The goal of the proposed project is to compare survey-based with claims- based risk assessment in a primary care ambulatory adult patient population by predicting charges for health services during a one-year period following a baseline clinic visit. The survey-based tool to be used is the Duke Case-mix System (DUMIX), which is based upon survey data on patient-reported functional health status (i.e., health-related quality of life) and provider-reported severity of illness at the time of one clinic visit. The DUMIX will be compared with the Ambulatory Care Groups (ACG), a claims-based tool which is based upon data on cumulative patient diagnoses for all clinic visits during a one-year period.
The specific aims are to compare the predictive accuracy of the two tools during the same one-year period on one group of 956 patients, and then compare the predictive accuracy of the two tools for forecasting future one-year changes of a second group of 1085 patients when using predictive coefficients derived from the first group of patients. Patients will be randomly selected from the Duke Family Medicine Center, a large primary care facility. For used in the DUMIX, functional health and severity of illness survey data will be collected during one clinic visit. For use in the ACG, diagnosis and visit data for each participant will be downloaded from the clinic computer for the one-year period after the baseline clinic visit. Total health services charges for the one-year period will be used as the outcome variable in the statistical analyses. Statistical models will include ordinary linear regression, mean absolute error, and Cox proportional hazards to allow for the known skewness of health care charges. It is hypothesized that both the survey and the claims tools will be statistically significant predictors of health service charges, and that there will be no difference in the predictive accuracy of the two tools. If these hypotheses are supported by the study, the two methods of risk assessment will offer complementary alternatives for used under different circumstances. When computerized diagnostic datasets are not available for use of the ACG, the DUMIX will offer a valid alternative that will have the additional advantages of direct input from both patients and providers and lack of dependence on poor diagnostic and/or utilization data.
