Operationalization of Artificial Intelligence Applications in the Intensive Care Unit: A Systematic Review
Authors' conclusion
Despite substantial growth in AI research within intensive care medicine in recent years, the transition from development to clinical implementation still remains limited and has made little progress over time.
Study context
- Population
- Studies evaluating artificial intelligence (AI) applications designed for use within intensive care units (ICUs) for adults (aged ≥16 years) using data collected during ICU stays, published July 28, 2020, to June 10, 2024
- Participants
- 1,263 studies
- Outcome
- AI system operationalization in the ICU: progress over time, technical maturity (technology readiness level, TRL) and risk of bias
Methodological details5 databases, dual independent screening, PROBAST risk of bias assessment
- 5 databases
- Dual independent screening
- PROBAST risk of bias assessment
Pooled result
As reported in the abstract
Technology readiness level (TRL) of AI studies
Studies classified as technology readiness level (TRL) 4 or below (early-stage development or initial validation)
936
74%
External validation achieved (TRL 5)
24%
Progressed to clinical integration (TRL≥6)
25
2%
Studies reaching full implementation (TRL 9)
0
Further pooled results
Dataset origins (among TRL 4 or below studies)
- Internal datasets
- 447 (37%)
- MIMIC datasets (I-IV)
- 562 (46%)
- eICU Collaborative Research Database
- 78 (6%)
Generative AI models
- Reached TRL 5
- 14 [47%]
- Reached clinical integration
- 0
Reference to reporting standards
- Studies referencing reporting standards
- 207 (16%)
- Adherence in 2021
- 14%
- Adherence in 2024
- 23%
High risk of bias (PROBAST)
- Studies at high risk of bias, primarily due to methodologic shortcomings in the analysis domain
- 581 of 1103 (53%)
Original publication
Operationalization of Artificial Intelligence Applications in the Intensive Care Unit: A Systematic Review. JAMA Netw Open. Published 1 July 2025. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →