{"resourceId":"npj-dylm-ohca-emergency-calls-20260303","versions":[{"version":"external-42ea534e13e3023b33bec309df7943594ea2e079f3cfca84510a83c71b075862","resource":{"id":"npj-dylm-ohca-emergency-calls-20260303","title":"npj Digital Medicine: DyLM-OHCA language model detects OHCA in 60s of emergency calls (AUROC 0.937)","organization":"Korea University / Chung-Ang University / Incheon Fire Headquarters / Catholic University of Korea (npj Digital Medicine)","sector":"Peer-reviewed academic EMS dispatch research","geography":"South Korea (Seoul, Incheon, Gwangju 119 fire-based EMS); transferability limits for U.S. SLED","publishedAt":"2026-03-03","publicationDate":"2026-03-03","eventDate":"2026-03-03","sourceName":"npj Digital Medicine","sourceLabel":"Open-access npj Digital Medicine article on DyLM-OHCA early OHCA recognition from emergency call transcripts","sourceUrl":"https://www.nature.com/articles/s41746-026-02498-5","evidenceClass":"academic-research","outcomeClass":"mixed","topics":["knowledge-work","developers-agents","operating-model","accessibility-workforce"],"finding":"Open-access npj Digital Medicine article published 3 March 2026 develops DyLM-OHCA, a GPT-2-based dynamic language model for early out-of-hospital cardiac arrest recognition from 158,973 Korean 119 emergency call transcripts (2023; Seoul/Incheon/Gwangju). Within 60 seconds, DyLM-OHCA outperformed conventional ML (AUROC 0.937; AUPRC 0.456). At recall ~0.80, precision was 0.155, specificity 0.924 (FP rate 7.6%). Median recognition time was 19.7 s. Authors emphasize conversational context over keyword spotting; 52.6% of false positives later dropped below threshold. Limitations: retrospective single-country Korean-language data; no prospective dispatcher-in-the-loop outcomes; prior RCT (Blomberg) showed ML alerts did not improve recognition when dispatchers did not trust them. Framed as decision support, not automated decision-making.","sledRelevance":"Measured EMS/dispatch language-model evidence for life-safety call recognition with explicit HITL and transferability limits. Why now: not archived; newly relevant as U.S. PSAPs scale assistive 911 AI (QA/answering) and need outcome-grade call-AI evidence beyond vendor QA case studies after 2026-09-27.","evidence":"Opened nature.com HTML via curl. Publication 03 March 2026; n=158,973; AUROC 0.937; AUPRC 0.456; median 19.7 s; precision 0.155 at recall 0.80; FP 7.6%; Korean 119 setting and decision-support framing from article text.","architectureImplications":"Real-time transcript/ASR → incremental LM risk score → dispatcher alert with explainability cues; do not auto-dispatch or auto-start DA-CPR without human accept.","governanceImplications":"Require prospective HITL trials, language/protocol external validation, and trust/adoption metrics before operational alerts—retrospective AUROC is insufficient.","securityPrivacyImplications":"Emergency-call transcripts are highly sensitive; training/eval needs lawful access regimes (here AI Hub) and strict retention/minimization for any U.S. adaptation.","caveats":"Korean-language retrospective internal validation; PPV 15.5% implies substantial false alarms; no measured dispatcher adoption or survival benefit in this study.","streamIds":["emergency-services"],"roles":{"sales":"Frame DyLM-OHCA's AUROC 0.937 and ~20-second median OHCA cues as promising EMS decision-support evidence—not a proven U.S. dispatch product. Discovery should map DA-CPR protocols, ASR quality, language needs, and willingness to run prospective HITL trials given prior RCT trust failures. Bounded engagement: retrospective local transcript validation plus simulation of alert burden at ~15% PPV. Unsupported claims include U.S. readiness, survival improvement, or autonomous arrest recognition.","engineering":"Fit is incremental call-transcript language modeling for early OHCA risk with explainability. Architecture needs low-latency ASR, streaming inference, thresholding, and dispatcher UI that shows why the alert fired. Prerequisites: labeled local calls, protocol alignment, and compute at the ECC. Constraints: Korean-only training; 7.6% FP at target recall; no live RCT here. Proof of value is prospective dispatcher recognition/time-to-DA-CPR under HITL, not AUROC alone.","delivery":"Delivery requires dataset governance, ASR tuning, threshold calibration, and dispatcher trust training before any live OHCA alerts. Dependencies include the EMS medical director, ECC IT, and research partners. Ownership: medical director for clinical thresholds; ECC for workflow. Acceptance: documented human accept/reject of alerts, calibrated false-positive burden, and an external validation plan. Risks include ignored alerts (Blomberg-type) and resource strain from false positives."},"roleDetails":{"sales":{"customerProblem":"Dispatchers still miss a substantial share of OHCA in the first minute of emergency calls.","stakeholders":["EMS medical directors","ECC managers","Telecommunicators","Research/QI leads"],"discoveryQuestions":["Current OHCA recognition rate/time?","ASR already in place?","Language/protocol match?","Appetite for prospective HITL trial?"],"valueHypothesis":"Context-aware call LMs may speed OHCA cues if dispatchers trust and act on alerts—retrospective AUROC does not prove that.","potentialEngagement":"Local retrospective validation and alert-burden simulation before any live pilot.","unsupportedClaims":["U.S. performance equals Korean AUROC","Survival benefit proven","Autonomous DA-CPR triggering"]},"engineering":{"fit":"Streaming transcript LM for early OHCA decision support in EMS call-taking.","architectureIntegration":"Call audio → ASR transcript stream → DyLM-like risk score → dispatcher alert + attribution → human DA-CPR decision.","prerequisites":["ASR pipeline","Labeled OHCA outcomes","ECC compute/network","Alert UI"],"constraints":["Korean-trained weights","Low PPV (~0.155)","Retrospective only"],"security":"Minimize retention of call transcripts; control research vs operational data paths.","proofOfValue":["Prospective recognition rate","Time-to-DA-CPR","Alert accept rate and FP workload"]},"delivery":{"work":"Validate locally, calibrate thresholds, train dispatchers, and monitor alert fatigue.","dependencies":["Medical direction","ECC IT/ASR","QI analytics"],"ownership":"EMS medical director owns clinical thresholds; ECC supervisor owns live use.","skillsAdoption":"Teach dispatchers to treat scores as prompts for consciousness/breathing probes.","governanceCheckpoints":["IRB/data-use review","Go-live HITL checklist","Weekly FP review"],"acceptanceCriteria":["Human decision required","Documented reject reasons","External validation roadmap"],"risks":["Ignored alerts","False-alarm burden","Language transfer failure"]}},"retrievedAt":"2026-09-28T22:06:13-05:00","enrichedAt":"2026-09-28T22:06:13-05:00","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"May support dispatcher OHCA recognition if alerts are trusted and interpretable; risk of alert fatigue at 7.6% FP and low PPV.","procurementImplications":"Buy as research-backed decision support only with local validation budget—not as a turnkey OHCA detector from foreign retrospective metrics.","operatingModelImplications":"Keep medical dispatchers as final recognizers for DA-CPR; use model risk as an early cue during the first minute.","sourceVerification":{"responseId":"curl-npj-dylm-ohca-20260928T220613","openedUrl":"https://www.nature.com/articles/s41746-026-02498-5","referenceExcerpt":"DyLM-OHCA markedly outperformed all other benchmarks (AUROC = 0.937; AUPRC = 0.456).","promptVersion":"sled-research-v4.0","model":null,"basis":"agent-reported inspection"}}}]}