Generative artificial intelligence, patient safety and healthcare quality: a review

The capabilities of artificial intelligence (AI) have accelerated over the past year, and they are beginning to impact healthcare in a significant way. Could this new technology help address issues that have been difficult and recalcitrant problems for quality and safety for decades? While we are early in the journey, it is clear that we are in the midst of a fundamental shift in AI capabilities. It is also clear these capabilities have direct applicability to healthcare and to improving quality and patient safety, even as they introduce new complexities and risks. Previously, AI focused on one task at a time: for example, telling whether a picture was of a cat or a dog, or whether a retinal photograph showed diabetic retinopathy or not. Foundation models (and their close relatives, generative AI and large language models) represent an important change: they are able to handle many different kinds of problems without additional datasets or training. This review serves as a primer on foundation models' underpinnings, upsides, risks and unknowns-and how these new capabilities may help improve healthcare quality and patient safety.