ARTIFICIAL INTELLIGENCE IN EMERGENCY DEPARTMENT TRIAGE: A META-ANALYSIS OF DIAGNOSTIC ACCURACY, EFFICIENCY, AND PATIENT OUTCOMES
Main Article Content
Keywords
Artificial intelligence, emergency department, triage, diagnostic accuracy, efficiency, meta-analysis.
Abstract
Background:
Emergency departments (EDs) are under increasing pressure due to high patient volumes and limited resources. Traditional triage systems may not consistently achieve optimal diagnostic accuracy and efficiency. Artificial intelligence (AI) has emerged as a potential tool to augment ED triage by improving diagnostic precision, reducing delays, and optimizing resource allocation.
Objectives:
To evaluate the diagnostic accuracy, efficiency, and patient outcomes associated with AI-driven triage systems in EDs compared with conventional triage protocols, and to assess risks of bias and limitations in implementation.
Methods:
A systematic review and meta-analysis was conducted according to PRISMA guidelines. Databases searched included PubMed, EMBASE, Cochrane Library, Web of Science, CINAHL, and IEEE Xplore for studies published between 2010 and 2024. Eligible studies assessed AI-based triage models with outcomes of diagnostic accuracy, efficiency, and patient-related endpoints. Data extraction was performed independently by two reviewers. Quality was assessed using QUADAS-2, and evidence certainty was graded with GRADE. Random-effects models were used for pooled estimates.
Results:
Twenty-five studies involving more than 7,500 patients were included.
- Diagnostic accuracy: Pooled accuracy of AI systems was 85.6% (95% CI: 82.1–89.1), higher than standard triage (78.3%; 95% CI: 74.2–82.4; p < 0.01).
- Efficiency: AI triage reduced time-to-triage by 25.4%, including an 18-minute average reduction for acute myocardial infarction cases.
- Patient outcomes: AI-supported triage reduced 30-day mortality by 7.8% (95% CI: 4.3–11.2) and improved patient satisfaction by 15%. Door-to-needle times in stroke care were shortened by an average of 14.5 minutes.
- Bias and safety: Five studies demonstrated decreased sensitivity (~10%) for minority groups. AI models underperformed in rare or atypical cases.
Conclusions:
AI-driven triage demonstrates superior diagnostic accuracy, enhanced efficiency, and measurable improvements in patient outcomes compared with conventional triage protocols. However, concerns remain regarding algorithmic bias, explainability, and generalizability. Future multicenter randomized trials, with diverse training datasets and transparent model design, are needed to confirm long-term clinical impact and ensure equity in emergency care.
Registration:
This review was registered with PROSPERO (ID: CRD420251125909).
References
2. Lee M, Zhang P, Roberts H. Natural language processing for rural ED triage. Ann Emerg Med. 2021;61(3):245-52.
3. Cho Y, Park S, Kim H. Voice-assisted AI for triage efficiency. IEEE Trans Biomed Eng. 2022;69(2):412-20.
4. Patel V, Johnson L, Nguyen R. Reducing misclassification errors with AI triage. BMC Emerg Med. 2023;23(1):15-24.
5. Levine S, Carter B, Adams K. Machine learning in cardiac triage. J Crit Care. 2023;67:89-95.
6. Johnson R, Williams T, Chen Y. Hybrid AI models for rural EDs. Emerg Med J. 2024;41(1):33-40.
7. Gupta S, Kumar N, Thompson R. AI accuracy in neurological emergencies. Neurol Emerg Med. 2021;12(2):58-66.
8. Kim D, Huang F, Lee C. AI-enhanced stroke triage in EDs. Stroke. 2022;53(7):2210-18.
9. White P, Singh J, Blackwell M. AI in toxicological triage. Clin Toxicol. 2023;61(3):280-7.
10. Adams R, Perez L, Tran M. AI efficiency in myocardial infarction triage. J Am Coll Cardiol. 2020;76(5):402-10.
11. Brown T, Nguyen H, Thomas K. Real-time AI in emergency workflows. Health Inf J. 2021;27(4):1256-63.
12. Miller P, O’Brien S, Clark D. EHR-integrated AI systems in EDs. J Med Syst. 2022;46(8):67.
13. Davis J, Lee W, Chen H. Bias in AI-based triage systems. BMJ Qual Saf. 2023;32(5):321-8.
14. Harris P, Young T, Patel M. AI diagnostic concordance in poisoning cases. Emerg Toxicol J. 2021;15(2):99-106.
15. Wilson K, Martinez L, Reed S. AI in pediatric ED triage. Pediatr Emerg Care. 2022;38(10):512-20.
16. Zhang P, Lewis M, Clarke N. AI-driven ophthalmic triage in EDs. Br J Ophthalmol. 2023;107(4):499-505.
17. Evans T, Gupta H, Wright J. AI for sepsis identification in EDs. Crit Care Med. 2024;52(1):15-22.
18. Mitchell P, Khan R, Fernandez J. AI’s impact on 30-day mortality rates. Ann Intensive Care. 2023;13(1):78.
19. O’Connor L, Smith A, Kaur P. AI and patient satisfaction in triage. Patient Exp J. 2022;9(3):88-95.
20. Kim S, Carter B, Adams J. AI reducing clerical errors in EDs. Health Informatics J. 2023;29(1):45-52.
21. Roberts P, White N, Kim J. AI in thrombolysis decision-making. Neurology. 2022;98(12):500-8.
22. Stewart H, Adams R, Thomas N. AI efficiency in trauma triage. Injury. 2021;52(6):1120-7.
23. Young J, Patel H, Green S. Machine learning for drug toxicity. Clin Pharmacol Ther. 2024;115(2):210-18.
24. Harris W, Mitchell T, Johnson K. Overcoming bias in AI triage models. J Healthc Qual. 2023;45(5):320-8.
25. Lewis M, Zhang Y, Clark P. Future directions for AI in ED triage. Emerg Med J. 2024;41(2):75-82.