Study III

Study III · Preprint · Research Square

Infection suspicion and triage concordance

A retrospective machine-learning study of concordance between dispatch suspicion, on-scene phenotype, and high-risk time-sensitive triage in prehospital infectious presentations.

Open preprint Back to thesis All publications
Study profile

Concordance across dispatch, phenotype, and triage

The study examines how EMCC Fever/Infection categorisation, the first on-scene ESS label, and high-risk RETTS triage align during the early prehospital course.

  • Preprint: Hill P, Jonsson D, Lederman J, Bolin P, Vicente V. Concordance Between Dispatch Suspicion, On-Scene Phenotype, and Time Sensitive Triage in Prehospital Infectious Presentations: A Retrospective Machine-Learning Study. Research Square. 2026.
  • DOI: 10.21203/rs.3.rs-7651316/v1.
  • Design: Population-based retrospective observational study with exploratory machine learning.
  • Participants/data: 34,779 EMS assignments initially categorised as Fever/Infection at EMCC in Region Stockholm, 2017–2022.
  • Outcome: High-risk triage at first EMS contact, defined as RETTS Red/Orange.
  • Methods: Gradient boosting, random forest, logistic regression and ensemble models using routinely recorded dispatch, patient, time, weather, workload and on-scene ESS variables.
  • Key contribution: The study makes visible where dispatch suspicion, early clinical phenotype, and high-risk triage align, while explicitly separating concordance from independent predictive validity.
Core findings

Where infectious suspicion aligns with severity

34,779Fever/Infection assignments included in the analytic cohort.
64.0%Had infection-coded ESS labels at first EMS contact.
71.5%HRTS among infection-coded ESS cases.
44.3%HRTS among other ESS categories.

Moderate concordance

Infection-coded ESS labels were common among Fever/Infection calls, indicating meaningful but incomplete alignment between dispatch suspicion and on-scene phenotype.

Age threshold pattern

Partial dependence analyses indicated a threshold-like rise in predicted high-risk triage probability beyond approximately 60 years.

Concordance, not independent prediction

Because ESS contributes to RETTS, ESS-HRTS alignment should be interpreted as agreement across system classifications rather than independent predictive validity.

Scientific contribution

Making alignment and mismatch visible

The study clarifies how early dispatch information, on-scene phenotype, and triage severity converge or diverge in infectious EMS presentations.

Study III supports secondary triage and quality-improvement thinking by identifying where dispatch suspicion aligns with later observed phenotype and where high-risk patients may appear outside infection-coded ESS categories.

Fever/Infection ESS phenotype RETTS Red/Orange Concordance Secondary triage

Interpretation note

This study is a preprint and should be interpreted as work under review. The analysis is useful for system diagnostics and secondary triage design, but it does not establish deployable dispatch-time prediction. Because ESS contributes to RETTS, relationships involving ESS and HRTS are presented as concordance rather than independent prediction.