Paramedic PapersHome Cardiac arrest Machine learning predicts good recovery after OHCA
Cardiac arrest
Machine learning predicts good recovery after OHCA
Early Prediction of Good Neurological Recovery in Comatose Out-of-Hospital Cardiac Arrest Survivors Using Machine Learning: A Multicenter Prospective Registry Study
Resuscitation Plus · 2026-10 · Added 30 October 2026
A South Korean multicentre prospective registry study developed a machine-learning model for comatose OHCA survivors that predicted good 6-month neurological recovery before targeted temperature management with AUROC 0.929, outperforming conventional scores like CAHP.
- Study type
- Registry
- Design
- Multicentre prospective registry (Korean Hypothermia Network, KORHN-PRO)
- Participants
- 1,358 comatose adult OHCA survivors treated with targeted temperature management
- Main finding
- The voting-classifier model predicted 6-month neurologically intact survival (29.3% of patients) with AUROC 0.929 and AUPRC 0.850; the top five predictors were prehospital initial rhythm, arrival rhythm, age, PaCO2 and total collapse duration
- Caveat
- Single-country registry from 2015–2018 and the model has not been externally validated
- Bottom line
- Hypothesis-generating
Why it matters in Australia (Medium relevance). Prognostication happens in intensive care, but the strongest predictors are the prehospital variables crews generate — initial rhythm, collapse duration — so early, accurate handover documentation is where this touches the road.
Learning goal
Update my judgement on early prognostication for comatose cardiac arrest survivors in the Australian context.
More
- Topic: Cardiac arrest
- Daily scan: 30 October 2026
- Podcast: Paramedic Papers Daily, 30 October 2026
Related papers on Cardiac arrest
-
AED coverage depends on how responders travel
A Chinese spatial optimisation study of 17,842 OHCA dispatch records in Wuhan modelled AED coverage under different responder travel modes and found demand-weighted coverage ranged from 85% (driving) to 32% (walking), with universal walking coverage needing 521 extra AEDs versus 68 by car.
-
Central venous access during arrest feasible in simulation
A US simulation study in a standard ambulance found central venous cannulation technically feasible during simulated refractory shockable cardiac arrest, completed in all runs with left femoral access identified as the most practical site and ACLS interventions maintained throughout.
-
ECPR timing in refractory OHCA: survival gap persists
A target trial emulation in 58,668 patients with refractory OHCA from the Korean nationwide registry found ECPR was associated with higher 30-day survival at 15, 30 and 60 minutes after ED arrival (+15.0, +14.4 and +7.7 percentage points); an exploratory analysis found the benefit narrowed but remained significant to 120 minutes.
-
Hot nights linked to more OHCA
A nationwide time-stratified case-crossover study of 135,898 emergency hospitalisations for OHCA in Japan found hot nights (minimum temperature above the regional 92.5th percentile) were associated with a 9% higher OHCA hospitalisation risk, with the highest risks in people aged 75+ and women in northern regions.