Beyond Validation: Operationalising Post-Deployment Surveillance of AI Medical Devices in Clinical Practice
Clinical Snapshot
PICO Framework
| P — Population | Healthcare institutions deploying artificial intelligence medical devices in clinical practice |
| I — Intervention | A structured, decision-oriented framework for post-deployment surveillance and monitoring of AI medical devices |
| C — Comparator | No comparator explicitly defined; implicitly compared to existing ad hoc or undefined post-deployment monitoring approaches |
| O — Outcomes | Effective performance assessment, governance-linked corrective action, safer and more accountable integration of AI into routine clinical care |
Bottom Line
This letter proposes a decision-oriented framework for post-deployment surveillance of AI medical devices, drawing on deployment experience at a UK NHS tertiary centre and a commercial AI vendor. The paper identifies a genuine and pressing governance gap — the absence of standardised post-market monitoring for AI diagnostics in clinical settings — and offers a pragmatic conceptual structure for addressing it. However, clinicians and governance leaders should interpret this contribution with appropriate caution. It is expert opinion at the base of the evidence hierarchy, with no empirical validation, no patient outcome data, and a notable conflict of interest given three authors' commercial affiliations with Qure.AI. The framework's generalisability beyond radiology AI in well-resourced NHS settings is unproven. For Australian institutions, the TGA's evolving SaMD regulatory framework provides a more authoritative governance anchor, and this paper may serve as a useful supplementary resource for operationalising surveillance processes rather than a standalone evidence base. Institutions should not adopt this framework uncritically but may find its decision-oriented structure a useful starting point for locally adapted, governance-approved surveillance protocols. Independent empirical evaluation of the framework's clinical impact is urgently needed.
Key Findings
P Value: Not reported
Effect Size: Not applicable — no quantitative outcomes reported
Primary Outcome: Proposal of a structured, decision-oriented post-deployment surveillance framework for AI medical devices in clinical institutions
Nnt Or Sensitivity: Not applicable — no diagnostic, therapeutic, or prognostic performance data presented
Confidence Interval: Not reported
Clinical Application
The decision-oriented framework is conceptually accessible and does not require specialist epidemiological expertise to implement. However, operationalisation would require dedicated governance infrastructure, data linkage capabilities, and ongoing clinical champion engagement — resources that may be limited in smaller or regional healthcare facilities. The framework's feasibility in low-resource settings is not addressed. This paper is directly relevant to Australian healthcare given the TGA's increasing regulatory activity around Software as a Medical Device (SaMD) and AI-enabled medical devices under the Therapeutic Goods (Medical Devices) Regulations 2002. The TGA's 2023 guidance on SaMD and its alignment with the International Medical Device Regulators Forum (IMDRF) framework for AI/ML-based SaMD creates a regulatory imperative for post-market surveillance that this paper partially addresses. Australian public hospitals deploying AI diagnostic tools (e.g., AI-assisted chest X-ray triage, AI colonoscopy aids) would benefit from structured surveillance protocols. The RACGP and relevant specialist colleges have not yet issued specific guidance on AI medical device monitoring in primary or specialist care. PBS implications are indirect but relevant where AI devices influence diagnostic pathways that trigger subsidised investigations or treatments. State health departments (e.g., NSW Health, Victorian DHHS) with active AI procurement programs represent the most immediate implementation context. Hospital clinical governance committees, radiology departments, health informatics teams, and clinical safety officers at institutions that have deployed or are planning to deploy AI medical devices in diagnostic or clinical decision support roles
Abstract
Artificial intelligence medical devices are increasingly deployed in clinical practice, yet practical approaches to post-deployment monitoring remain poorly defined. We present a structured, decision-oriented approach to monitoring within healthcare institutions, grounded in our own deployment experience. By framing surveillance as a set of interdependent decisions, this model supports effective performance assessment and governance-linked corrective action, enabling safer and more accountable integration of AI into routine clinical care.
References
- 1.Lampignano, J. A. P., Kale, A., Kumar, S., Shah, D., Reddy, B., & Lowe, D. J. (2026). Beyond validation: Operationalising post-deployment surveillance of AI medical devices in clinical practice. Journal of Medical Systems. https://doi.org/10.1038/s41746-022-00611-y
Related Research
Neural networks : the official journal of the International Neural Network Society
MFBD: Model-free backdoor defense based on vision-language pre-trained models
2 Aug 2026
Balkan medical journal
Bias and Fairness Across the Healthcare AI Lifecycle: A Clinician-Oriented Review
1 Aug 2026
Proceedings of the National Academy of Sciences of the United States of America
The backfiring effect of weak AI safety regulation
28 July 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