LAUSR

BACKGROUND Interpretation of hospital quality requires objective evaluation of both inpatient and postdischarge adverse outcomes (AOs). OBJECTIVE To develop risk-adjusted predictive models for inpatient and 90-d postdischarge AOs in elective craniotomy and apply those models to individual hospital performance to provide benchmarks to improve care. METHODS The Medicare Limited Dataset (2012-2014) was used to define all elective craniotomy procedures for mass lesions in patients ≥65 yr. Predictive logistic models were designed for inpatient mortality, inpatient prolonged length of stay, 90-d postdischarge deaths without readmission, and 90-d readmissions after exclusions. The total observed patients with one or more AOs were then compared to predicted AO values, and z-scores were computed for each hospital that met minimum volume requirements. Risk-adjusted AO rates allowed stratification of eligible hospitals into deciles of performance. RESULTS The hospital evaluation was performed for 223 facilities with 7624 patients that met criteria. A total of 849 patients (11.1%) died inclusive of 90 d postdischarge; 635 (8.3%) were 3σ length-of-stay outliers; and 1928 patients (25.3%) with one or more 90-d readmissions; 2716 patients experienced one or more AOs (35.6%). Six hospitals were 2 z-scores better than average, and 8 were 2 z-scores poorer. The median risk-adjusted AO rate was 18% for the first decile and 53.4% for the 10th decile. CONCLUSION There was a 35% difference between best and suboptimal performing hospitals for this operation. Hospitals must know their risk-adjusted AO rates and benchmark their results to inform processes of care redesign.