstract:
|
When considering influenza surveillance data, a primary measure of epidemic severity at the population level is excess mortality. Serfling's method to estimate excess mortality of influenza is a two-step process: (1) establish seasonal baseline by fitting a periodic regression model to a time series of recent mortality data; and (2) for future observations, calculate aggregated positive deviation above predicted baseline. Although excess mortality is a primary outcome in many epidemiological studies of influenza data, the variability of its estimator has rarely been studied, discussed, or accounted for analytically. We use simulations to quantify the variability of Serfling's estimator, and compare our results to theoretically derived values. We then apply these methods to historical influenza mortality data from the United States.
|
