Generative AI may become a more prominent part of caregiving for skilled nursing operators who need to save time due to staffing shortages.
WellSky, a software analytics company that services healthcare providers, unveiled a collaboration with Google Cloud late last week to develop advanced data analytics tools, machine learning capabilities, and artificial intelligence in healthcare settings. Joel Dolisy, Chief Technology Officer at WellSky, told Skilled Nursing News that the move is aimed at helping address burnout in the industry and increase clinician time at the bedside.
Skilled Nursing News connected with Dolisy to discuss the specific goals WellSky aims to achieve through its partnership with Google Cloud and the integration of the generative AI tools known as Vertex AI. He also spoke about the possible impact of his organization’s technology on the post-acute care and skilled nursing sectors.
The following transcript has been edited for length and clarity.
SNN: Can you provide examples of how generative AI tools will be specifically applied in skilled nursing settings?
Dolisy: Last summer, we went to all of our different domain experts to understand where the clinicians are actually spending a lot of time with the systems and where we could actually kind of get a better experience overall for them. We talked to pretty much all of our end markets on the post-acute side, including home health, hospice, skilled nursing, and long-term care. What came out is that if you take the really high-level patient journey from referral intake to pre-admission to admission, and then there’s the cycle of visits and encounters to discharge, there was so much similarity there. It became clear that each of those steps in the patient journey between the different end markets had common problems that would apply to all of those end markets. Some of those are, for instance, when you receive any incoming referral, and you get a [Continuity Care Document] CCD, for instance, with 100 pages of clinical documentation, somebody’s got to look at this right now and try to decipher whether it’s the list of problems, allergies, or current medications. This is one of the key uses, for instance, at which generative AI is really good at. It’s good at summarizing and extracting information from documentation. And, instead of imagining things, the models are really pulling key information and allowing you to give you a reference back into the document where it found the fact. That’s what’s super important so that we can use the summarization capabilities of the technology while also making sure that it’s actually rooted in facts and not just the model’s imagination.
SNN: What are some other examples for which generative AI can be helpful in SNFs?
Dolisy: Talking with our long-term care and skilled nursing facilities, and their domain experts for discharge, generative AI can be helpful for very meaningful yet short summary discharge for the patient before that patient is discharged – and that can mean the difference between that patient being admitted into another home health or other care setting quickly. Again, we have access to all the patient information as part of the patient record that we maintain in our system. Wouldn’t it be great if we could actually just summarize that information based on the last several encounters, the last assessment that has happened, and create a relevant, summarized view of the patient’s state that is really helpful for ensuring that the discharged patient can actually be accepted into another care setting very quickly?
SNN: Can you talk about the measures taken to protect patient information?
Dolisy: We are trusted to take care of the data on behalf of the providers that we are serving and the patients. So, we are using this responsible AI framework that we are working closely with Google Cloud because they adhere to the same philosophy in terms of their technology and the separation of data from their models. We really want to make sure at that level that we apply thoroughness, we apply privacy and security, and that we provide transparency about how we get to the results that we are providing. That is key for the users to always be in the center of it.
SNN: Will AI completely eliminate the need for clinician judgement?
Dolisy: Our goal right now is not to replace humans. I don’t believe we’re going to get there, at least not while I’m working, and maybe I’ll be wrong on that. But right now, we want to have clinicians in the loop. They are in the middle of the decision-making, either taking what is suggested by the AI engine as a way to prefill, autocomplete, to some extent, in a much smarter way than is escapable right now. But they always have a choice to actually see, “Yeah, I trust that,” or “No, that doesn’t really match what I heard during this encounter, for instance. Let me edit that and then get this back.” We’re not anywhere near autofilling completely on assessment, for instance, and just saying, “Hey, it’s good to go,” without any type of review at all.
At WellSky, we have the beginning of a responsible AI governance board that ensures any AI project goes through a set of reviews, that make sure that the different capabilities planned for the product really go through a set of gates and filters around providing transparency and ensuring unbiased features. So, it really takes a set of deliberate steps for WellSky to ensure we are dealing with this in a very responsible manner for patients and our users.
SNN:You mentioned seeing “no shortage of possibilities” in the emerging space of AI. Could you provide insights into the potential future developments?
Dolisy: Right now, we’re focusing on clinical, but there’s another area also that is really interesting, which is the back office on the revenue cycle management side, on the medical coding side, where a lot of the solutions right now, to some extent, are using what used to be the bleeding edge of AI technologies, which was called RPA, robotic process automation. A lot of those capabilities now, I think, will be enhanced with generative AI adaptive capabilities going forward, with even higher accuracy to some extent of the automatic coding that is happening and really minimizing the amount of human review that is needed….And I think that’s what’s important there is to really think that this is a really, a clinician-centric approach and patient-centric approach…That’s the key thing that I would like to make sure that our people, our clients understand. We’re not there to replace them or that we’re really trying to make them spend more time with what they all signed up to do in healthcare, which is helping patients.
Companies featured in this article:
Centers for Medicare & Medicaid Services, Google Cloud, WellSky
