UCSF Health is in the early stages of testing various AI tools to tackle workforce burnout and staff shortages - with mixed feedback from nursing staff. Use cases they are testing: 🔮 Predictive analytics to flag patients deteriorating in-hospital The challenge with predictive analytics is always the “so what?” Unless there is a clear protocol on how to respond to the outputs of a predictive analytics engine, it’s just an academic exercise. We have clear protocols for when a patient suddenly develops a fever. But what do we do if predictive analytics “warn” me that a patient has a 20% risk of mortality in the next 30 days but no other symptoms? Until someone creates a protocol for these situations, it’s an insight with no action. 👩‍⚕️ AI matching engine to pair nurses with patients based on the nurse’s background, education and experience I guess perhaps language is the obvious matching parameter. Maybe clinical areas of interest…? But do we create a risk of frontline staff developing too narrow an experience of patient care if they end up treating a homogenous set of patients? Is increased specialization of this type net beneficial for patient care? I guess someone’s going to test it… 🏥 Predictive analytics for patient flow - recommending which patients should flow to which beds in the hospital OK this I like and is probably universally accepted (and I’m pretty sure already exists as 3rd party solutions) 📚 Custom GPTs for staff to query training materials E.g. asking the GPT “how to change a dressing” instead of needing to be taught live. This I like and we will likely see an internal FAQ chatbot for every organization - even outside of healthcare. While beneficial, this can’t be the killer use case… I mean, how much time does this actually save? Meanwhile a local nursing union has raised concerns that AI will reduce the clinical workforce, de-personalize patient care, and increase risks of error (e.g. AI bias). I think most of those concerns are largely addressable, especially with time and experience. And actually, the workforce shortage is so bad, I doubt AI will reduce the clinical workforce anytime soon. One could also ask: if AI could reduce the required clinical workforce while improving access/quality of car, why is that bad? That said I think this is a good example for getting the right initial use case - one that is high ROI and shows the potential of AI while being seen as universally beneficial by most parties - is so important for educating the healthcare system on the potential good for AI. Help everyone experience the potential good for AI and then people will be more open-minded to new use cases. What do you think? What use cases would get the most nursing buy-in while driving a high ROI?