OBJECTIVES: National hospital performance evaluation programs depend on risk adjustment using comorbidities identified in administrative claims, yet the degree to which these data are comparable to clinical registry data is unknown. Our objective was to compare capture of comorbidities and hospital outlier classification based on risk-adjusted rates of readmission and mortality estimated from clinical registry and administrative claims data.  METHODS: We linked detailed clinical data for hospitalized NSTEMI patients >65 years of age in the CRUSADE study (Can Rapid Risk Stratification of Unstable Angina Patients Suppress ADverse Outcomes with Early Implementation of the ACC/AHA Guidelines; 2002–2007) to Medicare inpatient and outpatient claims to obtain longitudinal outcomes. We created risk adjustment models using 8 comorbidities captured in the CRUSADE registry and as International Classification of Diseases, 9^th edition (ICD-9) codes from claims data in the prior year for: myocardial infarction (MI) congestive heart failure (CHF) , diabetes, hypertension, renal insufficiency, stroke, percutaneous coronary intervention, coronary artery bypass grafting. We identified outliers using observed to expected ratios (O/E) for 30-day readmission and mortality from logistic generalized estimating equation models. RESULTS: Of 68,199 patients in our study cohort, 48.1% were female, 86.9% were white, and the median age was 78.0 (interquartile range=72.0 – 83.0). Comorbidity percent agreement between data sources ranged from 67.8% for MI to 89.3% for diabetes. Despite differences in individual model coefficients for comorbidities in claims-based and registry-based models, overall model performance was similar (mortality c-statistics= 0.69-0.71; readmission c-statistics=0.59). In mortality models, 82 hospitals were identified as outliers using claims-based models; of these, n=70 (85.4%) were identified as outliers in registry-based models.  Similar patterns were observed in readmission models. CONCLUSIONS: Comorbidity ascertainment may differ in clinical registries versus administrative claims. While overall model performance is comparable, individual hospital outlier status may differ based on the comorbidity data source used.