Alternative Methods for Determining Resource Requirements: The Chile Example

Abstract

Many more or less attractive techniques have been proposed for statistical analyses involving multiple sources of variation, for example in examining the possible contributors to differential patterns of health services utilization and expenditure. A large scale investigation of such patterns among 10,000 households in Chile provided a useful basis for comparison of alternative analytical approaches. The multiple regression, Automatic Interaction Detection, and Multisort techniques were applied to the survey data separately and in combination and results were compared. The Multisort technique was found to be the single most useful and convenient approach, but the most meaningful results were obtained from the systematic application of the three approaches combined.