A review of artificial intelligence in healthcare supply chains: untapped potential?

Abstract

Supply chain management (SCM) is a central function of operational strategy and provides assurance of continuous delivery of services and products across different industries. While Artificial Intelligence (AI) has transformed SCM in other industries through predictive analytics, its application within healthcare remains poorly defined. We conducted a systematic review to analyse the extent of AI adoption, the specific methodologies employed, and the empirical evidence of its impact on healthcare SCM. We performed a systematic review following Cochrane guidelines across MEDLINE (PubMed), LILACS, and SciELO databases up to May 2024. We screened for studies describing AI applications in healthcare SCM. Because of the heterogeneity of the included studies, we applied a narrative synthesis framework. Quality of reporting and evidence maturity were assessed using the STROBE tool. The protocol was registered in PROSPERO (CRD42024528856). From 8,697 initial records, only 13 met the inclusion criteria. The narrative synthesis revealed a significant conceptual gap: most “AI” applications (N = 6) relied on theoretical fuzzy multi-criteria decision making (MCDM) and optimisation logic, rather than data-driven machine learning techniques common in other sectors. The literature is concentrated on supplier selection (N = 4) and logistics simulation. Crucially, none of the studies reported application in real-world settings. STROBE assessment indicated a low maturity of evidence, with consistent lack of reporting on bias and external validity. The integration of AI in healthcare SCM is in a theoretical stage. There is a marked disconnect between the computational potential of AI and its actual deployment. Current research is dominated by mathematical simulations of supplier selection rather than operational implementation. The lack of empirical evidence and the reliance on theoretical models highlight a significant “untapped potential” that requires a shift from simulation to real-world validation.

Ähnliche Arbeiten