Very little has been done to model the utilization of ambulatory care empirically. OBJECTIVE: The purpose of this study is to estimate the expected frequency of outpatient visits from the socio-demographic characteristics, health status and comorbidities. METHOD: A randomized sample of 6,000 Southern California patients with chronic diseases in a managed care environment were surveyed longitudinally during 1992-95. Simultaneously, health care utilization data were collected from electronic data files. An expected frequency count data model was developed by poisson regression. The frequency of outpatient visits from baseline was used as a lagged dependent variable in the equation. Socio-demographic variables were used as covariates. Model overdispersion was corrected by appropriate power transformation. Model validity was also compared with the ordinary least square model and the general linear model. RESULTS: More than 75% of the sample had five or more outpatient visits during demonstration period with a mean of 15 visits per patient. The model showed that age, gender and baseline visit were significant predictors (p<0.0001) of future outpatient visit. Patient’s chronic disease status along with three of the eight SF-36 health status domains measured at baseline (bodily pain, general health, role limitation due to physical problem) were also statistically significant (p<0.005) in explaining the variations in future outpatient visits. Asian patients were less likely to use ambulatory care facilities than other races and female patients experienced greater utilization than males. Income and employment status significantly affected outpatient utilization. The count data model was superior to the other models. CONCLUSION: This study provided a useful alternative empirical method to model count data in ambulatory care.