*163. Validation of Case-mix Measures Derived from Self-Reports of Diagnoses and Health
VS Fan, University of Washington; DH Au, VA Puget Sound Health Care System; P Heagerty, University of Washington; RA Deyo, University of Washington; MB McDonell, VA Puget Sound Health Care System; SD Fihn, VA Puget Sound Health Care System
Objectives: Self-reported chronic diseases and health status have been associated with use of health services. There is, however, little information available about whether these measures are related to the risk of mortality or hospitalization. We sought to determine whether self-reported chronic medical conditions and the SF-36 could be used individually or in combination to develop a comorbidity index that could be applied in the outpatient setting to predict mortality and hospitalizations.
Methods: We used data from the Ambulatory Care Quality Improvement Project (ACQUIP) to conduct a prospective cohort study of patients enrolled in the primary care clinics at 7 Veterans Affairs medical centers. Our primary outcome was all-cause mortality and first hospitalization. Of the 34,103 subjects who were eligible for ACQUIP, 12,388 subjects returned both a health inventory and the SF-36 at entry to the ACQUIP study and were eligible for this analysis. The health inventory asked whether patients had been told by a clinical provider that they had any of 25 common chronic illnesses. We used Cox-modeling to estimate the hazard ratio for mortality and for first hospitalization. Because patients below 50 years of age violated the proportional hazards assumptions, the analysis was restricted to the 10,947 patients over 50 years of age. These patients were followed for a mean of 722.5 (plus or minus 84.3) days.
Results: Using a Cox proportional hazards model in a development set of 5469 patients, a comorbidity index was constructed from age, smoking status and 7 of 25 self-reported medical conditions that were associated with increased mortality univariately. The comorbidity index was predictive of both mortality and hospitalizations when validated in the validation set of 5478 patients. Two-year Kaplan Meier estimated mortality for different scores was: comorbidity score of "0-3," 1.3%, "4-5," 3.4%, "6-7," 7.6%, ">= 8," 15.2% (p < 0.00005 for difference in the survival curves). A separate model was constructed in which only age and the PCS and MCS scores of the SF-36 were entered to predict mortality. Both the SF-36 component scores and the comorbidity score had comparable ROC curves (AUC=0.70). However, combining the comorbidity score and SF-36 modestly improved discrimination for all cause mortality (AUC=0.74).
Conclusions: A new outpatient comorbidity score developed using an easy self-administered, health inventory was predictive of 2-year mortality and hospitalizations in the primary care setting. Adding the PCS and MCS scales from the SF-36 significantly increased the discriminatory ability of the model in predicting both mortality and hospitalizations.
Impact: Self-reported disease conditions and the SF-36 can be used to assess an ambulatory care population's risk of subsequent mortality and hospitalizations.