The effects of applying artificial intelligence to triage in the emergency department: A systematic review of prospective studies.
Nayeon Yi, Dain Baik, Gumhee Baek
PMID 39262027WHAT IT FOUND
AI triage showed varied accuracy, from 80.5% to 99.1% in four studies.
One voice-assisted system was 27 seconds faster than manual documentation, but records were complete only 81.84%. Nurses should not assume AI replaces triage judgment.
Key findings
01Seven prospective cohort studies of ED patients, with sample sizes from 146 to 17,072, were included; four reported AI triage prediction accuracy from 80.5% to 99.1%.
02One study found AI voice-assisted triage was 27 seconds faster than manual input, but record completion was 81.84%.
03A symptom app compared with manual MTS overtriaged 57.1% and undertriaged 8.9%, and 94.7% of cases were considered safe; a machine-learning protocol reported life-threatening mis-triage of 0.9% in intervention and 1.2% in control.
STILL TO COME
How it was doneWhat they foundWhat it means for RNs
Read the rest of this summary
You get three full summaries a month, free, and we do not ask for a card. Search, the TL;DRs and your library stay unlimited either way.
What it does not show
The review included only a small number of prospective observational studies, not randomized trials, so it cannot establish that AI triage improves patient outcomes. Included studies were heterogeneous across countries, ED populations, triage scales, AI methods, and comparison groups. AI accuracy was judged against nurse or physician triage, which may introduce subjective bias. Some included studies did not fully report study-size calculation, missing-value handling, confounding control, or sensitivity analysis. Severe or life-threatening patients were excluded or poorly classified in some studies, limiting applicability to highest-acuity cases.
Declared interests
The authors declared no conflict of interest.
The easy way to misread this
Do not read the highest reported accuracy as proof that AI triage can replace nurse judgment. The review included only prospective observational studies, and performance varied by triage scale, AI method, patient population, and acuity.