A practical framework for appropriate implementation and review of artificial intelligence (FAIR-AI) in healthcare
Open accessPublished version of a preprint
Abstract
Health systems face the challenge of balancing innovation and safety to responsibly implement artificial intelligence (AI) solutions. The rapid proliferation, growing complexity, ethical considerations, and rising demand for these tools require timely and efficient processes for rigorous evaluation and ongoing monitoring. Current AI evaluation frameworks often lack the practical guidance for health systems to address these challenges. To fill this gap, we developed a prescriptive evaluation framework informed by a literature review, in-depth interviews with key stakeholders, including patients, and a multidisciplinary design workshop. The resulting framework provides health systems an outline of the resources, structures, criteria, and template documents to enable pre-implementation evaluation and post-implementation monitoring of AI solutions. Health systems will need to treat this or any alternative framework as a living document to maintain relevance and effectiveness as the AI landscape and regulations continue to evolve.
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Integrity signals
Registry
Cites
Statistics
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
Evidence from activated systems undergoing longitudinal monitoring provides the strongest basis for judging whether a system remains safe, equitable, clinically useful, and governable once embedded in everyday care.
“Guidance documents can define what should matter; deployment-proximal evaluations can show whether activation appears plausible in a target setting; however, evidence from activated systems undergoing longitudinal monitoring provides the strongest basis for judging whether a system remains safe, equitable, clinically useful, and governable once embedded in everyday care.”
From Beyond Model Development in Healthcare AI: Post-Development Robustness, Post-Deployment Monitoring, and Lifecycle Governance—A Scoping Review of Reviews (2026) · cites this paper as [50]
In Results (Framework for the appropriate implementation and review of AI; page unavailable), the paper calls for continuous monitoring of deployed systems to check benefits, emerging risks, bias, and alignment with the approved use case. The narrative review also calls for assessing clinical utility and equity. It does not establish that evidence from longitudinally monitored systems is the strongest basis for judgment relative to guidance or deployment-proximal evaluations. (LLM, conf=0.83)
Experimental — accuracy not yet measuredThese assumptions may not hold equally across low- and middle-income countries, smaller hospitals, rural systems, or institutions with limited digital maturity, where the infrastructure required for continuous monitoring, subgroup surveillance, drift detection, formal change control, and incident response may be more difficult to sustain.
“These assumptions may not hold equally across low- and middle-income countries, smaller hospitals, rural systems, or institutions with limited digital maturity, where the infrastructure required for continuous monitoring, subgroup surveillance, drift detection, formal change control, and incident response may be more difficult to sustain [49,50,51,52].”
From Beyond Model Development in Healthcare AI: Post-Development Robustness, Post-Deployment Monitoring, and Lifecycle Governance—A Scoping Review of Reviews (2026) · cites this paper as [50]
Split into 2 atoms — 0 confirmed, 2 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.
Experimental — accuracy not yet measuredThe feasibility of lifecycle governance is likely to vary substantially across health systems because many recommendations assume access to data infrastructure, informatics expertise, technical staffing, procurement capacity, legal support, regulatory maturity, and multidisciplinary oversight structures.
“The feasibility of lifecycle governance is likely to vary substantially across health systems because many recommendations assume access to data infrastructure, informatics expertise, technical staffing, procurement capacity, legal support, regulatory maturity, and multidisciplinary oversight structures [49,50,51,52].”
From Beyond Model Development in Healthcare AI: Post-Development Robustness, Post-Deployment Monitoring, and Lifecycle Governance—A Scoping Review of Reviews (2026) · cites this paper as [50]
Split into 2 atoms — 0 confirmed, 2 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.
Experimental — accuracy not yet measuredRecent implementation-oriented perspectives emphasise that safe deployment depends on structured governance, staged review, and durable post-deployment accountability rather than one-time technical approval alone.
“Recent implementation-oriented perspectives make a similar point by emphasising that safe deployment depends on structured governance, staged review, and durable post-deployment accountability rather than one-time technical approval alone [15,50,51].”
From Beyond Model Development in Healthcare AI: Post-Development Robustness, Post-Deployment Monitoring, and Lifecycle Governance—A Scoping Review of Reviews (2026) · cites this paper as [50]
Comparable emphases on lifecycle management, local governance, and organisational readiness are visible in recent clinical AI implementation and oversight frameworks and in WHO guidance on AI for health.
“Comparable emphases on lifecycle management, local governance, and organisational readiness are visible in recent clinical AI implementation and oversight frameworks, as well as in WHO guidance on AI for health [50,51,52].”
From Beyond Model Development in Healthcare AI: Post-Development Robustness, Post-Deployment Monitoring, and Lifecycle Governance—A Scoping Review of Reviews (2026) · cites this paper as [50]
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Cites
Collins, G. S., Moons, K. G. M., Dhiman, P., Riley, R. D., Beam, A. L., Van Calster, B., Ghassemi, M., Liu, X., Reitsma, J. B., van Smeden, M., et al. (2024) · BMJ · doi:10.1136/bmj-2023-078378 · Cited by 4,718 · cited 2× · see citing works
“Numerous publications and guidelines such as TRIPOD and TRIPOD-AI have described the reporting necessary to properly evaluate a risk prediction model, regardless of the underlying statistical or machine learning method 12,13 .”
Rajpurkar, P., Chen, E., Banerjee, O., & Topol, E. J. (2022) · Nature Medicine · doi:10.1038/s41591-021-01614-0 · Cited by 3,212 · see citing works
Collins, G. S., Reitsma, J. B., Altman, D. G., & Moons, K. (2015) · BMC Medicine · doi:10.1186/s12916-014-0241-z · Cited by 1,927 · see citing works
Bandi, A., Adapa, P. V. S. R., & Kuchi, Y. E. V. P. K. (2023) · Future Internet · doi:10.3390/fi15080260 · Cited by 638 · see citing works
Artificial Intelligence in Health Care: A Report From the National Academy of Medicine
Matheny, M. E., Whicher, D., & Thadaney Israni, S. (2020) · JAMA · doi:10.1001/jama.2019.21579 · Cited by 628 · see citing works
Paulus, J. K., & Kent, D. M. (2020) · npj Digital Medicine · doi:10.1038/s41746-020-0304-9 · Cited by 274 · see citing works
Liu, X., Glocker, B., McCradden, M. M., Ghassemi, M., Denniston, A. K., & Oakden-Rayner, L. (2022) · The Lancet Digital Health · doi:10.1016/s2589-7500(22)00003-6 · Cited by 240 · cited 1× · see citing works
“Assess model performance across key patient subgroups 10,30,31 .”
Evaluation framework to guide implementation of AI systems into healthcare settings
Reddy, S., Rogers, W., Makinen, V.-P., Coiera, E., Brown, P., Wenzel, M., Weicken, E., Ansari, S., Mathur, P., Casey, A., et al. (2021) · BMJ Health & Care Informatics · doi:10.1136/bmjhci-2021-100444 · Cited by 230 · see citing works
Evaluating the impact of prediction models: lessons learned, challenges, and recommendations
Kappen, T. H., van Klei, W. A., van Wolfswinkel, L., Kalkman, C. J., Vergouwe, Y., & Moons, K. G. M. (2018) · Diagnostic and Prognostic Research · doi:10.1186/s41512-018-0033-6 · Cited by 215 · see citing works
Assessing the research landscape and clinical utility of large language models: a scoping review
Park, Y.-J., Pillai, A., Deng, J., Guo, E., Gupta, M., Paget, M., & Naugler, C. (2024) · BMC Medical Informatics and Decision Making · doi:10.1186/s12911-024-02459-6 · Cited by 187 · see citing works
A framework for the oversight and local deployment of safe and high-quality prediction models
Bedoya, A. D., Economou-Zavlanos, N. J., Goldstein, B. A., Young, A., Jelovsek, J. E., O’Brien, C., Parrish, A. B., Elengold, S., Lytle, K., Balu, S., et al. (2022) · Journal of the American Medical Informatics Association · doi:10.1093/jamia/ocac078 · Cited by 119 · see citing works
Dankwa-Mullan, I., Scheufele, E. L., Matheny, M. E., Quintana, Y., Chapman, W. W., Jackson, G., & South, B. R. (2021) · Journal of Health Care for the Poor and Underserved · doi:10.1353/hpu.2021.0065 · Cited by 108 · see citing works
Race and Ethnicity: A Part of the Equation for Personalized Clinical Decision Making?
Paulus, J. K., & Kent, D. M. (2017) · Circulation: Cardiovascular Quality and Outcomes · doi:10.1161/circoutcomes.117.003823 · Cited by 35 · see citing works
Flagged first, then by citations
Research works printed but not linked: 25
- [2]Dedehayir, O. & Steinert, M. The hype cycle model: A review and future directions. Technol. Forecast. Soc. Change 108, 28–41 (2016).
- [4]Thirunavukarasu, A. J. et al. Large language models in medicine. Nat. Med. 29, 1930–1940 (2023). 37460753
- [10]Vollmer, S. et al. Machine learning and artificial intelligence research for patient benefit: 20 critical questions on transparency, replicability, ethics, and effectiveness. BMJ 368, l6927 (2020). 32198138 PMC11515850
- [11]van der Vegt, A. H. et al. Implementation frameworks for end-to-end clinical AI: derivation of the SALIENT framework. J. Am. Med. Inform. Assoc. JAMIA 30, 1503–1515 (2023). 37208863 PMC10436156
- [14]Steyerberg, E. W., et al. Assessing the performance of prediction models: A framework for some traditional and novel measures. Epidemiol. Camb. Mass 21, 128 (2010). PMC3575184 20010215
- [15]Vickers, A. J. & Elkin, E. B. Decision curve analysis: A novel method for evaluating prediction models. Med. Decis. Mak. Int. J. Soc. Med. Decis. Mak. 26, 565–574 (2006). PMC2577036 17099194
- [17]Moons, K. G. M., Altman, D. G., Vergouwe, Y. & Royston, P. Prognosis and prognostic research: application and impact of prognostic models in clinical practice. BMJ 338, b606 (2009). 19502216
- [18]Moons, K. G. M. et al. Risk prediction models: II. External validation, model updating, and impact assessment. Heart 98, 691–698 (2012). 22397946
- [21]Wiens, J. et al. Do no harm: A roadmap for responsible machine learning for health care. Nat. Med. 25, 1337–1340 (2019). 31427808
- [22]Scott, I., Carter, S. & Coiera, E. Clinician checklist for assessing suitability of machine learning applications in healthcare. BMJ Health Care Inform. 28, e100251 (2021). 33547086 PMC7871244
- [24]Osheroff, J. A. et al. A roadmap for national action on clinical decision support. J. Am. Med. Inform. Assoc. JAMIA 14, 141 (2007). 17213487 PMC2213467
- [25]Blackman, R. Ethical Machines: Your Concise Guide to Totally Unbiased, Transparent, and Respectful AI. (Harvard Business Review Press, 2022).
- [29]O’Neill, J. et al. Applying an equity lens to interventions: Using PROGRESS ensures consideration of socially stratifying factors to illuminate inequities in health. J. Clin. Epidemiol. 67, 56–64 (2014). 24189091
- [31]Vasey, B. et al. Reporting guideline for the early stage clinical evaluation of decision support systems driven by artificial intelligence: DECIDE-AI. BMJ 377, e070904 (2022). 35584845 PMC9116198
- [32]Zerilli, J., Bhatt, U. & Weller, A. How transparency modulates trust in artificial intelligence. Patterns3, 100455 (2022). PMC9023880 35465233
- [33]Hurley, M. E., Lang, B. H., Kostick-Quenet, K. M., Smith, J. N. & Blumenthal-Barby, J. Patient consent and the right to notice and explanation of AI systems used in health care. Am. J. Bioeth. AJOB 25, 102–114 (2025). 39288291 PMC12143229
- [34]Pflanzer, M. Balancing transparency and trust: Reevaluating AI disclosure in healthcare. Am. J. Bioeth. AJOB 25, 153–156 (2025). 39992851
- [35]de Hond, A. A. H. et al. Guidelines and quality criteria for artificial intelligence-based prediction models in healthcare: a scoping review. NPJ Digit. Med. 5, 2 (2022). 35013569 PMC8748878
- [36]Crossnohere, N. L., Elsaid, M., Paskett, J., Bose-Brill, S. & Bridges, J. F. P. Guidelines for artificial intelligence in medicine: Literature review and content analysis of frameworks. J. Med. Internet Res. 24, e36823 (2022). 36006692 PMC9459836
- [37]Kramer, J. et al. Developing a Framework for the Review and Oversight of Artificial Intelligence at a Large Healthcare Enterprise: Assessing the Needs and Priorities of Senior Health System Leadership, Providers, and Community Stakeholders. in (AcademyHealth, 2024).
- [40]Feng, J. et al. Clinical artificial intelligence quality improvement: towards continual monitoring and updating of AI algorithms in healthcare. Npj Digit. Med. 5, 1–9 (2022). 35641814 PMC9156743
- [41]Lu, J. et al. Learning under concept drift: A review. IEEE Trans. Knowl. Data Eng. pp. 1–1 10.1109/TKDE.2018.2876857(2018). · DOI
- [42]Gallifant, J. et al. The TRIPOD-LLM reporting guideline for studies using large language models. Nat. Med. 31, 60–69 (2025). 39779929 PMC12104976
- [43]Tam, T. Y. C. et al. A framework for human evaluation of large language models in healthcare derived from literature review. Npj Digit. Med. 7, 1–20 (2024). 39333376 PMC11437138
- [44]Campbell, S. et al. Purposive sampling: Complex or simple? Research case examples. J. Res. Nurs. JRN 25, 652 (2020). 34394687 PMC7932468
Cited by
Sorted by citations
Boussi Rahmouni, H., Hassine, N. B. E. H., Chouchen, M., Ceylan, H. İ., Muntean, R. I., Bragazzi, N. L., & Dergaa, I. (2025) · Healthcare · doi:10.3390/healthcare13202553 · Cited by 25 · see citing works
El Arab, R. A., Mustafa, M. H., Almagharbeh, W. T., Saleem, N. H., Al Abdulmohsen, S., Boathab, R., & Bu Washl, M. (2026) · Healthcare · doi:10.3390/healthcare14111459 · Cited by 4 · see citing works
“Recent implementation frameworks in clinical AI similarly emphasise phased evaluation, progressive real-world validation, and explicit post-deployment monitoring, reinforcing the importance of distinguishing pre-activation evidence from genuine operational oversight [15,20,50].”
Influential
- Authors
- Wells, B. J., Nguyen, H. M., McWilliams, A., Pallini, M., Bovi, A., Kuzma, A., Kramer, J., Chou, S.-H., Hetherington, T., Corn, P., Taylor, Y. J., Cuison, A., Gagen, M., Isreal, M., FAIR-AI Consortium, Akbilgic, O., Barr, K., Bowers, A., Caffrey, R., Carroll, M. S., CiRullo, M., Downs, S. M., Hardy, N., Heuay, J., Katzovitz, K., Kirkendall, E., Lindgren, E., Lonergan, L., McKinley, E., Pajewski, N. M., Sak-Castellano, L., Setliff, E., & Wright, G.
- Affiliations
- Wake Forest University Health Sciences (US)
- Advocate Health Care (US)
- Corresponding author
- Nguyen, Hieu M.
- ORCID
- Venue
- npj Digital Medicine
- Published
- 2025-08-11
- Volume
- 8
- Issue
- 1
- Pages
- 514-514
- ISSN
- 2398-6352 (ISSN-L)
- PMC
- PMC12340025
- OpenAlex
- W4413107286
- Cited by
- 90
- References
- 13
- Citations by year
- 2026: 80 · 2025: 10
- FWCI
- 21.66
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Locations
- npj Digital Medicine (publishedVersion, cc-by)
- Research Square (acceptedVersion, cc-by)
- PubMed Central (publishedVersion, cc-by)
- PubMed (publishedVersion)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
- Funders
- Duke Endowment (AWD00002292)
- Topics
- Health Sciences › Medicine › Health Informatics › Artificial Intelligence in Healthcare and Education
- Social Sciences › Safety Research › Ethics and Social Impacts of AI
- Physical Sciences › Engineering › Safety, Risk, Reliability and Quality › Safety Systems Engineering in Autonomy
- Keywords
- artificial intelligence in healthcare, algorithmic fairness, AI evaluation framework, pre-implementation evaluation, post-implementation monitoring, health system implementation, AI ethics, AI safety, stakeholder engagement, living document
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
