Presentation (CTI)

OBJECTIVES: AI-enabled medical devices are subject to mandatory post-market monitoring under EU regulatory frameworks, and continuous monitoring is essential for detecting biases, performance degradation, or unintended harms, representing both a regulatory mandate and a moral imperative. Furthermore, effective interventions can alter the incidence of predicted outcomes, creating feedback loops that confound performance metrics, requiring continuous clinical evaluation. Yet the costs of fulfilling these obligations are insufficiently accounted for in health technology assessment (HTA). This review aims to document the prevalence and nature of post-market monitoring cost omissions in economic evaluations of AI-enabled medical devices.
METHODS: A PRISMA systematic review was conducted (MEDLINE, EMBASE, Cochrane Library, Web of Science) for economic evaluations of AI-enabled medical devices published between 2020 and 2025. Grey literature (regulatory guidance documents, HTA agency reports, and EU/EEA policy frameworks governing post-market surveillance), was searched in parallel. Inclusion criteria: reported costs of post-market monitoring requirements for AI-enabled devices in clinical settings. Data extraction: presence, absence, or partial reporting of post-market monitoring cost components, and post-deployment performance estimation.
RESULTS: Preliminary findings indicate that lifecycle costs, including personnel time for surveillance, data infrastructure, model auditing, and organizational overhead for managing detected performance drift or bias, are inconsistently reported and contribute to a systematic overestimation of cost-effectiveness of clinical AI. The literature reveals an expectation of continuous post-market monitoring that generates substantial institutional costs at the point of care, currently insufficiently accounted for in HTA cost estimates.
CONCLUSIONS: A disconnect exists between post-market monitoring obligations under current regulatory frameworks, including MDR 2017/745 and the AI Act, and real-world deployment cost estimates for AI-enabled medical devices. Future evaluations must adopt a standardized cost framework not only for acquisition and implementation, but also for monitoring and maintenance costs, to ensure investments in clinical AI deliver sustainable value to patients and healthcare systems.