Bayesian Analysis of the 1918 Influenza Pandemic in Baltimore, MD and Newark, NJ

Thomas A. Louis, Ph.D.
Johns Hopkins Bloomberg
School of Public Health

Wednesday, April 25, 2007
*3:45pm
NHH 2-101
Minneapolis Campus

*Please note the change in time and location

Abstract:
The influenza pandemic of 1918 killed from 20 to 40 million people worldwide. It has been well studied, but several features remain unexplained. For example, a mechanism for multiple temporal waves in incidence has not been identified. To get additional insight, we implement a Bayesian analysis of a stochastic Susceptible-Infected-Recovered (S-I-R) model using fall 1918 daily disease incidence and influenza related death data from Baltimore, MD and Newark, NJ. We find a time-varying influenza transmissibility {(social contact rate) x (infectivity)} and consequently a time-varying reproductive number (the expected number of individuals infected by an infective). In Baltimore, transmissibility increases then decreases; in Newark it is strictly decreasing. The S-I-R model fits the Baltimore data poorly, but fits the Newark data well.

We report these results, display fit diagnostics and sensitivity analysis; and speculate on what produces the poor fit in Baltimore and the need for time-varying transmissibility. Candidate causes include inhomogeneous spatial spread, aggregation of S-I-R compatible sub-epidemics and a time-varying mixing rate (due to changes in social dynamics). Furthermore, conclude that the multiple mortality waves in time series from the US are consistent with estimated time-varying transmissibility.
A social tea will be held at 3:00 P.M. in A434 Mayo. All are Welcome.
For more details contact 612-624-4655 or see http://www.biostat.umn.edu/seminar_academic.html