Early Health Technology Assessment of Artificial Intelligence in the Intensive Care Unit: Validation of a Generic Health-Economic Model

OBJECTIVES: Recently, an increasing number of studies emerge on improving care for intensive care unit (ICU) patients by using artificial intelligence (AI). However, the health economic impact of AI systems is quite rarely investigated. In previously published research, we have developed a generic health-economic model suitable for early health technology assessment (HTA) of different AI systems in the ICU. In this study we aim to validate this model, using cases taken from literature.

METHODS: The generic health-economic model simulates patients’ life trajectory starting from their hospitalisation until their death. Care with the AI system was compared to care as usual. We applied the generic health-economic model to an AI system predicting sepsis, and an AI system predicting ICU discharge. We compared the Incremental Cost Effectiveness Ratio (ICER) from the generic health-economic model with results from two previously published health-economic studies on these AI systems. Moreover, by varying input parameters and assumptions, we examined the cost-effectiveness of these AI systems under different scenarios.

RESULTS: The generic health-economic model was able to estimate the cost-effectiveness for both AI systems across a wide range of different clinical settings. Overall, the ICERs for both studies were comparable to the ICERs obtained in the previously published health-economic studies.

CONCLUSIONS: The validation of the generic health-economic model demonstrates the adaptability, flexibility, and reliability of the model. While results of more tailored approaches are likely to be more precise, the time gain of using the generic health-economic model is considerable, making it ideal for early HTA of AI systems in the ICU.