PhD defence J.D. (Jessica) Workum

Below is a brief summary of the dissertation:

Although artificial intelligence (AI) can vastly improve healthcare, many promising ideas often remain in the research phase. With the arrival of large language models, such as ChatGPT, putting AI into practice has become even more urgent and complex. These models can assist with administrative tasks as well as medical decision-making, but they also introduce new risks, especially in highly complex acute settings like the intensive care unit (ICU).
This thesis investigates how hospitals can safely and responsibly deploy these language models. It provides a clear step-by-step plan and practical tools. For example, it offers a governance framework at the application level to keep patient data secure, while also offering guidance through complex European AI legislation. Furthermore, the research shows that current language models already possess extensive medical knowledge, and are capable of generating summaries or ICU family meeting notes of comparable quality to those written by physicians.
The core of the research is a model based on four key pillars: clinical validation, technical governance, regulatory compliance, and a well-prepared organization. By starting small with administrative tasks before scaling up to medical decision support, AI can be successfully and responsibly integrated into healthcare.