JMIR Publications

ABSTRACT

Background:

Healthcare supply chains face persistent challenges such as information asymmetry, fragmented coordination, and limited technological integration, vulnerabilities starkly exposed during the COVID-19 pandemic. While artificial intelligence (AI) has shown promise in clinical applications, its use in healthcare supply chain management remains understudied.

Objective:

This systematic review examined the extent to which AI implementation in healthcare supply chain management (SCM) has been evaluated in the peer-reviewed literature, focusing on: (1) types of AI tools reported as implemented in real-world settings, (2) the degree to which implementation science frameworks were applied in these deployments, and (3) operational outcomes reported following implementation.

Methods:

Following PRISMA 2020 guidelines, three electronic databases (PubMed, Scopus, and Web of Science) were searched on April 9, 2024, using keywords related to implementation science, artificial intelligence, healthcare, and supply chain management. Searches were limited to English-language, peer-reviewed articles published between 2009 and 2024. Six reviewers independently screened 5,499 unique records using predefined inclusion and exclusion criteria. Studies were included only if they documented actual AI implementation beyond pre-implementation modeling or simulation. Data extraction focused on study characteristics, AI implementation contexts, supply chain domains, implementation science framework use, and reported outcomes.

Results:

From 5,499 initial unique records, 54 proceeded to full-text review; only 3 met final inclusion criteria, a 99.95% exclusion rate, revealing a fundamental gap between widespread industry AI adoption and rigorous research on the implementation of AI across supply chain functions. The three included studies reported on AI implementation across distinct supply chain functions: laboratory specimen transport optimization (genetic algorithms achieving 20-30% cost savings while maintaining ISO quality standards), acute stroke care coordination (machine learning-enabled platform reducing door-to-treatment times by 32-39% and communication burden by 30%), and integrated smart hospital operations (comprehensive AI platform supporting >12,000 daily uses with sub-second response times). While all implementations demonstrated measurable operational improvements, none employed formal implementation science frameworks (e.g., CFIR, RE-AIM, NASSS) to guide planning or evaluation, and follow-up periods were limited to six months or less in the studies that reported them.

Conclusions:

This review reveals a critical paradox: despite widespread industry AI adoption in the healthcare supply chain, the implementation evidence is absent. The lack of implementation research shows more than a methodological gap; it signals substantial risk for healthcare organizations looking to implement AI without evidence-based guidance on the implementation process, organizational prerequisites, or sustainability factors. Future research must prioritize implementation science approaches, longitudinal sustainability assessment, and evaluation of downstream patient outcomes. Interdisciplinary collaboration between engineers, healthcare managers, and implementation scientists is essential to transform AI from a promising concept into an equitable, sustainable component of healthcare supply chain operations.