EUVIMEDEUVIMED entdecken
Systematische ÜbersichtMetaanalyse

Artificial Intelligence–Based Predictive Models for Operational Performance and Failure Risk Assessment in Healthcare Systems: A Systematic Review and Meta-Analysis

Anil Rajbhandari

SSN Journal of Management & Technology Research Communications · 2025

DOI: 10.21786/mntrc/2.3.2 ↗

Kurz gefasst

Systematische Übersicht in der Zeitschrift SSN Journal of Management & Technology Research Communications, 2025. Verfasst von Anil Rajbhandari.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Healthcare systems operate as high-reliability organisations, where consistent operational performance is essential for patient safety, quality of care, and financial sustainability. Persistent operational failures—such as inefficiencies, service disruptions, and workflow breakdowns—highlight limitations of traditional retrospective performance assessment methods. Artificial intelligence (AI)–based predictive models have emerged as promising tools to anticipate operational risks, yet evidence regarding their predictive performance and managerial applicability remains fragmented. This systematic review and meta-analysis aimed to evaluate the predictive accuracy, methodological robustness, and managerial relevance of AI-based models used for operational performance and failure-risk assessment in healthcare systems. A systematic search of PubMed/MEDLINE, Embase, Scopus, Web of Science, Cochrane CENTRAL, and IEEE Xplore was conducted in accordance with PRISMA 2020 guidelines. Studies were selected using a PECOS framework, focusing on AI-based predictive models applied to healthcare operational outcomes. Risk of bias was assessed using the QUADAS-AI tool. Random-effects meta-analyses were performed to pool predictive performance metrics, including area under the receiver operating characteristic curve (AUC), accuracy, precision, recall, and F1-score.Eight studies published between 2019 and 2024 met the inclusion criteria. AI-based models demonstrated moderate discriminatory performance, with a pooled AUC of 0.646 (95% CI: 0.563–0.721) and substantial heterogeneity (I² = 86.8%). Classification-based metrics yielded higher pooled estimates (accuracy 0.814; F1-score 0.825) but exhibited pronounced heterogeneity. Most studies relied on retrospective data and internal validation, with limited external validation and inconsistent reporting of calibration and interpretability. Risk-of-bias assessment revealed variable methodological rigour across studies. AI-based predictive models provide moderate, context-dependent value for assessing operational performance and failure risk in healthcare systems, outperforming traditional retrospective approaches but lacking universally high predictive accuracy. Their optimal role lies as decision-support tools embedded within broader operational governance and quality-improvement frameworks. Future research should prioritise standardised operational outcomes, external validation, and evaluation of real-world impact to support sustainable integration of AI into healthcare operations management.

Zitieren

APA

Rajbhandari, A. (2025). Artificial Intelligence–Based Predictive Models for Operational Performance and Failure Risk Assessment in Healthcare Systems: A Systematic Review and Meta-Analysis. SSN Journal of Management & Technology Research Communications. https://doi.org/10.21786/mntrc/2.3.2

Vancouver

Rajbhandari A. Artificial Intelligence–Based Predictive Models for Operational Performance and Failure Risk Assessment in Healthcare Systems: A Systematic Review and Meta-Analysis. SSN Journal of Management & Technology Research Communications. 2025. doi:10.21786/mntrc/2.3.2

BibTeX
@article{Rajbhandari2025,
  author = {Rajbhandari, Anil},
  title = {Artificial Intelligence–Based Predictive Models for Operational Performance and Failure Risk Assessment in Healthcare Systems: A Systematic Review and Meta-Analysis},
  journal = {SSN Journal of Management & Technology Research Communications},
  year = {2025},
  doi = {10.21786/mntrc/2.3.2}
}
Kostenlos starten →