SHM Converge Scientific Abstract ePoster Gallery-A Machine Learning Algorithm Predicts Post Acute Care Needs to Enhance Discharge Planning

Pdf Summary

This study evaluated the Johns Hopkins Early Discharge Planning Calculator (JH-EDPC), a machine-learning tool adapted into a clinical app to predict whether hospitalized patients would be discharged home or to post-acute care (PAC). The goal was twofold: first, to see whether making the prediction visible to hospitalists increased early involvement of rehab services; second, to test how accurately the tool predicted discharge location in real time.<br /><br />The study included 162 adult hospitalist patients hospitalized for at least 48 hours. On day 2 of admission, staff entered patient data such as age, living situation, surgical status, and AM-PAC mobility scores into the app. In the intervention group, hospitalists could see the prediction; in the control group, they could not. Researchers tracked physical and occupational therapy use, time to consult, and actual discharge outcomes.<br /><br />Results showed that awareness of the app’s prediction increased rehab involvement. Within 5 days of admission, about 35% of intervention patients received a physical therapy consult versus about 25% in the control group. Among patients predicted to need PAC, consult rates were especially higher in the intervention group (about 80% vs. 50%).<br /><br />The calculator also performed well in predicting discharge disposition, with 85% accuracy, ROC 0.77, sensitivity 68%, and specificity 87%.<br /><br />Overall, the JH-EDPC appears to be a useful early discharge planning tool that can improve triage of limited rehab resources and encourage earlier, more appropriate therapy involvement.

Asset Subtitle

Becca C. Engels

Meta Tag

Author List Abigail Heubeck, Becca C. Engels, Daniel Young, Erik Hoyer

Category Research

Concept Early Discharge Planning

Concept Johns Hopkins Early Discharge Planning Calculator

Concept Non-medical Barrier

Concept Delayed Hospital Discharge

Concept Post-Acute Care

Distinguished Non-Finalist

Presenter Organization Johns Hopkins Hospital

Presenting Author Becca C. Engels

Track Transitions of Care

Keywords

Johns Hopkins Early Discharge Planning Calculator

JH-EDPC

machine learning

discharge prediction

post-acute care

rehabilitation services

hospitalist

physical therapy consult

occupational therapy

discharge disposition

Early Discharge Planning

Johns Hopkins Early Discharge Planning Calculator

Non-medical Barrier

Delayed Hospital Discharge

Post-Acute Care