2070. Development of a Casemix Adjustment Model for VA CHF Patients
UshaSubramanian, MD, Roudebush VAMC and Indiana University School of Medicine, F Hopp, VA Ann Arbor Healthcare System, M Kim,
VA Ann Arbor Healthcare System and University of Michigan Medical School, K Austin,
VA Ann Arbor Healthcare System and University of Michigan Medical School, J Lowery,
VA Ann Arbor Healthcare System and University of Michigan Medical School
Objectives: (1) To evaluate the effectiveness of the Charlson (generic) and Polanczyk (CHF-specific) case mix adjustment models for predicting mortality in CHF patients; and
(2) To determine whether clinical variables (alcohol dependence and depression), as well as laboratory and medication variables, can improve on the predictive powers of existing models.
Methods: The study cohort included CHF patients (one inpatient or 2+ outpatient visits) at six VISN 11 hospitals in FY1999. C-statistics were calculated from 2-year mortality models. The models included age, gender, hospitalization, generic and CHF-specific co-morbidities, clinical variables, laboratory values, and medication variables.
Results: Two-year mortality was 19.94%. The C-statistic for the CHF-specific model (0.70) was only slightly higher than for the generic model (c=0.66). The final model (with both generic and CHF-related co-morbidies, clinical variables, and laboratory and medication data) provided only marginal predictive improvement (c=0.74). Significant (P<0.05) predictors of mortality in the final model included having had a CHF-related stay within one year of the index date (OR=1.44) and age (OR=1.52).
Conclusions: The predictive powers of the co-morbidity models were not high, and were not substantially improved by inclusion of additional clinical, laboratory, and medication values.
Impact: While it is tempting to use existing co-morbidity models for casemix adjustment, the results of this study suggest that additional work is needed to develop VA-specific case mix adjustment models for CHF patients.