- Only 52% of inpatient admissions are accurately predicted at the time of triage by emergency medicine (EM) physicians, but more discharges from the ED are accurately predicted.
Why this matters
- Better predictive capacity could decrease time spent in ED for patients ultimately admitted, increasing efficiency.
- Prospective cohort study of 35 EM physicians (with ≥5 years postresidency) who made predictions on 398 patients using nursing notes, vital signs, and electronic medical record (EMR) charts.
- Physician predictions were compared against the full EMR.
- Sensitivity was defined as the proportion of actual admissions that were accurately predicted.
- Specificity was defined as the proportion of actual discharges that were accurately predicted.
- Positive predictive value (PPV) was the proportion of admission predictions resulting in admission.
- Negative predictive value (NPV) was the proportion of discharge predictions resulting in discharge.
- Funding: HealthPartners Institute.
- Sensitivity of predicting admission was 51.8% for the entire cohort.
- Specificity was 89.1%.
- PPV was 56.4%.
- NPV was 87.2%.
- Sensitivity was increased to 61.5% and specificity was increased to 95.7% when predictions were made with a confidence level of >90%.
- For physicians who accurately predicted admission, the admitting service was correctly predicted 88.6% of the time.
- Single-center study.