Dr. Patrick Brown
Cancer Care Ontario and University of Toronto
Title: " Bayesian Statistics and Swine Flu: Estimating Hospitalization and
Fatality Rates "
In the initial stages of an epidemic, estimating the fraction of cases
which result in hospitalization or death is complicated by the fact that
many (or most) of the cases are recent and sufficient time has not elapsed
for hospitalization or death to occur. Further, the number of deaths
recorded can be small, a particular problem when rates by age group or for
individuals with and without previous medical conditions are required. A
simple multi-state model is presented which allows for uncertainty about
the future progression of known cases, as well as for changes in rates with
age. To overcome the problem of limited data, Bayesian prior distributions
are assigned to the unknown parameters using expert opinion. One result is the
ability to translate possible scenarios of new case numbers into predictions
of the future load on hospital beds.