Variation in Practice for Discretionary Admissions

Standardized hospital mortality ratios based on analyses that do not account for variation in admission practice among hospitals are biased. Variation in admission practice will cause any outcome measure based solely on hospitalized patients to be similarly biased. Correction for selection bias is required to produce valid measures of hospital quality.