Use of artificial intelligence in critical care: opportunities and obstacles

Michael R. Pinsky,Armando Bedoya,Azra Bihorac,Leo Celi,Matthew Churpek,Nicoleta J. Economou-Zavlanos,Paul Elbers,Suchi Saria,Vincent Liu,Patrick G. Lyons,Benjamin Shickel,Patrick Toral,David Tscholl,Gilles Clermont

DOI: https://doi.org/10.1186/s13054-024-04860-z

IF: 15.1

2024-04-10

Critical Care

Abstract:Perhaps nowhere else in the healthcare system than in the intensive care unit environment are the challenges to create useful models with direct time-critical clinical applications more relevant and the obstacles to achieving those goals more massive. Machine learning-based artificial intelligence (AI) techniques to define states and predict future events are commonplace activities of modern life. However, their penetration into acute care medicine has been slow, stuttering and uneven. Major obstacles to widespread effective application of AI approaches to the real-time care of the critically ill patient exist and need to be addressed.

critical care medicine

What problem does this paper attempt to address?

This paper aims to explore the challenges and obstacles faced in the application of Artificial Intelligence (AI) in Intensive Care Units (ICU) and proposes solutions to facilitate its effective implementation.

### Main Issues Addressed in the Paper:
1. **Challenges with Data and Algorithms**:
- Existing Clinical Decision Support Systems (CDSS) cannot fully replace doctors in the ICU because these systems lack contextual awareness.
- There are fundamental biases in large databases, which may not reflect the actual patient population being treated, leading to fairness issues.
- Technical barriers limit timely access to effective data and its useful presentation in clinical workflows.

2. **Trust and Acceptance Issues**:
- Many predictive algorithms and CDSS have a "black box" nature, which lowers their credibility and acceptance in the medical community.
- The real-time integration and curation of data streams from multiple sources and their transformation into clinical decision support formats that adapt to individual patient responses are often overlooked.

3. **Legal and Commercial Barriers**:
- Legal and commercial barriers restrict access to existing clinical databases, affecting the fairness of research and the generalizability of models.

4. **Social and Ethical Issues**:
- Using existing Electronic Health Records (EHR) to build AI-CDSS and evaluate models can encode and perpetuate structural inequalities due to existing health disparities.
- Regulatory frameworks need to be established to prevent AI from further marginalizing populations already burdened by disease and social injustice.

5. **Implementation and Validation Issues**:
- There is a significant "implementation gap" in deploying AI tools in actual clinical environments, especially in predicting clinical deterioration and sepsis.
- Rigorous evaluation frameworks are needed to ensure the effectiveness and safety of AI technologies, along with continuous monitoring of their performance in the real world.

By comprehensively addressing these issues, the paper aims to promote the effective application and development of AI technology in intensive care environments.