HomeAsk EMS Allocation

Prehospital emergency care · Patient safety

Crafting the
patient safety net.

How should ambulance systems allocate scarce capacity when several patients need help at the same time?

Research on ambulance allocation, waiting, system pressure and patient safety.

Explore the observatoryResearch

EMS is a dynamic patient safety net.

Evidence at a glance

Four studies. One question.

4
Linked studies
3
Peer-reviewed articles
1
Preprint · not peer reviewed
2017–2022
Study II–IV data period
See the evidence

Choose your path

I am a…

What happens when ambulance demand approaches capacity?

In brief

EMS is a dynamic patient safety net.

  1. Dispatchers described balancing current patient needs with readiness for the next emergency.

    Stewarding scarce response capacity

  2. Relationships between response time and clinical urgency varied with patient characteristics; the exploratory model was not validated for individual prediction.

    Breathing emergencies and nonlinear risk

  3. Response-time distributions varied with system conditions. Typical response times alone can conceal long waits.

    Understanding EMS response times

These studies describe professional experience and observational associations. They do not demonstrate that a particular allocation policy improves clinical outcomes. The interactive model is illustrative, not validated for individual prognosis.

Policy brief — Research summary

Understand the research through the observatory

The observatory is a learning tool for the four studies: explore a question, compare synthetic scenarios, then examine the research and its limitations.

How the four studies connect
  1. 01 · How is scarce response capacity shared?Peer reviewedStewarding scarce response capacity
  2. 02 · Why might the same wait matter differently?Peer reviewedBreathing emergencies and nonlinear risk
  3. 03 · What is uncertain at the emergency call?Preprint · not peer reviewedInfectious presentations and concordance
  4. 04 · What can the longest response times reveal?Peer reviewedUnderstanding EMS response times

Study findings are empirical evidence. Response times here are synthetic calculations. Patient colours illustrate different sensitivity to waiting; they are not validated clinical predictions. Neither the simulation nor weather and traffic inputs establish a causal effect on patient outcomes.

Preparing the observatory…

Four studies. One question.

From dispatch decisions to clinical uncertainty and system delays. Three peer-reviewed articles and one preprint contribute different parts of the picture.

  1. Emergency call

  2. Dispatch assessment

    Peer reviewed

    Stewarding scarce response capacity

    Dispatchers described a professional responsibility to steward scarce response capacity through proportionate allocation, preserved availability and coordinated system response.

    Question, findings and limitations
    Research question
    How dispatchers prioritise patients, govern the queue and preserve readiness when ambulance capacity is constrained.
    Study population · Method
    Qualitative interviews with 13 emergency medical dispatchers working across 10 emergency medical communication centres; analysed using inductive content analysis.
    Interpretation boundaries
    The findings describe professional experience in one organisational context and should be transferred to other EMS systems with attention to local governance.

    How is scarce response capacity shared?

    Explore this research question

    Dispatch & queue governance

    BMJ Open · 2026Explore study 01

  3. Resource allocation / waiting

    Peer reviewed

    Breathing emergencies and nonlinear risk

    Age was the most important model feature. Patients older than 60 showed complex rising HRTS probability at prolonged response times beyond two hours; discrimination was modest (AUC 0.66).

    Question, findings and limitations
    Research question
    How response time, age and sex interact with high-risk time-sensitive triage among 132,395 breathing-problem missions.
    Study population · Method
    Retrospective observational analysis of 132,395 breathing-problem missions in Stockholm, 2017–2022. Multiple models were explored, with PD and ICE plots used to examine nonlinear relationships.
    Interpretation boundaries
    The exploratory model is not validated for patient-level prediction or dispatch automation and was based on limited variables from one region.

    Why might the same wait matter differently?

    Explore this research question

    Patient vulnerability & waiting

    BMC Medical Informatics and Decision Making · 2025Explore study 02

  4. First EMS assessment

    Preprint · not peer reviewed

    Infectious presentations and concordance

    The three information layers did not represent the same clinical construct; concordance changed depending on where in the care pathway infection and urgency were assessed.

    Question, findings and limitations
    Research question
    How dispatch suspicion, on-scene phenotype and high-risk time-sensitive triage align in infectious presentations.
    Study population · Method
    Retrospective machine-learning study comparing dispatch suspicion of infection, the on-scene symptom phenotype and high-risk time-sensitive triage.
    Interpretation boundaries
    This preprint has not completed journal peer review. Observational associations do not establish causal effects or operational thresholds.

    What is uncertain at the emergency call?

    Explore this research question

    Information across the care pathway

    Research Square · 2026Explore study 03

  5. System response-time distribution

    Peer reviewed

    Understanding EMS response times

    Priority was the strongest overall feature, but workload, call reason, geography, precipitation, temperature and operational intervals interacted. Lower priorities had markedly long-tailed response-time distributions.

    Question, findings and limitations
    Research question
    How priority, workload, geography, weather and operational intervals shape variability and tail delays.
    Study population · Method
    Retrospective analysis of 1,144,754 EMS missions in Stockholm, 2017–2022, integrating operational, geographic and weather variables with regression and machine-learning analyses.
    Interpretation boundaries
    Traffic information was unavailable, the study was conducted in one region and predictive performance does not make the models ready for real-time deployment.

    What can the longest response times reveal?

    Explore this research question

    System capacity & response-time variation

    BMC Medical Informatics and Decision Making · 2025Explore study 04

Why this matters

A safety net is more than a response-time target.

Speed matters. But it cannot, on its own, describe patient safety.

Ambulance systems operate with uncertain information, fluctuating demand, geographic variation and finite resources. Every allocation decision shapes who receives help first, who waits and how safety margins are distributed across the system.

Explore the three principles

From speed to system behaviour

Response time emerges from dispatch, triage, workload, geography, availability, call handling and travel.

From queue to virtual waiting room

Queued patients require active monitoring, reassessment, escalation and organisational ownership.

From data to responsible governance

Routine EMS data can reveal patterns, provided uncertainty and interpretation boundaries remain visible.

A patient’s journey · about 45 seconds

The queue is a virtual waiting room.

A general care pathway · no patient data or clinical advice.

A patient has called for help. No ambulance is available yet. Waiting is already part of the care pathway.

  1. Emergency call
  2. Dispatch assessment
  3. Queue / waiting
  4. Possible reassessment
  5. Resource allocation
  6. EMS arrival
  7. First EMS assessment

The emergency call

A person calls for help. The dispatcher receives incomplete information while several other patients also need a response. Available resources and the initial assessment shape priority and allocation.

What this means for EMS systems

Use the evidence to frame system questions, then evaluate changes in the local service.

Protect readiness

Follow available capacity and queue pressure together. Preserve room to absorb simultaneous needs, especially close to capacity.

Look beyond the typical response

Examine tail delays and geographic variation in ambulance response time. A system average can conceal where waiting accumulates.

Treat risk as dynamic

Make reassessment and ownership of waiting patients explicit. Clinical urgency can change while resources remain committed.

Evaluate resilience

Test how the system absorbs disruption and recovers readiness. Assess patient safety, distributional effects and human judgement before operational use.

Doctoral thesis & publications

Doctoral thesis · 2026Peer-reviewed publicationsResearch in progress

Peter Hill · Karolinska Institutet · DOI 10.69622/31262914

Research question and planned approach only. Unpublished results are not disclosed.

RESEARCH GUIDE

Ask EMS Allocation

Explore the research through questions.

This AI research guide uses the published research, thesis and explanatory material on this site. Answers should cite their sources and distinguish evidence status. It can make mistakes and is not clinical decision support. Questions are processed by Mistral AI; do not enter personal or patient information. Research integrity

Research · Collaboration · Speaking

Let’s take the research further

EMS & health-system development

Explore readiness, allocation and patient safety in your system.

Start a conversation

Speaking & media

Discuss a conference contribution, interview or research briefing.

Get in touch
Peter Hill · peter.hill@ki.se
Karolinska Institutet profileLinkedIn

Research integrity

The research is observational and explanatory. It does not provide patient-level advice or a validated real-time dispatch tool. Operational translation requires prospective evaluation, governance, human-factors design and safeguards.

Methodological limitations & research integrity

References & research resources

Bring the evidence into teaching, research and conversations about prehospital care. Reference files include the thesis, three journal articles and one preprint.

Policy brief · Research summaryAll references · RIS / EndNoteAll references · BibTeXFollow publications · RSS
Peer-reviewed publications
  1. Hill P, Lederman J, Jonsson D, Bolin P, Vicente V. Stewarding scarce response capacity: an inductive qualitative interview study of emergency medical dispatchers prioritising ambulance resources. BMJ Open. 2026. doi:10.1136/bmjopen-2026-118269
  2. Hill P, Jonsson D, Lederman J, Bolin P, et al. Uncovering nonlinear patterns in time-sensitive prehospital breathing emergencies: an exploratory machine learning study. BMC Med Inform Decis Mak. 2025;25:205. doi:10.1186/s12911-025-03046-z
  3. Hill P, Lederman J, Jonsson D, Bolin P, et al. Understanding EMS response times: a machine learning-based analysis. BMC Med Inform Decis Mak. 2025;25:143. doi:10.1186/s12911-025-02975-z
Doctoral thesis · 2026

Hill P. Adaptive emergency medical services allocation: crafting the patient safety net in prehospital emergency care. Stockholm: Karolinska Institutet; 2026. doi:10.69622/31262914

Preprint · not peer reviewed

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

Methodological limitations & research integrity

Exploration before deployment

The research is observational and explanatory. It does not provide patient-level advice or a validated real-time dispatch tool. Operational translation requires prospective evaluation, governance, human-factors design and safeguards.

Illustrative teaching model · no live data · not clinical decision support. Values are neither empirical findings nor patient-level predictions.

Methodological limitations & research integrity
LinkedIn