A Statistical Model for Assessing the Need for Medical Care in a Health Screening Program

This is a statistical model (the "multiple logistic model") for pooling data from many variables into a single probability estimate that a person examined in a health screening system requires medical care. Because discrete variables are frequently studied in health screening programs, the required assumption of multivariate normality of the predictor variables related to many statistical techniques is seldom fulfilled. A computing method is given that avoids this assumption by directly assuming the appropriateness of logistic function and obtains the corresponding maximum likelihood estimates (MLE) of the unknown parameters. Fisher’s linear discriminant function (LDF) may be used to provide the initial estimates of these parameters, which are necessary for the application of the MLE computing procedure. The algorithms to estimate these coefficients are discussed for both the LDF and MLE models, and an application of these methods to a set of data is presented and other applications of this procedure are proposed