What is the clinical evidence from this study?

Study design: Systematic Review. Population: Acute coronary syndrome (ACS) (n=319709). Intervention: Artificial intelligence (AI) models vs. Standard clinical assessment. Primary outcome: Diagnostic performance for ACS/STEMI.

What does this research mean for the field?

Artificial intelligence models, particularly ECG-based deep learning and multimodal approaches, demonstrate strong diagnostic and risk prediction performance for the prehospital assessment of acute coronary syndrome, though current evidence is limited by methodological constraints. Novelty: ClaimNovelty.SYNTHESIS. Consensus alignment: ConsensusAlignment.NEUTRAL.

What question did this study set out to answer?

To explore the current evidence on AI applications for prehospital assessment of suspected acute coronary syndrome and identify knowledge gaps.

June 8, 2026BMC Emergency Medicine0 citationsOpen Access

Artificial intelligence in prehospital assessment of acute coronary syndrome: a scoping review

KKKimia KhonakdarMazandaran University of Medical Sciences RFReyhaneh Sadat FaghaniMazandaran University of Medical Sciences FKFatemeh KeyhanianMazandaran University of Medical Sciences

Key Result

Artificial intelligence models demonstrated promising diagnostic performance for prehospital acute coronary syndrome assessment, achieving AUC values between 0.81 and 0.99.

Key Points

To explore the current evidence on AI applications for prehospital assessment of suspected acute coronary syndrome and identify knowledge gaps.
Conducted a scoping review following PRISMA-ScR guidelines.
Performed a comprehensive literature search up to May 2026.
Included 19 studies evaluating AI-based models using prehospital data for ACS diagnosis, prediction, or triage.
AI models displayed diagnostic performance with AUC values from ~0.81 to 0.99, sensitivity from 73% to 94%, and specificity from 56% to 99%.
Risk prediction models had moderate to good performance (AUC ~ 0.71–0.95) with variability.
Emerging applications included risk stratification and decision support, such as predicting cardiogenic shock and revascularization needs.

Study Design

Type

Systematic Review (n=319,709)

Structured PICO

Do artificial intelligence-based models improve the prehospital diagnosis, risk stratification, and triage of suspected acute coronary syndrome?

Population

319,709 adult patients with suspected or confirmed acute coronary syndrome assessed in prehospital settings across 19 included studies.

Intervention

Artificial intelligence (AI)-based models, including ECG-based deep learning and multimodal approaches, for prehospital assessment

Outcome

Diagnostic performance (AUC, sensitivity, specificity) for ACS diagnosis, prediction, or triagesurrogate

AI-based models, particularly ECG-based deep learning and multimodal approaches, show considerable potential to improve the prehospital diagnosis and triage of suspected acute coronary syndrome.

Limitations

Most available studies are retrospective and highly heterogeneous
Considerable variation in model performance across studies
Limited real-time assessment in clinical settings
Current evidence remains limited by methodological constraints
Demographic reporting was limited and heterogeneous across studies

Abstract

Acute coronary syndrome (ACS) remains a leading cause of morbidity and mortality worldwide, where timely diagnosis is critical. Prehospital assessment is challenging due to limited diagnostic resources and high clinical uncertainty. Artificial intelligence (AI) has emerged as a potential tool to support early diagnosis, risk stratification, and triage. This scoping review aimed to map the current evidence on AI applications in the prehospital assessment of suspected ACS and to identify existing knowledge gaps. A scoping review was conducted in accordance with PRISMA-ScR guidelines. A comprehensive search of PubMed, Scopus, Web of Science, and Embase was performed up to May 2026. Studies evaluating AI-based models using prehospital data for ACS diagnosis, prediction, or triage were included. 19 studies involving 319,709 patients were included. AI-based models, particularly ECG-based deep learning and multimodal approaches, demonstrated promising diagnostic performance, with AUC values generally ranging from approximately 0.81 to 0.99, sensitivity from 73% to 94%, and specificity from 56% to 99%. These models improved sensitivity, reduced diagnostic variability, and enhanced triage efficiency in some settings. Risk prediction models showed moderate to good performance (AUC ~ 0.71–0.95) but were more variable. Emerging applications extended beyond diagnosis to risk stratification and decision support, including prediction of cardiogenic shock and need for revascularization. AI shows considerable potential to improve prehospital assessment of suspected ACS, particularly through enhanced ECG interpretation and multimodal data integration. However, current evidence remains limited by methodological constraints. Future prospective, multicenter studies with standardized approaches are essential to support clinical implementation.

Ask AI

Helpful

Bookmark

View Full Paper