Evaluating risk adjustment by partitioning variation in hospital mortality rates

The variation in mortality rates among hospitals has often been described informally as having three major components: patient severity, quality of care and random variation. These informal concepts are characterized formally by partitioning sums of squares and finding their expected values. The partition relates to commonly used tests for whether individual hospitals have unusual mortality rates. Application of the partition to the hospital mortality reports by the Health Care Financing Administration shows that their models for patient risk account for about one‐half the variation among hospital mortality rates. An example using clinical measures of severity accounts for about two‐thirds of mortality variation among hospitals.