Modeling denominator populations for mental health care.

Difficulties in using basic epidemiologic survey techniques in estimating the prevalence of disease in the community have led health services researchers to develop alternative methods using routinely collected health care data. The purpose of this paper is to illustrate the application of a model that estimates the prevalence of diagnosable mental disorder in a community, including the proportion of untreated cases. A mathematical model is developed based on a well-known probability distribution (the negative binomial) and simple calculation formulae are presented. This model has been used to estimate total practice population size. This paper extends that work by illustrating the utility of the model vis-a vis a specific type of disorder and shows pitfalls when caseness is not well established. An application of the model to data from two studies in an organized health care setting shows that the estimated period prevalence (one year) of mental disorder is between 17.4% and 21.8%. Of those individuals, it is estimated that 28.6% seek no health care during the year. The implications of these findings and the utility of the model are discussed.