SHM Converge Scientific Abstract ePoster Gallery-Comparing Methods for Attributing Hospital Quality Metrics to Individual Hospitalists: A Retrospective Cohort Study

Pdf Summary

This study compared five methods for attributing hospital quality metrics to individual hospitalists across five Utah hospitals from 2020–2024, using physician-attributed length of stay (LOS) as the benchmark outcome. The attribution methods were Majority, Plurality, Standard/discharging hospitalist, Partial/day-weighted, and IMPAQ, which assigns the discharging hospitalist only if they oversaw at least 30% of the encounter.<br /><br />The authors evaluated agreement between methods in two ways: whether patient encounters were included or excluded, and how physicians were ranked into performance quartiles within each hospital site group. Agreement was measured using kappa coefficients.<br /><br />Among 95,112 eligible hospital encounters, the Majority and Plurality methods behaved similarly and mostly excluded short LOS encounters while retaining longer stays, showing near-perfect agreement in encounter inclusion. In contrast, the other methods had only moderate to no agreement with one another, with kappa values ranging from 0.54 to -0.10. IMPAQ tended to keep short LOS encounters and exclude the longest ones.<br /><br />Physician performance rankings varied substantially by attribution method. Agreement in quartile rank ranged from none (κ=0.11 for Partial vs Majority/Plurality) to moderate (κ=0.76 for Majority vs Plurality). Individual hospitalists could shift by as much as five standard deviations in rank depending on the model used. The Partial attribution model also showed a strong negative correlation between z-scores and the percentage of admitting shifts (r=-0.70, p<0.001).<br /><br />Overall, the study found that common attribution methods do not agree on which encounters to assign to hospitalists and can produce very different quality rankings. This has important implications for pay-for-performance programs and quality improvement efforts.

Asset Subtitle

Harris L. Carmichael

Meta Tag

Author List Harris L. Carmichael, Ithan Peltan, Michael Pirozzi, Scott Woller

Category Research

Concept Hospitalist Care Attribution

Concept Length of Stay

Concept Majority Method

Concept Plurality Method

Concept Clinical Encounter

Distinguished Finalist

Presenter Organization Intermountain Healthcare

Presenting Author Harris L. Carmichael

Track Outcomes Research

Keywords

hospitalist attribution

hospital quality metrics

length of stay

attribution methods

Majority method

Plurality method

IMPAQ

kappa agreement

physician performance ranking

pay-for-performance

Hospitalist Care Attribution

Length of Stay

Majority Method

Plurality Method

Clinical Encounter