DailyDoctor Archive Specialties Get app
Back to the 21 October 2026 edition

Research · 02 of 06

A five-item function score predicts nursing-facility discharge better than clinicians

Record prior and current physical function daily; it flags patients who will need a facility discharge earlier than clinical judgement.

Design
Retrospective cohort, 21 hospitals
Population
493,035 inpatient and observation encounters
Primary outcome
Prediction of discharge disposition within 24 hours
Effect
SNF discharge AUC 0.89 with PF5 vs 0.75 with clinician estimate (day 3)

A Journal of Hospital Medicine cohort study examined the PROMIS physical function five-item tool (PF5), recorded daily in the electronic record across 21 Kaiser Permanente hospitals. It captures prior function from the patient and current function from a clinician.

Across 493,035 encounters, models including PF5 predicted discharge to a skilled nursing facility within 24 hours with an AUC of 0.89 on day 3, against 0.75 using clinicians' expected discharge disposition alone. Current function was the strongest predictor. The tool was most useful early in the stay, when clinicians' estimates were often missing.

Discharge planning that starts on the day before discharge is a common cause of delay. A simple, repeated function score gives the team an earlier and more objective signal of who will need a placement or home services.

The study shows prediction, not benefit; it did not test whether acting on the score shortens stays. But recording function daily costs little.

  • Record each patient's function before admission and their current function, and update it daily
  • Start placement or home-service planning early for patients whose current function has dropped
  • Do not rely on an expected-discharge field that is often blank early in the stay
  • Involve therapy teams early for patients with a large gap between prior and current function

Why it matters

Discharge delays often start with function that nobody measured on day one.

Don't overread it

Better prediction has not yet been shown to shorten stays or improve outcomes.

The statistics, in plain English

An AUC of 0.89 means the model ranks a patient who goes to a facility above one who does not 89% of the time; 0.75 is useful, 0.89 is strong. These were retrospective models in one health system.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

QR code to install Daily Doctor
Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.

evidenceinpatienttransitions

Tomorrow morning, before your first patient

One edition a day for internal medicine, written by the desk, every claim tied to its paper. Six minutes.

Get the app — free
Daily Doctor All 27 specialties, every morning. Free.
Get the app