Using Medical Malpractice Data to Predict the Frequency of Claims: A Study of Poisson Process Models with Random Effects

I use the Florida state data base (1975–1987) of settled malpractice claims to develop Poisson process models for the frequency of claims filed against individual physicians. These models incorporate random effects and covariates that represent physician attributes and are natural generalizations of the negative binomial model that is typically used to study claims frequency. I predict claims frequencies during 1981–1982 using models that are selected and estimated from claims data during 1975–1980 and then compare these predictions to the actual frequencies.