In 2020, I wrote about how New Jersey-based RWJBarnabas Health (RWJBH) had begun working with a startup called Wellsheet, whose solution pulls data from the EHR, prioritizes clinical content through machine learning algorithms, and optimizes the provider workflow. Another health system that has been working with Wellsheet is Concord Hospital in New Hampshire. Its chief quality officer, Christopher Fore, M.D., recently spoke with Healthcare Innovation about the impact it has had.
Newark, N.J.-based Wellsheet’s predictive clinical workflow platform uses the FHIR API standards to work within an existing EHR to surface the most relevant content for physicians in a view that is contextualized and prioritized for the clinician’s needs. For instance, Wellsheet brings risk tools to the forefront. If the patient has Afib, Wellsheet brings the appropriate calculator in the EHR right to the forefront. Fore said that Concord’s clinicians have been using it for mobile chart review, real-time notifications, care team collaboration, handoff, and discharge planning.
Healthcare Innovation: Were there some limitations with your EHR at Concord that made you take interest in Wellsheet?
Fore: We started the collaboration back in 2019 just before the pandemic. We leveraged Wellsheet initially just to make lists for teams and customized ways to look at information to make our day a little easier. It turns out, that is not actually the best reason to use Wellsheet. It also allows you to better manage patient flow and access and capacity, set notifications and communicate better.
HCI: So that issue about managing flow and access and capacity — is that for administrators, or someone managing a hospital unit?
Fore: It tends to be used by people of a clinical background in an administrative position like myself. Our utilization management teams and care management teams use it all the time. Our nursing unit leadership looks at it all the time, and clinical teams at our whiteboards.
HCI: You’re a chief quality officer. Do you look at it from a quality or patient safety lens?
Fore: I do. I will tell you that in every quality review that I facilitate I have Wellsheet open and on the screen in the room. Because if you want to trend vital signs, look at laboratories, read providers’ notes, look at imaging results, it’s all there. The way the data is trended, the visualization, is just so far superior to what we can get inside the electronic health record. It makes it really, really easy. I can’t remember the last time I actually went into Cerner to review a case, as opposed to going into Wellsheet to review the data from Cerner regarding the case, because it is a lot easier.
HCI: Can we come up with an example of the return on investment from using Wellsheet?
Fore: One of the things that’s an accountability of mine is managing length of stay — reducing the time inpatients need to be in a hospital bed. I think everyone recognizes it’s a super-scarce resource. Being in the hospital is expensive and sometimes not even the best place to be if you don’t need to be there. So we’ve been working really hard due to issues of capacity constraints and access for the community to try to reduce length of stay, like many health systems, and we’ve actually achieved a length-of-stay reduction of about a day in the last year. We’ve gone from 5.5 days to 4.5 days in a year, which is pretty remarkable.
We have found that the key is teamwork — people communicating. With a patient being in the hospital, it’s like this big multi-contributor process. You’re trying to get the nursing team, multiple consultative services, interventional radiology, imaging, laboratory results, care management, utilization management all on the same page. You can’t all be in the same place at the same time, so communicating is key. We’ve developed ways to develop team notes inside Wellsheet that allow us to leave notes regarding what’s going on and they can be handoff notes. They can also be group notes that are more related to a challenging disposition, a homeless patient, someone who is going to a facility but hasn’t yet qualified for Medicaid, things like that.
HCI: You mentioned that you started working with Wellsheet right before the pandemic? Did that help with some of the critical issues that came up during the pandemic?
Fore: Yes, it was our go-to tool. Inside the electronic health record, you have all this information, all this discrete data. But you know, big EHR vendors, Epic and Cerner, can’t be nimble and change the way they operate, or the way they visualize data on a dime, but literally within a couple of weeks, we saw that we needed to make a list of every patient with COVID in the hospital. We wanted to know what’s the precaution status. How long have they been here? When did they last test positive? When does their next test need to be done? What kind of oxygen delivery are they on? And we built a list we maintained, and all we had to do is just slide patients onto that list, and at a glance, you can tell how many COVID patients are in the hospital. How are they doing? How long have they been there? When did they test? What kind of test did they have? It was really slick.
HCI: Did that help you manage resources that you needed to deploy?
Fore: Yes, the height of the pandemic was really challenging from a capacity standpoint, and we had built out the facility literally overnight to be able to to manage 60 patients on ventilators. I think we got up into the mid-50s, but we never ticked into not having a ventilator for somebody. But you can imagine when you’re at 55 and you can do 60, being able to check at 10 o’clock how many people we have was really game-changing for us.
HCI: I’ve been talking to a lot of health systems that are deploying generative AI tools to help generate clinical notes from patient encounters. Has Concord looked at doing that for its physician encounters with patients?
Fore: Yes, we are actually looking at it. The Wellsheet team has done something that isn’t necessarily generative, but is definitely on the ML spectrum, and saves you a lot of time as well. Wellsheet is really good about pulling discoverable discrete data out of the EHR. If you go into a patient chart and that patient has coronary disease and they’re in the emergency department for chest pain, it may actually calculate some of the cardiac risk scores with available data before you’re even doing it. I also like the way that it pulls the laboratory data and just puts it in an easily digestible fashion in the exact way that you would write it in longhand on a piece of paper in your pocket. We have had really good uptake with our cardiology group, because it brings together all the laboratories they regularly check in one place.
Joining the conversation was Craig Limoli, CEO and founder of Wellsheet.
HCI: Craig, do you have to work with Cerner and Epic on these implementations — and are they happy with you guys, or is it frustrating for them to hear someone like Dr. Fore say that this is so much better than having to go into Cerner to find this information?
Limoli: The integration itself is very arms length. We have fully API-based integration models, so we integrate through the FHIR standards. That’s something we can dependably access across all health systems, since it’s federally regulated. So it’s not as if we need to bring in teams from Cerner to deploy our product. And because of this API-based integration model, we can also go live very quickly — within a month or two. In terms of how they perceive it, I think at the end of the day, the EHR vendors that are successful are ones that care about their clients having the best outcomes, and this is clearly something that’s contributing to that goal. So, we’ve been fairly well received across the board.
By the way, one of the capabilities that large language models have is the ability to sort through unstructured text in documents, pull out key data elements and surface those to physicians as well. That can really enable expedited review of the most relevant information. Pre-filling calculators and the like can be expedited with that type of application of these new AI tools. So it’s not just the generation of notes that large language models enables, although we are doing that as well.
