SHM Converge Scientific Abstract ePoster Gallery-Mobility Scores Help Predict Early Discharge Disposition in Hospitalized Patients

Pdf Summary

This study examined whether a patient’s early hospital mobility score can predict discharge destination. Researchers retrospectively analyzed records from Johns Hopkins Hospital for 34,432 admissions, comparing discharge home versus discharge to a post-acute care (PAC) facility. The key predictor was the lowest AM-PAC “6-clicks” mobility score recorded within the first 48 hours of admission.<br /><br />Using a random forest machine learning model, the team found strong predictive performance, with an AUC of 0.80, indicating good ability to distinguish patients likely to go home from those likely to need PAC. AM-PAC was the most important predictor in the model, ahead of BMI, age, and admitting service. Other factors such as prior function, payor, gender, living alone, ICU/mechanical ventilation status, and prior residence were less influential.<br /><br />Patients with lower AM-PAC scores were much more likely to require PAC. For example, scores below 38 were associated with at least a 0.26 probability of PAC discharge, and patients age 66 or older with scores below 31 had the highest likelihood. A decision threshold of 0.25 yielded 78% sensitivity and 66% specificity, while a threshold of 0.40 increased specificity and overall accuracy but reduced sensitivity.<br /><br />The findings were consistent across most medical services, suggesting the mobility measure is robust across diagnoses and care settings. The study concludes that baseline mobility assessment early in hospitalization is a highly useful tool for discharge planning and should be systematically incorporated into clinical workflows and electronic medical records to help identify patients at risk for needing post-acute care.

Asset Subtitle

Becca C. Engels

Meta Tag

Author List Becca C. Engels, Daniel Young, Elizabeth Colantuoni, Erik Hoyer, Lisa Friedman

Category Research

Concept Acute Medical Profile-Pathway with Home Care AM-PAC Mobility Score

Concept Hospital Discharge Disposition

Concept Post-Acute Care

Concept Home Discharge

Concept Predictive Model

Distinguished Non-Finalist

Presenter Organization Johns Hopkins University School of Medicine

Presenting Author Becca C. Engels

Track Transitions of Care

Keywords

AM-PAC

hospital mobility

discharge destination

post-acute care

machine learning

random forest

predictive model

hospital discharge planning

mobility score

electronic medical records

Acute Medical Profile-Pathway with Home Care AM-PAC Mobility Score

Hospital Discharge Disposition

Post-Acute Care

Home Discharge

Predictive Model