DyLM-OHCA identified out-of-hospital cardiac arrest within 60 seconds of emergency calls with an AUROC of 0.937, significantly outperforming conventional machine learning algorithms.
Observational (n=158,973)
Yes
Does a dynamic deep learning model (DyLM-OHCA) improve early out-of-hospital cardiac arrest (OHCA) detection compared to conventional machine learning algorithms in emergency calls?
A dynamic deep learning model analyzing real-time emergency call transcripts significantly outperformed conventional machine learning algorithms in detecting out-of-hospital cardiac arrest within 60 seconds.
Effect estimate: AUROC 0.937
p-value: p=<0.001
We developed a dynamic deep learning model (DyLM-OHCA) for early out-of-hospital cardiac arrest (OHCA) detection. Using 158,973 emergency call transcripts from three South Korean metropolitan regions, we trained DyLM-OHCA for 60 s OHCA identification and compared its performance against four conventional machine learning algorithms—Logistic Regression, XGBoost, Gradient Boosting, and Random Forest. DyLM-OHCA markedly outperformed all other benchmarks (AUROC = 0.937; AUPRC = 0.456). We analyzed global and sample-level word importance and temporally predicted OHCA risk patterns. Word attribution revealed differences in important words between callers and dispatchers. OHCA recognition was influenced more by conversational flow than by individual keywords. True-positive cases sustained high-risk scores, whereas over half of false-positive cases showed early risk score decline. DyLM-OHCA captures clinically meaningful dialog patterns, moving beyond simple keyword spotting. By providing real-time, context-aware, and interpretable risk assessments, our model is potentially valuable in decision support, enhancing dispatcher confidence, and improving early OHCA recognition.
Choi et al. (Tue,) conducted a observational in Out-of-hospital cardiac arrest (OHCA) (n=158,973). DyLM-OHCA (Dynamic deep language model) vs. Conventional machine learning algorithms (Logistic Regression, XGBoost, Gradient Boosting, Random Forest) was evaluated on OHCA identification within 60 seconds of call onset (AUROC 0.937, p=<0.001). DyLM-OHCA identified out-of-hospital cardiac arrest within 60 seconds of emergency calls with an AUROC of 0.937, significantly outperforming conventional machine learning algorithms.