Comparing Methods for Attributing Hospital Quality Metrics to Individual Physicians: Impact on Performance Assessment and Outlier Identification.

backgroundPhysician-attributed quality measurement underpins value-based healthcare and commonly informs physician reimbursement. However, accurate quality measurement for individual physicians is increasingly difficult with the rise of team-based care delivery and without gold standards defining accuracy in physician-attributed quality.

objectiveTo compare how different attribution methods affect which patient encounters are included in physician performance metrics and, subsequently, impact physician-attributed quality rankings and outlier identification.

designRetrospective comparative analysis of hospitalist encounters in a multi-hospital health system from January 2020 to September 2024.

participants95,112 eligible hospital encounters overseen by 135 physician/hospital pairs across four Utah hospitals.

interventionsFive attribution methods were compared: Standard (discharging physician), Intermountain Method for Physician Attributed Quality (IMPAQ, discharging physician with ≥ 30% encounter oversight), Partial (proportional day-weighted), Majority (physician overseeing ≥ 50% of encounter), and Plurality (physician overseeing largest proportion of encounter). MAIN MEASURES: Primary outcomes included agreement in encounter inclusion/exclusion between methods, physician quartile rankings, and identification of outlier performance using z-scores for physician-attributed length of stay (LOS). KEY

resultsThe proportion of encounters attributed to individual physicians ranged from 78% to 100% by method. Majority and Plurality methods showed near perfect agreement in encounter selection (κ = 0.91) and physician ranking but primarily excluded short LOS encounters while retaining prolonged encounters assigned to individual physicians. IMPAQ and Standard methods showed only moderate agreement (κ = 0.54-0.62). The IMPAQ method retained short LOS encounters and excluded the longest LOS encounters while identifying the most extreme physician outliers. Individual physicians changed positions as much as 5 standard deviations between attribution methods.

conclusionsAttribution method choice dramatically influences encounter inclusion/exclusion and hospitalist performance assessment. Methods retaining encounters without meaningful physician oversight may introduce noise and obscure observation of true physician differences. These findings highlight the urgent need for consensus on attribution methodologies to ensure valid performance assessment in the context of team-based care.