Abstract

The Second National Morbidity Survey, conducted in England and Wales between 1970 and 1976, contains a unique body of information on episodes of mental illness experienced by individuals registered in a representative sample of general practices around the country. This information is used to construct the episode distribution among the individuals surveyed. The Poisson and negative binomial distributions are then used to model the episodes. The Poisson model gives a very poor fit but the negative binomial model is found to fit the data very well. Deviations of the observed data from this model are discussed. The possibility of applying this model at the local practice level is then considered.