An event count model for studying health services utilization.

OBJECTIVES: The purpose of this article is to first conceptualize health services utilization behavior as event counts. Based on this concept and behavioral theory, the author presents the Generalized Event Count model as an alternative modeling tool for studying health services utilization. This model is theory driven and is consistent with behavioral assumptions.
METHODS: In presenting the Generalized Event Count model, the author first examines its model assumptions to see whether they conform to elements of behavior theory, assumptions of health services utilization, and the distribution assumption of the nature of the data. To demonstrate the Generalized Event Count model, the author applied this model to an empirical ambulatory care utilization data set from a 1988 household interview of Chinese-Americans in Boston's inner-city community.
RESULTS: The Generalized Event Count model analysis suggested that the regular source of medical care and the use of Chinese medicine were strong predictors of physician visits for this population. Further, the Generalized Event Count model was able to test that most of the ambulatory visits within an individual were correlated.
CONCLUSIONS: Compared with other models, the Generalized Event Count model is more consistent with health services utilization behavioral assumptions. Moreover, it makes an efficient use of information from the utilization data for model estimation. This model has the potential of having broad applications in studying various types of health services utilization, especially for analyzing cross-sectional utilization data.