Summary
Healthcare systems are struggling. There is a lack of time for comprehensive and empathetic communication with patients. Documentation and administration require an increasingly large proportion of healthcare providers’ time. Also, healthcare providers often fail to apply the medical knowledge that is relevant for a specific patient.
Large language models (LLMs) can address these problems and will therefore play an important role in healthcare. LLMs can free up time for clinicians by automating documentation and certain communication. They can also draw upon vast amounts of medical literature to give clinicians medical knowledge relevant for a specific patient.
As with all new technology, LLMs have risks and limitations that must be managed. LLMs can hallucinate and be wrong. They can also be with biased, with embedded and implicit preferences. It is unlikely that they will be able replace essential human qualities in healthcare, such as empathy. Furthermore, they can also create challenging ethical trade-offs which are difficult to solve.
Despite these limitations, LLMs are already being piloted and deployed in healthcare settings. Healthcare providers need to do several things to reap the benefits of LLMs while mitigating the risks. First, providers need to digitalize their healthcare processes in order to facilitate implementation of LLMs. Second, most providers will have to develop new capabilities in order to understand and successfully implement such systems. Third, due to the contextual nature of applying LLMs, it is important that providers share their experiences in order to help others avoid known pitfalls.
Hopefully, LLMs will help providers spend less time on monotonous tasks in front of computer screens, and more time with patients, practicing and enjoying the art of medicine.
Repport: Jonathan Ilicki – LLMs in Healthcare