Abstract 32: Variation in Discharge Practices and Use of Life-Sustaining Procedures: Potential Pitfalls for Adjusted Mortality Quality Measures

Abstract

Introduction: The Center for Medicare and Medicaid Services (CMS) is developing a risk-standardized stroke mortality measure intended for public comparison of hospital-level quality of care. We sought to determine the impact of discharge practices and use of life sustaining procedures on risk-standardized mortality. We could then address questions such as, how much does mortality differ between hospitals that routinely discharge patients to hospice and those that don’t? Methods: We estimated risk standardized stroke inpatient mortality for all hospitals in the Nationwide Inpatient Sample (NIS) from 2008-9, using hierarchical logistic regression following a similar approach used for existing CMS mortality measures. Hospital-level adjusted stoke mortality was then categorized as below average, average or above average using standardized Hospital Compare methodology. We then developed individual and hospital level variables to characterize hospital-level practices of interest: discharge to long term acute care (LTAC), discharge to hospice and a life-sustaining procedure index which combined the rates of gastrostomy, tracheostomy, hemicraniectomy and ventriculostomy. These variables were added to the base model in a second hierarchical logistic regression model. This model was used to compare differences in risk-standardized mortality from hospitals grouped in the highest and lowest practice quintiles and to estimate the effects of the hospital practices on grades. Results: A total of 186,689 stroke patients were identified. Median age was 72 (IQR 60-82) and 52% were female. Seventy six percent of strokes were ischemic, 19% were ICH and 5% were SAH. Of the 1,366 hospitals in the sample 1,210 received average, 73 below average and 83 above average grades for stroke mortality using the base model. Hospitals that more commonly discharged patients to hospice or LTAC or used more life-sustaining procedures had lower mortality. Mean risk standardized mortality was higher in the lowest quintile of hospice utilizing hospitals compared to the highest (11.4% vs. 10.0%, p < 0.01) and in the lowest quintile of LTAC utilizing hospitals compared to the highest (11.4% vs. 10.3%, p = 0.02). Similarly, mean risk standardized mortality was higher in the lowest quintile of hospital use of life-sustaining procedures compared to the highest (11.3% vs. 10.7%, p < 0.01). Accounting for changes in hospital-level practices resulted in significant hospital grade reassignment. Of the 73 initial below average grades, 19 (26%)were reclassified to average and of the 83 initial above average grades, 35 (42%) were reclassified to average. Overall, 6% were reclassified. Conclusions: Variation in discharge practices and use of life-sustaining procedures alters hospital mortality rankings and distorts perceptions of comparative hospital quality. ns of comparative hospital quality.