Advanced-analytics to improve critical care: Moving from bytes to bedside

AI research for healthcare increased rapidly, particularly in the ICU, but still most studies remain in development and prototyping stages and have limited impact on patient care. To transition AI from bytes (theoretical concepts) to bedside (practical applications) in patient treatment, it's crucial for healthcare professionals to develop a thorough understanding of AI technologies. Additionally, a methodical strategy is needed for the development and implementation of these technologies, with a strong emphasis on ethical design principles. Moreover, cross-disciplinary collaborations are important to ensure safe and responsible use. The DESIRE study demonstrates AI's potential in predicting safe hospital discharge. This thesis demonstrates the consistent model performance across diverse hospital settings and feasibility to use the system in real-world clinical practice. Despite AI demonstrating predictive utility, testing in operational real-world clinical settings is a crucial step to assess feasibility and safety. A large intervention study is warranted to determine its true clinical utility.
It is evident that AI offers significant potential for enhancing healthcare diagnostics, prognostics, and treatment methods. However, there exists a notable gap between the development of AI models and their clinical assessment, coupled with issues of potential biases and ethical considerations. This gap presents substantial obstacles to the safe and responsible deployment of AI in healthcare. Additional research is required to establish a framework for responsible AI development and adoption in healthcare, providing practical guidance on effectively addressing these challenges.