CoChat

Harnessing AI in critical care: opportunities, challenges and key steps for success

The DOI matches its Crossref record; 84 of the 85 references it cites were found in a registry or an index.

Abstract

BACKGROUND: The integration of artificial intelligence (AI) into critical care offers significant potential to enhance early diagnosis, predict patient deterioration, personalise treatment and inform clinical decision-making. Despite this promise, AI adoption in the intensive care unit (ICU) faces challenges, as illustrated by the limited number of AI tools which have been approved for clinical use and/or successfully deployed in critical care. METHODS: Aims of the review are to provide a synthesis of research on AI in critical care; assess approved tools; and consider challenges and opportunities, focusing on the different phases of the AI algorithm lifecycle in the ICU, including data collection, modelling, validation, implementation and post-deployment monitoring. Peer-reviewed publications were searched using terms relevant to AI and critical care spanning the years 2000-2025. RESULTS: Research on AI applications in the ICU is characterised by significant limitations including suboptimal data quality, retrospective analyses and a paucity of prospective validation studies. The few AI algorithms that have received Food and Drug Administration approval for use in the ICU have not gained widespread clinical adoption due, in part, to issues such as lack of user trust, integration challenges, unclear clinical impact or performance drift. Overcoming these barriers will require a structured approach that addresses the key challenges identified in the AI lifecycle, including the integration of real-world data, post-deployment performance monitoring, governance and ethical considerations. A successful implementation pathway should consider realistic goal-setting, greater model explainability, improved workflow integration and active end-user involvement. CONCLUSIONS: Advancing critical care with AI will require special attention to interdisciplinary collaboration, robust validation frameworks and adaptive governance models. The need for rigorous scientific evaluation needs to be balanced with the pressure for rapid deployment. Ensuring transparency, safety and alignment with clinical workflows will be critical to achieving meaningful AI integration in critical care.

Overview

Integrity signals

Registry

Crossref · 10.1136/thorax-2024-222125

Cites

84 of 84 references verified0 citation errors · 0 not checkable

Statistics

Recomputing p-values needs the paper’s full text.

Verification

Claims the paper makes

This paper’s full text isn’t openly available, so its claims and self-check weren’t run. The registry and citation checks still apply.

Citations

Cites

Show all 84 references

Cited by

See all 12

Metadata

Authors
Kalimouttou, A., Stevens, R. D., & Pirracchio, R.
Affiliations
  • Centre Hospitalier Universitaire de Grenoble (FR)
  • Johns Hopkins University (US)
  • University of California, San Francisco (US)
Corresponding author
Pirracchio, Romain
Venue
Thorax
Publisher
BMJ
Published
2025-06-15
Volume
81
Issue
2
Pages
183-192
ISSN
  • 0040-6376 (ISSN-L)
  • 1468-3296
OpenAlex
W4411319594
Cited by
12
References
85
Citations by year
2026: 8 · 2025: 4
FWCI
0.71
Type
journal article
Language
en
Open access
closed
Locations
  • Thorax (publishedVersion)
  • PubMed (publishedVersion)
Funders
  • Common Fund (R01EB035028-01)
  • Patient-Centered Outcomes Research Institute (P0562155)
  • Amomed Pharma (no grant number)
Topics
  • Health Sciences › Medicine › Health Informatics › Artificial Intelligence in Healthcare and Education
  • Physical Sciences › Computer Science › Artificial Intelligence › Machine Learning in Healthcare
  • Health Sciences › Medicine › Surgery › Healthcare Technology and Patient Monitoring
Keywords
intensive care unit, AI adoption barriers, model validation, clinical decision support, clinical deterioration prediction, early diagnosis, model explainability, post-deployment monitoring, performance drift, AI governance, ethical considerations, real-world data, clinical workflow integration, user trust
MeSH terms
Algorithms, 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.

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: Harnessing AI in critical care: opportunities, challenges and key steps for success. https://app.cochat.ai/papers/10.1136/thorax-2024-222125
The paper
Kalimouttou, A., Stevens, R. D., & Pirracchio, R. (2026). Harnessing AI in critical care: opportunities, challenges and key steps for success. Thorax. https://doi.org/10.1136/thorax-2024-222125

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.