USING SEVERITY-ADJUSTED STROKE MORTALITY-RATES TO JUDGE HOSPITALS

Mortality rates are commonly used to judge hospital performance. In co mparing death rates across hospitals, it is important to control for d ifferences in patient severity. Various severity tools are now activel y marketed in the United States. This study asked whether one would id entify different hospitals as having higher- or lower-than-expected de ath rates using different severity measures. We applied 11 widely-used severity measures to the same database containing 9407 medically-trea ted stroke patients from 94 hospitals, with 916 (9.7%) in-hospital dea ths. Unadjusted hospital mortality rates ranged from 0 to 24.4%, For 2 7 hospitals, observed mortality rates differed significantly from expe cted rates when judged by one or more, but not all 11, severity method s. The agreement between pairs of severity methods for identifying the worst 10% or best 50% of hospitals was fair to good. Efforts to evalu ate hospital performance based on severity-adjusted, in-hospital death rates for stroke patients are likely to be sensitive to how severity is measured.