Paramedic PapersHome Operations & systems LLMs identify paediatric asthma in EMS data
Operations & systems
LLMs identify paediatric asthma in EMS data
Large Language Models Exhibit High Performance as Computable Phenotypes for Pediatric Asthma Exacerbation Detection in a Retrospective, Observational Study of Emergency Medical Services Encounters
Prehospital Emergency Care · Added 4 June 2026
They tested whether large language models could accurately identify paediatric asthma exacerbations from EMS documentation, finding high performance but noting this is early exploratory work with real-world deployment challenges.
- Study type
- Cohort
- Design
- Retrospective observational study of LLM computable phenotypes
- Participants
- 24,283 EMS encounters, patients aged 2-18 years; 1,000 annotated encounters
- Main finding
- Best model (Llama-3.3-70B, unstructured+structured data, 10-shot chain-of-thought) achieved F1 score 0.894; chain-of-thought prompts outperformed basic prompts in 4 of 5 models
- Caveat
- Early exploratory work; real-world deployment not tested
- Bottom line
- Hypothesis-generating
Learning goal
Understand how large language models could support identifying paediatric asthma exacerbations in EMS documentation systems.
More
- Topic: Operations & systems
- Daily scan: 4 June 2026
Related papers on Operations & systems
-
Documenting social risk factors in EMS charts
This pilot implementation tested whether EMS clinicians can reliably identify and document patient social risk factors during routine care, finding that structured documentation is feasible but requires workflow integration and clinician training.
-
Electric ambulances feasible in most dispatches
Retrospective NEMSIS analysis examined operational feasibility of electric ambulance adoption — fleet integration appears viable under specific conditions, though infrastructure readiness remains key.
-
Tatra mountain rescue epidemiology over 10 years
Retrospective analysis of 9,477 mountain rescue missions over 10 years found a median NACA score of 2 with 38% classified as moderate-to-severe, seasonal variation in severity, and helicopter use correlated with higher acuity, providing epidemiological data that could inform training and resource planning for wilderness EMS.
-
Modelling air vs ground EMS in Alberta
Geospatial analysis comparing response and transport times for air versus ground EMS in Alberta to identify optimal triage criteria.