RESUMO
OBJECTIVES: To develop two practice-site-level measures of comprehensiveness and examine their associations with patient outcomes, and how their performance differs from physician-level measures. DATA SOURCES: Medicare fee-for-service claims. STUDY DESIGN: We calculated practice-site-level comprehensiveness measures (new problem management and involvement in patient conditions) across 5286 primary care physicians (PCPs) at 1339 practices in the Comprehensive Primary Care initiative evaluation in 2013. We assessed their associations with practices' attributed beneficiaries' 2014 total Medicare expenditures, hospitalization rates, ED visit rates. We also examined variation in PCPs' comprehensiveness across PCPs within practices versus between primary care practices. Finally, we compared associations of practice-site and PCP-level measures with outcomes. PRINCIPAL FINDINGS: The measures had good variation across primary care practices, strong validity, and high reliability. Receiving primary care from a practice at the 75th versus 25th percentile on the involvement in patient conditions measure was associated with $21.93 (2.8%) lower total Medicare expenditures per beneficiary per month (P < .01). Receiving primary care from a practice at the 75th versus 25th percentile on the new problem management measure was associated with $14.77 (1.9%) lower total Medicare expenditures per beneficiary per month (P < .05); 8.84 (3.0%) fewer hospitalizations (P < .001), and 21.27 (3.1%) fewer ED visits per thousand beneficiaries per year (P < .01). PCP comprehensiveness varied more within than between practices. CONCLUSIONS: More comprehensive primary care practices had lower Medicare FFS expenditures, hospitalization, and ED visit rates. Both PCP and practice-site level comprehensiveness measures had strong construct and predictive validity; PCP-level measures were more precise.