| To identify the important risk factors that contribute to the skewed health care utilization rates among patients, where the top percent of patients consume the majority of medical services, we recruited 334 study participants from Yale New Haven Hospital (YNHH) Primary Care Center. Those patients were chosen based on their health care utilization one year prior to the study so that roughly 50% of these patients will present top 20% costly of all YNHH patients. These patients were interviewed three times over the course of a year to get measurements on various instruments such as demographic characteristics, health status, mental disorder diagnosis and psychosocial measurements. In our analysis, study participants were stratified into four utilization categories base on their utilization data during the year of study: Low (below 50% in total cost); middle (50% to 75% in total costs); high (75% to 95% in total costs); and super high (highest 5% in total cost). Difference in demographic characteristics and instrument measurements between utilization categories were compared using analysis of variance (ANOVA), chi-square tests, or fisher's exact tests. Multivariate regression analysis and recursive partitioning tree analysis were performed to identify the important predictors of health care utilization. Results show that there may be a gender linked increased risk for high utilization to a degree. Age, employment status, who do you live with, Interpersonal Support Evaluation List (ISEL) score and physical health score (SF36) differ significantly between different utilization groups. Results from multivariate regression analysis indicate that among all variables, physical health score (SF36) is the most important predictor for health care utilization, followed by SCID diagnosis. Whereas recursive partitioning tree analysis shows that among patients who have lower physical health scores, those who have a current SCID diagnosis have higher average health care cost compared to patients who do not have a current SCID diagnosis, which may be explained by that primary care patients with psychiatric disorders often over present with physical impairments than those associated with medical diagnosis. |
