The DOI matches its Crossref record; 79 of the 79 references it cites were found in a registry or an index; 45 of the 67 sources its claims cite were checked for support.
Implementing Artificial Intelligence in Critical Care Medicine: a consensus of 22.
Open accessAbstract
Artificial Intelligence (AI) is rapidly transforming the landscape of critical care, offering opportunities for enhanced diagnostic precision and personalized patient management. However, its integration into ICU clinical practice presents significant challenges related to equity, transparency, and the patient-clinician relationship. To address these concerns, a multidisciplinary team of experts was established to assess the current state and future trajectory of AI in critical care. This consensus identified key challenges and proposed actionable recommendations to guide AI implementation in this high-stakes field. Here we present a call to action for the critical care community, to bridge the gap between AI advancements and the need for humanized, patient-centred care. Our goal is to ensure a smooth transition to personalized medicine while, (1) maintaining equitable and unbiased decision-making, (2) fostering the development of a collaborative research network across ICUs, emergency departments, and operating rooms to promote data sharing and harmonization, and (3) addressing the necessary educational and regulatory shifts required for responsible AI deployment. AI integration into critical care demands coordinated efforts among clinicians, patients, industry leaders, and regulators to ensure patient safety and maximize societal benefit. The recommendations outlined here provide a foundation for the ethical and effective implementation of AI in critical care medicine.
Overview
CoChat Summary
## TL;DR
This paper is an expert consensus calling for human-centric, ethically governed, and clinically validated AI in critical care, organized around four domains: human-centric use, clinician training, standardized data infrastructure, and governance/regulation. It argues that AI should augment rather than replace clinicians and that deployment must include transparency, bias monitoring, privacy protection, and post-market surveillance.
## Key findings
- The consensus identifies four main areas for AI integration in critical care: 1) human-centric AI, 2) clinician training on AI use, 3) standardization of data models and networks, and 4) AI governance.
- The paper emphasizes that AI should increase time for patient-physician communication and be used to automate burdensome tasks such as documentation and transfers of care.
- It argues that many AI tools are poorly validated and untested in real settings, and warns that premature deployment can cause harm, bias, and worsening health inequities.
- It recommends stronger safeguards including formal oversight mechanisms (for example, hospital AI committees), continuous post-market evaluation, and regulatory processes that address liability and insurance.
## Methods
- This is a multidisciplinary consensus paper, not a formal systematic review or evidence-graded guideline.
- The consensus emerged through expert discussions among selected participants chosen by the consensus chairs based on recognized contributions to AI in critical care; the panel included intensivists, AI researchers, data scientists, and experts.
- Recommendations were developed through iterative discussion and critical examination of real-world challenges, current research, and regulatory landscapes, rather than formal voting or evidence grading.
## Limitations
- The paper explicitly states that the consensus was not based on formal grading or voting on evidence; this limits how strongly the recommendations can be tied to a structured evidence hierarchy.
- It also states that AI in critical care is rapidly evolving and many critical questions remain unanswered, which the text makes evident as a limitation of the field and of the recommendations.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Figures



Verification
Claims the paper makes
32 Valid · 12 Not valid · 13 Not checked · 34 Own results and interpretations
Some cited sources support this claim and some don’t — read each source’s verdict below.
In the paper
“Despite some promising results [27, 28], the clinical application of AI remains limited [29–31].”
§ Implementing Artificial Intelligence in Critical Care Medicine: a consensus of 22 > Introduction > Recommendations
What the sources say
- Not checked yetcites [29] Hah H, Goldin D. Moving toward AI-assisted decision-making: Observation on clinicians’ management of multimedia patient information in synchronous and asynchronous telehealth contexts. Health Informatics J 2022; 28. https://doi.org/10.1177/14604582221077049/ASSET/IMAGES/LARGE/10.1177_14604582221077049-FIG4.JPEG. doi:10.1177/14604582221077049/asset/images/large/10.1177_14604582221077049-fig4.jpeg
- Disputedcites [30] Shen J, Zhang CJP, Jiang B, et al. Artificial intelligence versus clinicians in disease diagnosis: Systematic review. JMIR Med Inform 2019; 7. https://doi.org/10.2196/10010. sourceExperimental — accuracy not yet measured
The Introduction (Background; page not provided) says AI has been extensively used across medical fields, including to assist clinical diagnoses. Although the Discussion (Principal Findings; page not provided) identifies unresolved challenges, it does not characterize clinical application as limited; it describes AI as improving efficiency in clinical practice. (LLM, conf=0.78)
Checked by a reasoning model — may be wrong.
- Supportedcites [31] Zhang A, Xing L, Zou J, Wu JC. Shifting machine learning for healthcare from development to deployment and from models to data. Nature Biomedical Engineering 2022 6:12 2022; 6: 1330–45. source
The Data-limiting factors in the deployment of ML section says the past decade focused on model development and the next will be defined by deployment into clinical settings. The Abstract notes that some systems are beginning to enter clinics, while the Outlook describes implementation as a central challenge. Together, these support the broad claim that clinical application remains limited, without implying that it is absent. (LLM, conf=0.88)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
Every source the paper cites for this claim was checked and supports it.
In the paper
“For example, machine learning has been used to stratify critically ill patients based on their response to specific therapeutic interventions, potentially improving clinical outcomes [37].”
§ Implementing Artificial Intelligence in Critical Care Medicine: a consensus of 22 > Introduction > Recommendations
What the sources say
- Supportedcites [37] Zhang Z, Chen L, Sun B, et al. Identifying septic shock subgroups to tailor fluid strategies through multi-omics integration. Nat Commun. 2024;15:9028. source
The Methods describe a model that identifies septic-shock subgroups by their expected response to restrictive versus liberal fluid strategies. The Results (Subgroup identification; page not provided) report better survival when patients received the strategy recommended for their subgroup, supporting the claim’s qualified statement about potentially improved outcomes. (LLM, conf=0.91)
hedgedstates a possible cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“In a large ICU cohort of patients with traumatic brain injury (TBI), unsupervised clustering identified six distinct subgroups, based on combined neurological and metabolic profiles. [38]”
§ Implementing Artificial Intelligence in Critical Care Medicine: a consensus of 22 > Introduction > Recommendations
What the sources say
- Supportedcites [38] Åkerlund CAI, Holst A, Stocchetti N, et al. Clustering identifies endotypes of traumatic brain injury in an intensive care cohort: a CENTER-TBI study. Crit Care. 2022;26:1–15. source
The Methods passage identifies an ICU cohort of 1,728 patients with TBI and an unsupervised clustering model. The Discussion and Conclusions report six distinct subgroups characterized by GCS alongside systemic metabolic measures, matching the claimed neurological and metabolic profiles. (LLM, conf=0.95)
plainstates an association- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“Seymour et al. demonstrated how machine learning can stratify septic patients into clinically meaningful subgroups using high-dimensional data, which can subsequently inform risk assessment and prognosis [35].”
§ Implementing Artificial Intelligence in Critical Care Medicine: a consensus of 22 > Introduction > Recommendations
What the sources say
- Supportedcites [35] Seymour CW, Kennedy JN, Wang S, et al. Derivation, Validation, and Potential Treatment Implications of Novel Clinical Phenotypes for Sepsis. JAMA. 2019;321:2003. source
The Statistical Methods section reports using consensus k-means clustering on 29 clinical variables to derive sepsis phenotypes. The Relationship With Mortality and Organ Support section reports distinct short- and long-term outcomes across those phenotypes, supporting their clinical relevance for prognosis and risk assessment. No page numbers are available in our copy. (LLM, conf=0.93)
boostedstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“AI algorithms may be harmful when prematurely deployed without adequate control [9, 15–17].”
§ Implementing Artificial Intelligence in Critical Care Medicine: a consensus of 22 > Introduction > Recommendations
What the sources say
- Not checked — no full text of the sourcecites [9] Kwong JCC, Nguyen D-D, Khondker A, et al. When the Model Trains You: Induced Belief Revision and Its Implications on Artificial Intelligence Research and Patient Care — A Case Study on Predicting Obstructive Hydronephrosis in Children. NEJM AI 2024; 1. https://doi.org/10.1056/AICS2300004. source
- Partly supportedcites [15] Perry Wilson F, Martin M, Yamamoto Y, et al. Electronic health record alerts for acute kidney injury: multicenter, randomized clinical trial. BMJ 2021; 372. https://doi.org/10.1136/BMJ.M4786. sourceExperimental — accuracy not yet measured
The Results report worse outcomes in two non-teaching hospitals that received acute kidney injury alerts, and the Discussion (Comparisons with other studies) says seemingly benign alerts should not be implemented without robust evidence of safety and efficacy. This supports the warning about harm from insufficiently evaluated deployment, but the intervention was an electronic health record alert; the paper does not establish the claim specifically for AI algorithms. (LLM, conf=0.86)
- Partly supportedcites [16] Ji Z, Lee N, Frieske R, et al. Survey of Hallucination in Natural Language Generation. ACM Comput Surv 2023; 55. https://doi.org/10.1145/3571730. sourceExperimental — accuracy not yet measured
The Introduction says hallucinations in natural-language generation raise safety concerns in real-world applications and could harm patients; Section 13.3.2 discusses safety fine-tuning to mitigate such risks. The paper supports the possibility of harm, but does not establish the broader claim about AI algorithms generally or specifically link harm to premature deployment without adequate control. (LLM, conf=0.84)
- Supportedcites [17] Oniani D, Hilsman J, Peng Y, et al. Adopting and expanding ethical principles for generative artificial intelligence from military to healthcare. npj Digital Medicine 2023 6:1 2023; 6: 1–10. source
The Reliability section calls for thorough evaluation and testing during development and deployment to minimize failures and harm. The Governability section says deployed systems need meaningful human control, including monitoring and the ability to deactivate a system that risks harming patients. Together, these support the claim that deploying AI without adequate safeguards may be harmful. (LLM, conf=0.88)
hedgedstates a possible cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
17 more claims · 9 with their cited sources checked
Create a free account to see all 91 claims and to verify the 2 references still open.
- It may state two separate claims as one. 27
- Probably not a substantial claim. 58
Citations
Cites
High-performance medicine: the convergence of human and artificial intelligence
Topol, E. J. (2019) · Nature Medicine · doi:10.1038/s41591-018-0300-7 · Cited by 10,450 · cited 1×
“[1, 2] Without a structured approach to implementation, evaluation, and control, this transformation may concrete recommendations to guide the safe, effective, and meaningful adoption of AI into critical care.”
Survey of Hallucination in Natural Language Generation
Ji, Z., Lee, N., Frieske, R., Yu, T., Su, D., Xu, Y., Ishii, E., Bang, Y. J., Madotto, A., & Fung, P. (2023) · ACM Computing Surveys · doi:10.1145/3571730 · Cited by 5,025
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×
“For predictive models, the TRIPOD-AI extension of the TRIPOD guidelines focuses on transparent reporting for clinical prediction models with specific emphasis on calibration, internal and external validation, and fairness [68].”
Self-citationKung, T. H., Cheatham, M., Medenilla, A., Sillos, C., De Leon, L., Elepaño, C., Madriaga, M., Aggabao, R., Diaz-Candido, G., Maningo, J., et al. (2023) · PLOS Digital Health · doi:10.1371/journal.pdig.0000198 · Cited by 4,115 · cited 1×
“In addition to undergraduate education, it is essential to train experienced physicians, nurses, and other allied health professional [43].”
Kung, T. H., Cheatham, M., ChatGPT, Medenilla, A., Sillos, C., De Leon, L., Elepaño, C., Madriaga, M., Aggabao, R., Diaz-Candido, G., et al. (2022) · doi:10.1101/2022.12.19.22283643 · Cited by 4,115
Deo, R. C. (2015) · Circulation · doi:10.1161/circulationaha.115.001593 · Cited by 3,466
Derivation, Validation, and Potential Treatment Implications of Novel Clinical Phenotypes for Sepsis
Seymour, C. W., Kennedy, J. N., Wang, S., Chang, C.-C. H., Elliott, C. F., Xu, Z., Berry, S., Clermont, G., Cooper, G., Gomez, H., et al. (2019) · JAMA · doi:10.1001/jama.2019.5791 · Cited by 1,567 · cited 2×
“Before integrating these methodologies into clinical workflows, we need to make sure clinicians can accept the paradigm shift between broad syndromes and specific sub-phenotypes, ultimately supporting the transition toward personalized medicine [35, 39–41].”
Self-citationMathew, D., Giles, J. R., Baxter, A. E., Oldridge, D. A., Greenplate, A. R., Wu, J. E., Alanio, C., Kuri-Cervantes, L., Pampena, M. B., D’Andrea, K., et al. (2020) · Science · doi:10.1126/science.abc8511 · Cited by 1,509 · cited 1×
“Before integrating these methodologies into clinical workflows, we need to make sure clinicians can accept the paradigm shift between broad syndromes and specific sub-phenotypes, ultimately supporting the transition toward personalized medicine [35, 39–41].”
Liu, X., Cruz Rivera, S., Moher, D., Calvert, M. J., Denniston, A. K., The SPIRIT-AI and CONSORT-AI Working Group, SPIRIT-AI and CONSORT-AI Steering Group, Chan, A.-W., Darzi, A., Holmes, C., et al. (2020) · Nature Medicine · doi:10.1038/s41591-020-1034-x · Cited by 1,421
Moons, K. G. M., Damen, J. A. A., Kaul, T., Hooft, L., Andaur Navarro, C., Dhiman, P., Beam, A. L., Van Calster, B., Celi, L. A., Denaxas, S., et al. (2025) · BMJ · doi:10.1136/bmj-2024-082505 · Cited by 1,069 · cited 1×
“PROBAST-AI framework complements this by offering a structured tool to assess risk of bias and applicability in prediction model studies [69].”
Self-citation
Research works printed but not linked: 2
- Rony MKK, Parvin MR, Wahiduzzaman M, Debnath M, Bala S Das, Kayesh I. “I Wonder if my Years of Training and Expertise Will be Devalued by Machines”: Concerns About the Replacement of Medical Professionals by Artificial Intelligence. SAGE Open Nurs 2024; 10. https://doi.org/10.1177/23779608241245220/ASSET/IMAGES/LARGE/10.1177_23779608241245220-FIG1.JPEG. · doi:10.1177/23779608241245220/asset/images/large/10.1177_23779608241245220-fig1.jpeg
- Sauer CM, Dam TA, Celi LA, et al. Systematic Review and Comparison of Publicly Available ICU Data Sets—A Decision Guide for Clinicians and Data Scientists. Crit Care Med. 2022;50: e581. · DOI · Paper page
Cited by
AI in critical care: A roadmap to the future
Workum, J. D., Meyfroidt, G., Bakker, J., Jung, C., Tobin, J. M., Gommers, D., Elbers, P. W. G., van der Hoeven, J. G., & Van Genderen, M. E. (2026) · Journal of Critical Care · doi:10.1016/j.jcrc.2025.155262 · Cited by 17
Self-citationRecent approaches of artificial intelligence in intensive care unit: A review
Havelikar, U. A., Bembde, A., Patil, D., Shinde, A., Sharma, R. K., Patel, V. P., & Chaudhari, N. B. (2025) · Intelligent Hospital · doi:10.1016/j.inhs.2025.100030 · Cited by 11
Physiology-guided personalized mechanical ventilation to prevent ventilator-induced lung injury
Merola, R., Battaglini, D., Schultz, M. J., & Rocco, P. R. M. (2026) · Frontiers in Medicine · doi:10.3389/fmed.2026.1764151 · Cited by 10
“Emerging computational tools, including physiologic modeling, parameter estimation, and machine learning, offer new opportunities to further refine individualized ventilatory care (144, 145).”
From promise to practice: A roadmap for artificial intelligence in critical care
Agard, G., Hraiech, S., & Gauss, T. (2025) · Journal of Critical Care · doi:10.1016/j.jcrc.2025.155263 · Cited by 9
AI-Enabled Precision Nutrition in the ICU: A Narrative Review and Implementation Roadmap
Briassoulis, G. C., & Briassouli, E. (2025) · Nutrients · doi:10.3390/nu18010110 · Cited by 10
“In this context, “human-centric AI” is emphasized as a tool that supports rather than supplants clinical judgment, reinforcing empathetic care and protecting the therapeutic relationship [4].”
Metadata
- Authors
- Cecconi, M., Greco, M., Shickel, B., Angus, D. C., Bailey, H., Bignami, E., Calandra, T., Celi, L. A., Einav, S., Elbers, P., Ercole, A., Gómez, H., Gong, M. N., Komorowski, M., Liu, V., Park, S., Sarwal, A., Seymour, C. W., Zampieri, F. G., Taccone, F. S., Vincent, J.-L., & Bihorac, A.
- Affiliations
- Humanitas University (IT)
- University of Florida Health (US)
- University of Pittsburgh (US)
- Durham VA Medical Center (US)
- University of Parma (IT)
- Centre Hospitalier Universitaire Vaudois (CH)
- Beth Israel Deaconess Medical Center (US)
- Hebrew University of Jerusalem (IL)
- Amsterdam University Medical Centers (NL)
- University of Cambridge (GB)
- Albert Einstein College of Medicine (US)
- Imperial College London (GB)
- Kaiser Permanente (US)
- NewYork–Presbyterian Hospital (US)
- Atrium Health Wake Forest Baptist (US)
- University of Alberta (CA)
- Erasmus Hospital (BE)
- Corresponding author
- Cecconi, Maurizio
- ORCID
- Cecconi, Maurizio
- Greco, Massimiliano
- Shickel, Benjamin
- Angus, Derek C.
- Calandra, Thierry
- Celi, Leo Anthony
- Einav, Sharon
- Elbers, Paul
- Ercole, Ari
- Gómez, Hernando
- Gong, Michelle Ng
- Komorowski, Matthieu
- Liu, Vincent
- Park, Soojin
- Sarwal, Aarti
- Seymour, Christopher W.
- Zampieri, Fernando G.
- Taccone, Fabio Silvio
- Vincent, Jean-Louis
- Bihorac, Azra
- Venue
- Critical Care
- Published
- 2025-07-08
- Volume
- 29
- Issue
- 1
- Pages
- 290-290
- ISSN
- 1364-8535 (ISSN-L)
- PMC
- PMC12239397
- OpenAlex
- W4412105411
- Cited by
- 80
- References
- 78
- Citations by year
- 2026: 61 · 2025: 19
- FWCI
- 19.09
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Licence
- cc-by-nc-nd
- creativecommons.org/licenses/by-nc-nd/4.0 (tdm)
- creativecommons.org/licenses/by-nc-nd/4.0 (vor)
- Locations
- Critical Care (publishedVersion, cc-by-nc-nd)
- PubMed Central (publishedVersion, cc-by-nc-nd)
- PubMed (publishedVersion)
- Pure Amsterdam UMC (publishedVersion, other-oa)
- SERVAL (Université de Lausanne) (publishedVersion, cc-by-nc-nd)
- IRIS (submittedVersion, cc-by-nc-nd)
- Apollo (University of Cambridge) (submittedVersion, cc-by-nc-nd)
- Dépôt institutionnel de l'Université libre de Bruxelles (Université Libre de Bruxelles) (submittedVersion, other-oa)
- DSpace@MIT (Massachusetts Institute of Technology) (submittedVersion, cc-by-nc-nd)
- Apollo (University of Cambridge) (submittedVersion, cc-by)
- Apollo (cc-by)
- Funders
- National Science Foundation (2148451, Bridge2AI OT2OD032701, DS-I Africa U54 TW012043-01)
- Korea Health Industry Development Institute (RS-2024-00403047)
- National Institutes of Health (OT2OD032701)
- Fogarty International Center (U54TW012043)
- Topics
- Health Sciences › Medicine › Health Informatics › Artificial Intelligence in Healthcare and Education
- Physical Sciences › Computer Science › Artificial Intelligence › Machine Learning in Healthcare
- Physical Sciences › Computer Science › Artificial Intelligence › Ethics and Social Impacts of AI
- Keywords
- artificial intelligence, intensive care, intensive care unit, AI implementation, algorithmic bias, transparency, patient-clinician relationship, personalized medicine, data sharing, artificial intelligence in education, AI regulation, patient safety, AI ethics, consensus recommendations
- MeSH terms
- Artificial Intelligence, Critical Care, Humans, Intensive Care Units
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.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
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
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: Implementing Artificial Intelligence in Critical Care Medicine: a consensus of 22. https://app.cochat.ai/papers/10.1186/s13054-025-05532-2
- The paper
- Cecconi, M., Greco, M., Shickel, B., Angus, D. C., Bailey, H., Bignami, E., Calandra, T., Celi, L. A., Einav, S., Elbers, P., Ercole, A., Gómez, H., Gong, M. N., Komorowski, M., Liu, V., Park, S., Sarwal, A., Seymour, C. W., Zampieri, F. G., . . . Bihorac, A. (2025). Implementing Artificial Intelligence in Critical Care Medicine: a consensus of 22. Critical Care. https://doi.org/10.1186/s13054-025-05532-2
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.