Table 2.

Use of health care standards in the reviews mapped to the life cycle phases by van de Sande et al [23].

			Standards and corresponding reviewsa
Life cycle phase 0: preparation before AIb model development
	Define the problem and engage stakeholders	Do no harm road map (Wiens)
	Search for and evaluate available models	FDAc devices (Benjamens) ÉCLAIRd (Omoumi)
	Identify, collect data, and account for bias	FHIRe (Mandel) FAIRf (Wilkinson) Validation (Riley) PROBASTg (Moons and Wolff): Adamidi et al [37]a, Barboi et al [38], Abbasgholizadeh Rahimi et al [58], and Sahu et al [59]
	Handle privacy	HIPAAh (OOTAi): Battineni et al [39] GDPRj (EUk)
	Ethical principles, frameworks, and guidelines	WMAl (Helsinki declaration): Abd-Alrazaq et al [36]a Ethics for mental health technology (WEFm): Zidaru et al [68] Digital disease technology detection (SORMASn): Zhao et al [13]a Asadi framework of ethics (Asadi): Zhao et al [15]a Ethical guidelines (EU): Chew and Achananuparp [44] Ethical principles and framework (Montreal): Mörch et al [55] Ethics and governance of AI health (WHOo): Zidaru et al [68]
Life cycle phase I: AI model development
	Check applicable regulations	Proposed regulatory framework (FDA) Harmonized rules on AI (EU)
	Prepare data	Preprocessing data (Ferrao)
	Train and validate	MLp cardiac imaging (Juarez-Orozco)
	Evaluate performance and report results	AI guide (Park and Han): Mörch et al [55] TRIPODq (Collins): Adamidi et al [37] TRIPOD-MLr (Collins): Hassan et al [50] and Mörch et al [55] CLAIMs (Mongan): Shelmerdine et al [15]t CHARMSu (Moons): Barboi et al [38] and Sahu et al [59] PRISMA-DTAv (McInnes): Shelmerdine et al [15]t MI-CLAIMw (Norgeot): Shelmerdine et al [15]t MINIMARx (Hernandez-Boussard): Shelmerdine et al [15]t NOSy (Lo): Hassan et al [50] and Battineni et al [39] LOEz (Concato): Seibert et al [61]a MMATaa (Hong): Zhao et al [66]a Clinical Prediction Rule Checklist (CASPab): Bertini et al [40] STARDac (checklist): Zhao et al [66]a COREQad (checklist): Buchanan et al [42]a, Kaelin et al [52]a, and Seibert et al [61]a
Life cycle phase II: assessment of AI performance and reliability
	Externally validate model or concept	Validation (Ramspek and Riley) Generalizability (Futoma) Risk of bias (?): Talpur et al [63] MADE1.0ae (Dandala): Choudhury and Asan [46]a
	Simulate results and prepare for clinical study	DECIDE-AIaf (steering group): Shelmerdine et al [15]t
Life cycle phase III: clinically testing AI
	Design and conduct clinical study	SPIRIT-AIag (Cruz): Mörch et al [55] Explanations (Barda) CONSORT-AIah (Liu): Loveys et al [54] Revised Cochrane RoB 2ai (Sterne): Loveys et al [54] ROBINS-Iaj (RoB 2 for non-RCTsak; Sterne): Loveys et al [54] STROBEal (checklists): Adamidi et al [37]a
Life cycle phase IV: implementing and governing AI
	Legal approval	AI-MLam medical devices (Muehlematter)
	Safely implement model	TAMan (Jauk) ML model facts (Sendak)
	Model and data governance	FAIR (Wilkinson): Adamidi et al [37]a SaMDao clinical evaluation (FDA): Adamidi et al [37]a Quality management system (IMDRFap): Adamidi et al [37]a
	Responsible model use	Ethics ambient intelligence (Martinz-Martin)
Standards in the reviews mapped to multiple phases
	Design justice principles	10 principles—International Design Justice Network: Zidaru et al [68]
	Study quality	Reporting guidelines (EQUATORaq): Zidaru et al [68]
	Policy	China—AI governance (Laskai): Choudhury et al [45] Federal engagement plan (NISTar): Choudhury et al [45] Russian AI policy (OECDas): Choudhury et al [45] AMAat AI recommendations (link): Sapci and Sapci [60] CCCau AI road map (link): Zidaru et al [68]a EU (AI Watch): Zidaru et al [68]a
	Technical and interoperability	Software life cycle ISOav and IEEEaw-12207 (link): Zidaru et al [68]a OGCax: Seibert et al [61]a SWEay: Seibert et al [61]a SOSaz: Seibert et al [61]a AI concepts and terminology (ISO and IECba 22989): Seibert et al [61]a Framework for AI systems (ISO and IEC 23053): Wenzel and Wiegand [26]t AI risk management (ISO and IEC 23894): Wenzel and Wiegand [26]t AI bias (ISO and IEC 24027): Wenzel and Wiegand [26]t AI trustworthiness (ISO and IEC 24028): Wenzel and Wiegand [26]t AI robustness (ISO and IEC 24029-1): Wenzel and Wiegand [26]t AI use cases (ISO and IEC 24030): Wenzel and Wiegand [26]t Safety of machinery (ISO 22100-5): Choudhury et al [45]a
	Terminology standards	FAERSbb: Choudhury et al [45]a Medication information extract clinical notes (MedEx): Choudhury et al [45]a Medical prescription normalized (RxNorm): Choudhury et al [45]a MedDRAbc: Choudhury et al [45]a Patient-centered clinical research network (PCORnet): Choudhury et al [45]a UMLSbd: Choudhury et al [45]a
Robotics
	Partnership for R&Dbe and innovation	European robotics partnership (SPARCbf): Buchanan et al [42]a
	Robotic standardization and safety	Robotics standardization (ISO and TCbg 299): Buchanan et al [42]a
	Robotic devices for personal care	Personal care robots (13482: 2014): Buchanan et al [42]a Vocabulary (ISO 8373: 2021): Buchanan et al [42]a

^aItalicized references are original studies cited in the reviews, and references denoted with the footnote t are those cited in our paper but not present in any of the reviews.

^bAI: artificial intelligence.

^cFDA: Food and Drug Administration.

^dECLAIR: Evaluate Commercial AI Solutions in Radiology.

^eFHIR: Fast Healthcare Interoperability Resources.

^fFAIR: Findability, Accessibility, Interoperability, and Reusability.

^gPROBAST: Prediction Model Risk of Bias Assessment Tool.

^hHIPAA: Health Insurance Portability and Accountability Act.

ⁱOOTA: Office of The Assistant Secretary.

^jGDPR: General Data Protection Regulation.

^kEU: European Union.

^lWMA: World Medical Association.

^mWEF: World Economic Forum.

ⁿSORMAS: Surveillance, Outbreak Response Management and Analysis System.

^oWHO: World Health Organization.

^pML: machine learning.

^qTRIPOD: Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis.

^rTRIPOD-ML: Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis—Machine Learning.

^sCLAIM: Checklist for Artificial Intelligence in Medical Imaging.

^tReferences denoted with the footnote t are those cited in our paper but not present in any of the reviews.

^uCHARMS: Checklist for Critical Appraisal and Data Extraction for Systematic Reviews of Prediction Modeling Studies.

^vPRISMA-DTA: Preferred Reporting Items for Systematic Reviews and Meta-Analyses of Diagnostic Test Accuracy.

^wMI-CLAIM: Minimum Information About Clinical Artificial Intelligence Modeling.

^xMINIMAR: Minimum Information for Medical AI Reporting.

^yNOS: Newcastle-Ottawa Scale.

^zLOE: level of evidence.

^aaMMAT: Mixed Methods Appraisal Tool.

^abCASP: Critical Appraisal Skills Programme.

^acSTARD: Standards for Reporting of Diagnostic Accuracy Studies.

^adCOREQ: Consolidated Criteria for Reporting Qualitative Research.

^aeMADE1.0: Model Agnostic Diagnostic Engine 1.0.

^afDECIDE-AI: Developmental and Exploratory Clinical Investigations of Decision-Support Systems Driven by Artificial Intelligence.

^agSPIRIT-AI: Standard Protocol Items: Recommendations for Interventional Trials–Artificial Intelligence.

^ahCONSORT-AI: Consolidated Standards of Reporting Trials–Artificial Intelligence.

^aiRoB 2: Risk of Bias 2.

^ajROBINS-I: Risk of Bias in Non-Randomised Studies of Interventions.

^akRCT: randomized controlled trial.

^alSTROBE: Strengthening the Reporting of Observational Studies in Epidemiology.

^amAI-ML: artificial intelligence–machine learning.

^anTAM: Technology Acceptance Model.

^aoSaMD: Software as a Medical Device.

^apIMDRF: International Medical Device Regulators Forum.

^aqEQUATOR: Enhancing the Quality and Transparency of Health Research.

^arNIST: National Institute of Standards and Technology.

^asOECD: Organisation for Economic Co-operation and Development.

^atAMA: American Medical Association.

^auCCC: Computing Community Consortium.

^avISO: International Organization for Standardization.

^awIEEE: Institute of Electrical and Electronics Engineers.

^axOGC: Open Geospatial Consortium.

^aySWE: Sensor Web Enablement.

^azSOS: Sensor Observation Service.

^baIEC: International Electrotechnical Commission.

^bbFAERS: Food and Drug Administration Adverse Event Reporting System.

^bcMedDRA: Medical Dictionary for Regulatory Activities.

^bdUMLS: Unified Medical Language System.

^beR&D: research and development.

^bfSPARC: Scholarly Publishing and Academic Resources Coalition.

^bgTC: technical committee.