BRENNEN

Industries / Government, education and community / Government / Emergency services

Paramedic care

The treatment paramedics and other ambulance clinicians give at the scene, and the trials and rules that decide which treatments they carry.

6

Milestones achieved, from 1967 to 2024.

1

To watch for: announced, not yet confirmed outside.

2

Open deliverables. Nobody has shipped them.

The road

AchievedWatch forOpen

Achieved

  1. 1967

    Paramedics trained in Pittsburgh

    In 1967 Freedom House, a community group in Pittsburgh's Hill District, started an ambulance service whose crews, recruited largely from the neighborhood's Black residents, were trained by the anesthesiologist Peter Safar and the physician Nancy Caroline in medical care well beyond first aid. In its first year the service made 5,868 runs, by the University of Pittsburgh's count. The University of Pittsburgh's anesthesiology department and the EMS Museum both call its crews the first paramedics in the United States. In 1975 the city started its own ambulance service, did not renew Freedom House's contract, and gave its staff no special consideration for jobs.

    anesthesiology.pitt.eduemsmuseum.org

    Credit

    Freedom House; Peter Safar and Nancy Caroline

  2. 2012

    A CT scanner in ambulances

    A team in the German state of Saarland put a computed tomography (CT) scanner, a small laboratory and a telemedicine link into an ambulance, so stroke patients could be scanned and treated where they were found. In a randomized trial stopped at 100 patients, the median time from emergency call to treatment decision was 35 minutes, against 76 minutes for hospital care. In a larger study of 1,515 patients, funded by the US Patient-Centered Outcomes Research Institute and published in 2021, patients treated by such units got clot-dissolving drugs a median 72 minutes after their symptoms began, against 108 minutes with a standard ambulance, and had less disability at 90 days.

    ebi.ac.ukebi.ac.uk

    Credit

    S. Walter, K. Fassbender and colleagues, in The Lancet Neurology; J. C. Grotta and colleagues, BEST-MSU study

  3. 2018

    Epinephrine tested against salt water

    In the PARAMEDIC2 trial, paramedics in five NHS ambulance services gave 8,014 adults in cardiac arrest either epinephrine (adrenaline) or a saltwater placebo, without knowing which. At 30 days 3.2% of the epinephrine group were alive, against 2.4% of the placebo group, but the share who left hospital with good brain function did not differ significantly (2.2% against 1.9%). More survivors in the epinephrine group had severe brain damage (31.0% against 17.8%). By 2021, 5 of the 32 US states with statewide protocols had changed their epinephrine rules, mostly by capping the total dose, and none had dropped the drug, by a 2022 count.

    LessonA drug can restart more hearts without returning more people to the lives they had.

    ebi.ac.ukebi.ac.uk

    Credit

    G. D. Perkins and colleagues, in the New England Journal of Medicine; protocol count by E. Garfinkel and colleagues, Johns Hopkins University

  4. 2021

    Limits on sedation for police

    In 2019 paramedics in Aurora, Colorado, injected the sedative ketamine into Elijah McClain, a 23-year-old Black man, while police held him handcuffed on the ground. He died, and the autopsy was inconclusive about the cause. An independent report to the city found that the medics appeared to accept the officers' view that he had "excited delirium" without checking it, and gave a dose based on a gross overestimate of his weight. Colorado's House Bill 21-1251, signed in July 2021, limits paramedics' use of ketamine to sedate people to medical emergencies and bars officers from "compelling, directing, or unduly influencing" a paramedic to give it. The state health department suspended its waivers for ketamine sedation in "excited delirium" and agitation the same week.

    LessonThe decision to sedate has to rest on the paramedic's own examination, not on what officers at the scene believe.

    cdphe.colorado.govkunc.org

    Credit

    Colorado General Assembly; Colorado Department of Public Health and Environment; reporting by KUNC

  5. 2022

    Roadside blood, no gain

    Some ambulance and air ambulance teams carry blood to give to bleeding trauma patients before hospital. In the RePHILL trial, four UK prehospital critical care services gave 432 patients either red blood cells and dried plasma or saline; death or poor clearing of lactate, a sign of starved tissue, occurred in 64% against 65%, no real difference. In SWiFT, published in 2026, ten air ambulance services in England gave whole blood or standard blood components to 616 analyzed patients; death or massive transfusion within 24 hours occurred in 48.7% against 47.7%.

    LessonA treatment that works in hospital still has to be tested at the roadside, and two UK trials found no gain from the blood products they tested there.

    ebi.ac.ukebi.ac.ukaabb.org

    Credit

    N. Crombie, G. D. Perkins and the RePHILL collaborators; J. E. Smith and colleagues, SWiFT trial, funded by NHS Blood and Transplant

  6. 2024

    Bone or vein, same result

    Paramedics give cardiac arrest drugs either into a vein or through a needle into a bone (the intraosseous route), which may be quicker. In the PARAMEDIC-3 trial across 11 UK ambulance services, 6,082 adults were assigned to one route first. At 30 days 4.5% of the bone group and 5.1% of the vein group were alive, a difference that was not significant.

    ebi.ac.uk

    Credit

    K. Couper and colleagues, in the New England Journal of Medicine, published online October 2024

Watch for

  1. AI notes on ambulance tablets

    North East Ambulance Service in England is trialing software that listens to conversations between crews and patients through the iPads crews already carry, transcribes them and drafts a summary for the patient record. The eight-week trial involves 38 staff, a clinician reads and edits every note before it is filed, and it is due to end in November 2026. No supplier or results have been named.

    Achieved whenthe service or an outside group publishes how much time crews saved and how often the drafted notes got clinical facts, such as drugs and doses, wrong.

    resultsense.com

    Claimed by

    North East Ambulance Service, reported by the BBC, October 2026

Open

  1. 01

    Public test for ambulance scribes

    A public set of filmed and recorded simulated ambulance calls, run in a training lab with manikins and actors, each with a written answer key listing every finding, drug, dose, route and time. Score AI note-taking tools, open pipelines and commercial products used with their makers' permission, on how many of those facts the finished patient report gets wrong, leaves out or invents.

    The brief
    Credit · unclaimed
  2. 02

    Track protocols as trials land

    An open archive that downloads every published statewide and large-agency treatment protocol for ambulance crews on a schedule, keeps every version, and pulls fixed entries into one table: the epinephrine dose limit in cardiac arrest, which drug route to try first, whether crews carry blood, and the rules for sedation. Paramedics check a sample by hand. From the version history, report how long protocols take to change after trials such as PARAMEDIC2, PARAMEDIC-3 and SWiFT.

    The brief
    Credit · unclaimed

Open 01 · the brief

Public test for ambulance scribes

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

A public set of filmed and recorded simulated ambulance calls, run in a training lab with manikins and actors, each with a written answer key listing every finding, drug, dose, route and time. Score AI note-taking tools, open pipelines and commercial products used with their makers' permission, on how many of those facts the finished patient report gets wrong, leaves out or invents.

Why it is still open

The closest work, by Xiao Luo, Zhan Zhang and colleagues (2025), used 40 recordings of three pediatric simulations; four speech engines scored as low as 0.524 on airway details, on a 0-to-1 accuracy scale (F1), and a combination of four AI models pulling report fields from the transcripts reached 0.78 to 0.81. Those authors note a lack of publicly available EMS data, and their transcripts are available only on request. EgoEMS, released in November 2025 by a University of Virginia team, holds 233 simulated emergencies with time-stamped transcripts, available on request for non-commercial use; it labels the steps responders take, not the facts a patient report must hold, and no note-taking tool has been scored on it. A 2025 German benchmark used 99 synthetic dialogues and tested transcription only, and the North East Ambulance Service trial has published no results.

webpage.pace.eduwebpage.pace.eduarxiv.orgdataverse.harvard.edupmc.ncbi.nlm.nih.gov

Built with what exists

high-fidelity training manikins; tablet or body-worn microphones; speech-to-text such as Whisper or ElevenLabs Scribe v3 (see the Note "Fewer word errors"); general AI models to draft reports; the US National EMS Information System (NEMSIS) data dictionary as the list of report fields.

  1. First step

    Script 10 scenarios with a paramedic program's simulation lab, write the answer keys in NEMSIS fields, and record them.

  2. Done looks like

    100 scenarios covering at least 10 kinds of call, adult and child, released under an open license with answer keys, and at least three note-taking tools scored, with each tool's share of drugs recorded with the correct dose, route and time and its count of invented facts per report published.

Open 02 · the brief

Track protocols as trials land

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

An open archive that downloads every published statewide and large-agency treatment protocol for ambulance crews on a schedule, keeps every version, and pulls fixed entries into one table: the epinephrine dose limit in cardiac arrest, which drug route to try first, whether crews carry blood, and the rules for sedation. Paramedics check a sample by hand. From the version history, report how long protocols take to change after trials such as PARAMEDIC2, PARAMEDIC-3 and SWiFT.

Why it is still open

Comparisons are done by hand, one topic and one moment at a time. A 2022 Johns Hopkins review compared statewide cardiac arrest protocols from 2018 and 2021, found statewide protocols in only 32 of the 50 states and DC, and left out local and regional protocols, which its authors say can adjust to new research more quickly.

cambridge.org

Built with what exists

a web crawler; PDF text extraction such as pdftotext; an AI model that answers fixed questions about each document, such as Gemini Decision 1 (see the Note "A thousand questions, one pass"); Git for version history.

  1. First step

    Download the current cardiac arrest protocol of every state that has one, extract its epinephrine rules, and compare them with the 2022 review's table.

  2. Done looks like

    protocols for all 50 states and DC plus the 100 largest US ambulance agencies archived monthly for a year, 20 entries extracted from each, a random 5% of entries checked by two paramedics with at least 97% agreement, and a published table of months from each trial's publication to each protocol's change.