Objectives Performance measurement is used by health care providers, payers, and patients. Historically accomplished using administrative data, registries are used increasingly to track and improve care. We assess how measured diabetes care quality differs when calculated using claims versus registry. Data Sources/Study Setting Cross‐sectional analysis of administrative claims and electronic health records (EHRs) of patients in a multispecialty integrated health system in 2012 (n = 368,883). Study Design We calculated percent of patients attaining glycohemoglobin <8.0 percent, LDL cholesterol <100 mg/dL, blood pressure <140/90 mmHg, and nonsmoking (D4) in cohorts, identified by Medicare Accountable Care Organization/Minnesota Community Measures (ACO‐MNCM; claims‐based), Healthcare Effectiveness Data and Information Set (HEDIS; claims‐based), and registry (EHR‐based). Data Collection/Extraction Methods Claims were linked to EHR to create a dataset of performance‐eligible patients. Principal Findings ACO‐MNCM, HEDIS, and registry identified 6,475, 6,989, and 6,425 measurement‐eligible patients. Half were common among the methods; discrepancies were due to attribution, age restriction, and encounter requirements. D4 attainment was lower in ACO‐MNCM (36.09 percent) and HEDIS (37.51 percent) compared to registry (43.74 percent) cohorts. Conclusions Registry‐ and claims‐based performance measurement methods identify different patients, resulting in different rates of quality metric attainment with implications for innovative population health management.