The role of artificial intelligence in virtual emergency care: a systematic review
Clinical Snapshot
PICO Framework
| P — Population | Patients presenting to or managed within virtual emergency care settings (including telehealth-based triage, remote emergency consultations, and digital emergency platforms) |
| I — Intervention | Artificial intelligence applications (including decision trees, machine learning ensembles, and graph neural networks) integrated into virtual emergency care workflows |
| C — Comparator | Standard virtual emergency care without AI augmentation, or alternative AI approaches (comparators varied across included studies; several studies lacked explicit comparator arms) |
| O — Outcomes | Triage accuracy, diagnostic performance (accuracy, sensitivity, specificity), resource efficiency, and — notably absent — actual clinical patient outcomes (mortality, morbidity, time-to-treatment, patient safety events) |
Bottom Line
This systematic review — the first to comprehensively evaluate AI in virtual emergency care — searched eight databases and identified only eight eligible studies from 7,098 records, collectively enrolling approximately 500,000 participants. Despite this large aggregate sample, the evidence base is critically immature. AI performance varied dramatically (accuracy 77.5–100%, sensitivity 63–100%), all clinical studies carried serious risk of bias per PROBAST+AI, and reporting quality was poor (mean TRIPOD+AI compliance 36.9%). Most critically, no included study measured actual clinical outcomes — mortality, morbidity, time-to-treatment, or patient safety — relying instead on surrogate technical metrics. GRADE certainty was very low to low across all outcomes. The authors appropriately conclude that current evidence does not support widespread clinical implementation. For Australian clinicians and health services considering AI-augmented virtual emergency care, this review signals that enthusiasm must be tempered by rigorous evaluation. TGA SaMD regulatory requirements and RACGP telehealth safety standards cannot be met by the existing evidence. Future research must prioritise prospective controlled trials with external validation, real patient data, and patient-centred outcome measurement before AI tools can be responsibly integrated into virtual emergency workflows.
Key Findings
P Value: Not applicable — no inferential statistics reported at the review level; narrative synthesis only
Effect Size: Accuracy range: 77.5–100%; Sensitivity range: 63–100%; Specificity: 60% (single study only). TRIPOD+AI reporting compliance: mean 36.9% (range 30.9–48.1%). All clinical studies rated as having serious risk of bias. GRADE certainty: very low to low across all outcomes.
Primary Outcome: AI diagnostic and triage performance in virtual emergency care settings, assessed via technical accuracy metrics across eight included studies (total N ≈ 500,000 participants)
Nnt Or Sensitivity: Sensitivity range 63–100% (highly variable across studies and AI modalities); specificity available from one study only (60%) — insufficient data to calculate clinically meaningful likelihood ratios, predictive values, or NNT equivalents. The wide performance ranges preclude reliable diagnostic accuracy estimation.
Confidence Interval: Not reported — no pooled estimates or confidence intervals around performance metrics were calculable given study heterogeneity and narrative synthesis design
Clinical Application
Clinical implementation is not currently feasible based on this evidence. The review identifies critical gaps: no prospective controlled trials, no clinical outcome data, poor reporting quality (mean TRIPOD+AI compliance 36.9%), and serious risk of bias in all clinical studies. Any deployment of AI in virtual emergency care would require prospective validation with real patient data, regulatory approval, and clinical outcome monitoring prior to routine use. Virtual emergency care is an expanding domain in Australia, accelerated by the COVID-19 pandemic and supported by Medicare Benefits Schedule (MBS) telehealth item numbers introduced and extended since 2020. The Australian Digital Health Agency's National Digital Health Strategy 2023–2028 explicitly supports AI integration in healthcare, but the TGA's regulatory framework for Software as a Medical Device (SaMD) — aligned with international IMDRF guidance — requires demonstrated clinical safety and efficacy before deployment. This review's finding of very low to low GRADE certainty evidence means no currently available AI virtual emergency care tool would meet TGA SaMD evidence standards for high-risk clinical decision support. RACGP telehealth guidelines emphasise clinical safety and appropriate triage; the absence of clinical outcome data in this literature means AI tools cannot yet be recommended within RACGP-compliant virtual emergency workflows. Australian emergency medicine bodies (ACEM) have not yet issued specific guidance on AI-augmented virtual emergency triage, and this review reinforces that such guidance would be premature. The Singapore-based author affiliations are relevant — while the healthcare context differs from Australia, the virtual care challenges (workforce distribution, after-hours access, rural and remote populations) have direct parallels to Australian healthcare delivery challenges. Patients accessing emergency care through virtual or telehealth platforms, including those triaged remotely, managed via digital consultation services, or assessed through AI-augmented emergency decision support systems. The approximately 500,000-participant aggregate sample suggests potential generalisability, but the serious risk of bias in all clinical studies and absence of external validation severely limits confidence in applying these findings to any specific patient population.
Abstract
BACKGROUND: The integration of artificial intelligence (AI) into virtual emergency care represents a potentially transformative approach to healthcare delivery, yet the evidence base remains poorly characterized. This systematic review comprehensively evaluates the current state of AI applications in virtual emergency care settings. METHODS: We systematically searched eight databases (Embase, PsycINFO, MEDLINE, PubMed, Scopus, Web of Science, CINAHL, Cochrane Library) from inception through March 2025. Of 7,098 records identified and 4,935 screened after deduplication using Covidence, 8 studies met inclusion criteria following exclusion of one study lacking AI components. Studies were assessed using PROBAST + AI for risk of bias and quality assessment, TRIPOD + AI for reporting quality, and GRADE for certainty of evidence. RESULTS: The eight included studies (total participants: approximately 0.5 million) evaluated diverse AI applications including decision trees, machine learning ensembles, and graph neural networks across multiple virtual emergency contexts. Performance varied widely (accuracy 77.5-100%, sensitivity 63-100%, specificity 60% in single study reporting). All clinical studies demonstrated serious risk of bias. TRIPOD + AI compliance averaged only 36.9% (range 30.9-48.1%). GRADE assessment revealed very low to low certainty evidence across all outcomes, with no studies measuring actual clinical outcomes. CONCLUSIONS: Current evidence is insufficient to support widespread clinical implementation of AI in virtual emergency care. While preliminary results suggest potential benefits in triage accuracy and resource efficiency, critical gaps exist in validation, clinical outcome assessment, and reporting standards. Future research must prioritize prospective controlled trials with real patient data, clinical outcome measurements, and adherence to reporting guidelines.
References
- 1.Shankar, R., Wang, L., Hoe, H. S., Yee, I. L., Fong, L. M., Kumar Gollamudi, S. P., Wong, T. C., & Wong, S. (2026). The role of artificial intelligence in virtual emergency care: a systematic review. International Journal of Medical Informatics. https://doi.org/10.1016/j.ijmedinf.2026.106411
Related Research
Journal of critical care
Artificial intelligence and computerized decision support in adult intensive care: A systematic review of randomized controlled trials
3 Aug 2026
Journal of pediatric surgery
Interpretable deep learning model for pediatric strangulated small bowel obstruction on CT: A multicenter study
3 Aug 2026
Telemedicine journal and e-health : the official journal of the American Telemedicine Association
Application of Telehealth and Artificial Intelligence in Military Health Care: A Promising Combination for the Future
3 Aug 2026
This content is for educational purposes for healthcare professionals only and does not constitute clinical advice. Clinical decisions should be based on individual patient assessment, current guidelines, and appropriate specialist consultation. Editorial Standards · Privacy Policy · Terms of Service