SHM Converge Scientific Abstract ePoster Gallery-Provider-Level Length of Stay: Multivaritate Model vs Standard Metrics

Pdf Summary

This study examined whether individual hospitalists meaningfully affect patient length of stay (LOS), a key measure of hospital efficiency. Researchers analyzed 13,442 inpatient encounters (70,262 patient days) from a large academic tertiary referral center over two years. They collected patient demographics and many known LOS drivers, including discharge day, unit, service, DRG, expected LOS, ICU stay, month, and discharge disposition.<br /><br />To compare approaches, the team used several common provider-level LOS attribution methods, such as observed-to-expected LOS based on discharge provider, progress notes, and discharge efficiency. They then built a mixed-effects negative binomial regression model that adjusted for the many patient- and institution-level factors that also influence LOS.<br /><br />Thirty-eight providers with at least 75 discharges were included. The analysis showed that provider rankings changed substantially depending on the metric used, but once other major contributors were accounted for, the effect of any single provider on LOS was generally small. The model found that variables such as discharge disposition, DRG, ICU status, and patient demographics had much stronger associations with LOS than individual providers did.<br /><br />The main conclusion was that in a large academic hospital medicine group covering multiple service lines, individual providers are unlikely to be major drivers of LOS. The authors suggest that LOS improvement efforts should focus more on broader system factors and high-impact clinical and discharge-related variables than on provider-level comparisons alone.

Asset Subtitle

Matthew T. Cerasale

Meta Tag

Author List Andrew W. Schram, Eliot R. Weinstein, Matthew T. Cerasale

Category Research

Concept Length of Stay

Concept Provider-Level Attribution

Concept Discharge Disposition

Concept Multivariate Mixed-Effects Model

Concept Physician Impact

Distinguished Non-Finalist

Presenter Organization University of Chicago

Presenting Author Matthew T. Cerasale

Track Value in Hospital Medicine

Keywords

hospitalist

length of stay

LOS

inpatient encounters

hospital efficiency

mixed-effects regression

negative binomial model

discharge disposition

ICU stay

provider attribution

Length of Stay

Provider-Level Attribution

Discharge Disposition

Multivariate Mixed-Effects Model

Physician Impact