The aim of this study was to summarize the literature on the applications of machine learning (ML) and their performance in Emergency Medical Services (EMS).

Four relevant electronic databases were searched (from inception through January 2024) for all original studies that employed EMS-guided ML algorithms to enhance the clinical and operational performance of EMS. Two reviewers screened the retrieved studies and extracted relevant data from the included studies. The characteristics of included studies, employed ML algorithms, and their performance were quantitively described across primary domains and subdomains.

This review included a total of 164 studies published from 2005 through 2024. Of those, 125 were clinical domain focused and 39 were operational. The characteristics of ML algorithms such as sample size, number and type of input features, and performance varied between and within domains and subdomains of applications. Clinical applications of ML algorithms involved triage or diagnosis classification (n = 62), treatment prediction (n = 12), or clinical outcome prediction (n = 50), mainly for out-of-hospital cardiac arrest/OHCA (n = 62), cardiovascular diseases/CVDs (n = 19), and trauma (n = 24). The performance of these ML algorithms varied, with a median area under the receiver operating characteristic curve (AUC) of 85.6%, accuracy of 88.1%, sensitivity of 86.05%, and specificity of 86.5%. Within the operational studies, the operational task of most ML algorithms was ambulance allocation (n = 21), followed by ambulance detection (n = 5), ambulance deployment (n = 5), route optimization (n = 5), and quality assurance (n = 3). The performance of all operational ML algorithms varied and had a median AUC of 96.1%, accuracy of 90.0%, sensitivity of 94.4%, and specificity of 87.7%. Generally, neural network and ensemble algorithms, to some degree, out-performed other ML algorithms.

Triaging and managing different prehospital medical conditions and augmenting ambulance performance can be improved by ML algorithms. Future reports should focus on a specific clinical condition or operational task to improve the precision of the performance metrics of ML models.

The European Society of Cardiology (ESC) 2020 guidelines propose an algorithm for in-hospital management of non-ST-elevation myocardial infarction (NSTEMI) based on risk stratification according to clinical, electrocardiographic, and biological data. However, out-of-hospital management is not codified.

The objective of the present study was to evaluate the role of high-sensitivity cardiac troponin-I in out-of-hospital management of NSTEMI by Emergency Medical Services (EMS).

This monocentric, retrospective, observational study analyzed the files of all patients having received a troponin assay in the EMS of Beaujon University Hospital, AP-HP (Paris region, France) from January 1, 2020 through December 31, 2020. Patients were classified as low risk, high risk, or very high risk according to the ESC 2020 algorithm at the time of their hospital treatment. The relationship between troponin in point-of-care and risk level according to time to onset of pain was analyzed using logistic regression. A search for predictors of risk level was performed using multivariate analysis. A P value <.05 was considered significant.

Out of 309 patients in the file, 233 were included. Men were 61% and the median age was 63 years. A positive troponin assay was associated with high-risk or very high-risk stratification regardless of the time to onset of pain (P <.0001). Predictive factors for being classified as high or very high risk in hospital were: a history of atrial fibrillation (P = .03), electrocardiogram (ECG) modifications such as negative T wave or ST-segment depression (P <.0001), and positive troponin (P <.0001).

The use of point-of-care troponin in EMS, combined with clinical and electrical criteria, allows risk stratification of NSTEMI patients from the prehospital management stage and optimization of referral to an appropriate care pathway. Patients classified as low risk should be referred to the emergency department (ED) and patients classified as high risk or very high risk to the cardiac intensive care unit or percutaneous coronary intervention (PCI) center.

Existing peer-reviewed literature describing emergency medical technician (EMT) acquisition and transmission of 12-lead electrocardiograms (12L-ECGs), in the absence of a paramedic, is largely limited to feasibility studies.

The objective of this retrospective observational study was to describe the impact of EMT-acquired 12L-ECGs in Suffolk County, New York (USA), both in terms of the diagnostic quality of the transmitted 12L-ECGs and the number of prehospital percutaneous coronary intervention (PCI)-center notifications made as a result of transmitted 12L-ECGs demonstrating a ST-elevation myocardial infarction (STEMI).

A pre-existing database was queried for Emergency Medical Services (EMS) calls on which an EMT acquired a 12L-ECG from program initiation (January 2017) through December 31, 2019. Scanned copies of the 12L-ECGs were requested in order to be reviewed by a blinded emergency physician.

Of the 665 calls, 99 had no 12L-ECG available within the database. For 543 (96%) of the available 12L-ECGs, the quality was sufficient to diagnose the presence or absence of a STEMI. Eighteen notifications were made to PCI-centers about a concern for STEMI. The median time spent on scene and transporting to the hospital were 18 and 11 minutes, respectively. The median time from PCI-center notification to EMS arrival at the emergency department (ED) was seven minutes (IQR 5-14).

In the event a cardiac monitor is available, after a limited educational intervention, EMTs are capable of acquiring a diagnostically useful 12L-ECG and transmitting it to a remote medical control physician for interpretation. This allows for prehospital PCI-center activation for a concern of a 12L-ECG with a STEMI, in the event that a paramedic is not available to care for the patient.

Chest pain is one of the most common reasons for 999 calls and transfers to the emergency department (ED). In these patients, acute myocardial infarction (AMI) is often the diagnosis that clinicians are seeking to exclude. However, only a minority of those patients have AMI, causing a substantial financial burden to health services. Cardiac troponin (cTn) is the reference standard biomarker for the diagnosis of AMI. Several commercially available point-of-care (POC) cTn assays are portable and could feasibly be used in an ambulance. The aim of this paper is to systematically review existing evidence for the use of POC cTn assays in the prehospital setting to rule out AMI.

A systematic search was conducted on EMBASE, MEDLINE, and CINAHL Plus databases, reference lists, and relevant grey literature, including combinations of the relevant terms. Papers published in English language since the year 2000 were eligible for inclusion. A narrative synthesis of the evidence was then undertaken.

The initial search and cross-referencing revealed a total of 350 papers, of which 243 were excluded. Seven papers were included in the systematic literature review.

Current evidence does not support the use of POC troponin assays to exclude AMI due to issues with diagnostic accuracy and insufficient high-quality evidence.

Emergency Medical Services (EMS) systems have developed protocols for prehospital activation of the cardiac catheterization laboratory for patients with suspected ST-elevation myocardial infarction (STEMI) to decrease first-medical-contact-to-balloon time (FMC2B). The rate of “false positive” prehospital activations is high. In order to decrease this rate and expedite care for patients with true STEMI, the American Heart Association (AHA; Dallas, Texas USA) developed the Mission Lifeline PreAct STEMI algorithm, which was implemented in Los Angeles County (LAC; California USA) in 2015. The hypothesis of this study was that implementation of the PreAct algorithm would increase the positive predictive value (PPV) of prehospital activation.

This is an observational pre-/post-study of the effect of the implementation of the PreAct algorithm for patients with suspected STEMI transported to one of five STEMI Receiving Centers (SRCs) within the LAC Regional System. The primary outcome was the PPV of cardiac catheterization laboratory activation for percutaneous coronary intervention (PCI) or coronary artery bypass graft (CABG). The secondary outcome was FMC2B.

A total of 1,877 patients were analyzed for the primary outcome in the pre-intervention period and 405 patients in the post-intervention period. There was an overall decrease in cardiac catheterization laboratory activations, from 67% in the pre-intervention period to 49% in the post-intervention period (95% CI for the difference, -14% to -22%). The overall rate of cardiac catheterization declined in post-intervention period as compared the pre-intervention period, from 34% to 30% (95% CI, for the difference -7.6% to 0.4%), but actually increased for subjects who had activation (48% versus 58%; 95% CI, 4.6%-15.0%). Implementation of the PreAct algorithm was associated with an increase in the PPV of activation for PCI or CABG from 37.9% to 48.6%. The overall odds ratio (OR) associated with the intervention was 1.4 (95% CI, 1.1-1.8). The effect of the intervention was to decrease variability between medical centers. There was no associated change in average FMC2B.

The implementation of the PreAct algorithm in the LAC EMS system was associated with an overall increase in the PPV of cardiac catheterization laboratory activation.

Field identification of ST-elevation myocardial infarction (STEMI) and advanced hospital notification decreases first-medical-contact-to-balloon (FMC2B) time. A recent study in this system found that electrocardiogram (ECG) transmission following a STEMI alert was frequently unsuccessful.

Instituting weekly test ECG transmissions from paramedic units to the hospital would increase successful transmission of ECGs and decrease FMC2B and door-to-balloon (D2B) times.

This was a natural experiment of consecutive patients with field-identified STEMI transported to a single percutaneous coronary intervention (PCI)-capable hospital in a regional STEMI system before and after implementation of scheduled test ECG transmissions. In November 2014, paramedic units began weekly test transmissions. The mobile intensive care nurse (MICN) confirmed the transmission, or if not received, contacted the paramedic unit and the department’s nurse educator to identify and resolve the problem. Per system-wide protocol, paramedics transmit all ECGs with interpretation of STEMI. Receiving hospitals submit patient data to a single registry as part of ongoing system quality improvement. The frequency of successful ECG transmission and time to intervention (FMC2B and D2B times) in the 18 months following implementation was compared to the 10 months prior. Post-implementation, the time the ECG transmission was received was also collected to determine the transmission gap time (time from ECG acquisition to ECG transmission received) and the advanced notification time (time from ECG transmission received to patient arrival).

There were 388 patients with field ECG interpretations of STEMI, 131 pre-intervention and 257 post-intervention. The frequency of successful transmission post-intervention was 73% compared to 64% prior; risk difference (RD)=9%; 95% CI, 1-18%. In the post-intervention period, the median FMC2B time was 79 minutes (inter-quartile range [IQR]=68-102) versus 86 minutes (IQR=71-108) pre-intervention (P=.3) and the median D2B time was 59 minutes (IQR=44-74) versus 60 minutes (IQR=53-88) pre-intervention (P=.2). The median transmission gap was three minutes (IQR=1-8) and median advanced notification time was 16 minutes (IQR=10-25).

Implementation of weekly test ECG transmissions was associated with improvement in successful real-time transmissions from field to hospital, which provided a median advanced notification time of 16 minutes, but no decrease in FMC2B or D2B times.

D’ArcyNT, BossonN, KajiAH, BuiQT, FrenchWJ, ThomasJL, ElizarrarazY, GonzalezN, GarciaJ, NiemannJT. Weekly Checks Improve Real-Time Prehospital ECG Transmission in Suspected STEMI. Prehosp Disaster Med. 2018;33(3):245–249.

The History, Electrocardiogram (ECG), Age, Risk Factors, and Troponin (HEART) score is a decision aid designed to risk stratify emergency department (ED) patients with acute chest pain. It has been validated for ED use, but it has yet to be evaluated in a prehospital setting.

A prehospital modified HEART score can predict major adverse cardiac events (MACE) among undifferentiated chest pain patients transported to the ED.

A retrospective cohort study of patients with chest pain transported by two county-based Emergency Medical Service (EMS) agencies to a tertiary care center was conducted. Adults without ST-elevation myocardial infarction (STEMI) were included. Inter-facility transfers and those without a prehospital 12-lead ECG or an ED troponin measurement were excluded. Modified HEART scores were calculated by study investigators using a standardized data collection tool for each patient. All MACE (death, myocardial infarction [MI], or coronary revascularization) were determined by record review at 30 days. The sensitivity and negative predictive values (NPVs) for MACE at 30 days were calculated.

Over the study period, 794 patients met inclusion criteria. A MACE at 30 days was present in 10.7% (85/794) of patients with 12 deaths (1.5%), 66 MIs (8.3%), and 12 coronary revascularizations without MI (1.5%). The modified HEART score identified 33.2% (264/794) of patients as low risk. Among low-risk patients, 1.9% (5/264) had MACE (two MIs and three revascularizations without MI). The sensitivity and NPV for 30-day MACE was 94.1% (95% CI, 86.8-98.1) and 98.1% (95% CI, 95.6-99.4), respectively.

Prehospital modified HEART scores have a high NPV for MACE at 30 days. A study in which prehospital providers prospectively apply this decision aid is warranted.

StopyraJP, HarperWS, HigginsTJ, ProkesovaJV, WinslowJE, NelsonRD, AlsonRL, DavisCA, RussellGB, MillerCD, MahlerSA. Prehospital Modified HEART Score Predictive of 30-Day Adverse Cardiac Events. Prehosp Disaster Med. 2018;33(1):58–62.

Early recognition of an acute myocardial infarction (AMI) can increase the patient’s likelihood of survival. As the first point of contact for patients accessing medical care through emergency services, emergency medical dispatchers (EMDs) represent the earliest potential identification point for AMIs. The objective of the study was to determine how AMI cases were coded and prioritized at the dispatch point, and also to describe the distribution of these cases by patient age and gender.

No studies currently exist that describe the EMD’s ability to correctly triage AMIs into Advanced Life Support (ALS) response tiers.

The retrospective descriptive study utilized data from three sources: emergency medical dispatch, Emergency Medical Services (EMS), and emergency departments (EDs)/hospitals. The primary outcome measure was the distributions of AMI cases, as categorized by Chief Complaint Protocol, dispatch priority code and level, and patient age and gender. The EMS and ED/hospital data came from the Utah Department of Health (UDoH), Salt Lake City, Utah. Dispatch data came from two emergency communication centers covering the entirety of Salt Lake City and Salt Lake County, Utah.

Overall, 89.9% of all the AMIs (n=606) were coded in one of the three highest dispatch priority levels, all of which call for ALS response (called CHARLIE, DELTA, and ECHO in the studied system). The percentage of AMIs significantly increased for patients aged 35 years and older, and varied significantly by gender, dispatch level, and chief complaint. A total of 85.7% of all deaths occurred among patients aged 55 years and older, and 88.9% of the deaths were handled in the ALS-recommended priority levels.

Acute myocardial infarctions may present as a variety of clinical symptoms, and the study findings demonstrated that more than one-half were identified as having chief complaints of Chest Pain or Breathing Problems at the dispatch point, followed by Sick Person and Unconscious/Fainting. The 35-year age cutoff for assignment to higher priority levels is strongly supported. The Falls and Sick Person Protocols offer opportunities to capture atypical AMI presentations.

ClawsonJJ, GardettI, ScottG, FivazC, BarronT, BroadbentM, OlolaC. Hospital-Confirmed Acute Myocardial Infarction: Prehospital Identification Using the Medical Priority Dispatch System. Prehosp Disaster Med. 2018;33(1):29–35.

Percutaneous coronary intervention (PCI) appears to be superior to in-hospital fibrinolysis for most patients with ST-elevation myocardial infarction (STEMI). However, few hospitals have PCI capability. The optimal prehospital strategy for facilitating rapid coronary reperfusion in STEMI patients is unclear. We sought to determine whether direct transport of adult STEMI patients by emergency medical services to primary PCI centres improves 30-day all-cause mortality when compared with a strategy of transportation to the closest hospital.

We systematically searched MEDLINE, EMBASE, Cochrane “CENTRAL” database (1980-July 2007) and several other electronic databases. Two authors independently assessed citations for relevance. Two authors independently abstracted data from included studies. We included studies that, 1) transported patients directly to a PCI-capable centre for primary PCI, 2) had a control group that was transported to the closest hospital and 3) reported outcomes of treatment time intervals, all-cause mortality, reinfarction rate, stroke rate or the frequency of cardiogenic shock. We used a random effects model to provide pooled estimates of relative risk (RR) when data allowed.

We identified 2264 citations with the search. Five studies, including 980 STEMI patients, met inclusion criteria, and were clinically heterogeneous and of variable quality. Most studies were European (3/5) and involved physician out-of-hospital care providers. There was a trend toward increased survival with direct transport to primary PCI but this was not statistically significant (RR 0.51, 95% confidence interval [CI] 0.24–1.10). One study reported nonsignificant reductions in reinfarction (RR 0.43, 95% CI 0.11–1.60) and stroke (RR 0.33, 95% CI 0.01–8.06) with direct transport for primary PCI.

There is insufficient evidence to support the effectiveness of direct transport of patients with STEMI for primary PCI when compared with transportation to the closest hospital.