St. Louis-based Lumeris describes itself as a value-based care managed services company. For instance, it serves as an operating partner for health systems in the ACO REACH program by providing end-to-end services and technology to help them be successful in managing downside risk. In a recent interview with Healthcare Innovation, the company’s chief technology officer, Jean-Claude Saghbini, discussed his role with the company and the impact artificial intelligence is having in the value-based care space.
Before joining Lumeris in 2021, Saghbini served as chief technology officer for Wolters Kluwer Health, where he led engineering for a portfolio of solutions, including consumer/patient engagement applications and clinical decision support. He also created the AI organization that built AI applications for early detection of a patient’s deteriorating medical condition using predictive models from structured and unstructured clinical data.
HCI: Why were you interested in moving from Wolters Kluwer to Lumeris?
Saghbini: A series of conversations with Mike Long, CEO of Lumeris, got me super-excited about joining Lumeris. It’s the ability to apply this combination of technology with services, coupled with more than a decade of know-how and success in value-based care and actually taking those solutions and implementing them in markets with health systems and being extremely tied to the outcomes. Being there on the frontlines tied to the outcomes was extremely exciting.
HCI: Is one of your missions at Lumeris to beef up the technology stack?
Saghbini: Yes, we have two platforms in our tech stack. We have a population health intelligence platform. The path there is how do we get better at more comprehensive data ingestion, and then extraction of more actionable insights from a variety of data sources? It goes from claims to EHR to social determinants of health data, pharmacy, labs, data. The work there is around more data, better algorithms, and better AI models to extract insights.
The second platform is called Lumeris Engage, a care orchestration platform. It's taking all of our knowledge we've acquired over more than a decade to automate that knowledge and orchestrate care, whether it's to patients, whether it's to physicians, to others in the provider setting. So that one is building further capabilities in terms of outreach, interactive experiences with patients, with providers, to drive them to what we call the next best action.
HCI: Are you regularly getting feedback from clinicians about what's the most valuable thing to them in terms of analytics or care coordination?
Saghbini: Yes, we have teams on the technology side that are on the front line, engaging all the time with end users, with physicians, and we get two types of feedback that drive our roadmap. One type of feedback is about the challenges they're facing in terms of needed insights to drive actions. Because we are also on the frontline with them in managing and value-based care, we are highly aware of all the things that matter in value-based care. The second sort of very loud feedback we get, which is highly driving our integration strategy, is physicians want things integrated in their workflow, right? They do not want to be switching between applications, no matter how smart your insight is. So insights and workflow integration are the two biggest drivers of our roadmap.
HCI: Do some of these ACOs you work with have eight different EHRs being used across the ACO?
Saghbini: Sometimes eight and sometimes 50! There is usually some large concentration of practices on one large EHR, and then there's this spectrum of other EHRs. And this becomes both the challenge but also where we differentiate ourselves in that we have built capabilities to both bi-directionally integrate with large EHRs as well as with array of smaller EHRs. We can push insights whether you have the large EHR or you have any of the other EHRs.
HCI: You spoke about improving data ingestion of lots of different data types. Is there any one type of data that's more of a struggle than any other to actually get and ingest that way?
Saghbini: Claims data can be a challenge because it is coming from a variety of payer sources, each with its own format. We have built deep expertise in ingesting that data, making sense of it, normalizing it to look like other data. We've designed the system for the ‘unhappy path,’ which means that we've designed the system with an expectation that there will be continuous errors in data coming in. Sometimes data will show up missing formats or not have all the data and we've designed a system that can detect the errors and raise a flag, so that you can get it corrected as fast as possible.
HCI: Do you have AI tools helping to automate some tasks? Can you describe that?
Saghbini: We've been on an AI journey for quite some time. We're particularly interested in some of the recent advances in generative AI. We use AI as a tool in our stack to do things like predicting risk of readmission, risk of hospitalization. We also use it to compute a variety of other risk scores for individuals.
We have found that we have been able to add to the accuracy of our models if we don't only build them on clinical data, claims data and lab and pharmacy data, but if we also inject into our models social determinants of health data. That does two things. One is it makes our models way more predictive, because some of the things that impact the trajectory of a patient's care may not be articulated well only through claims and clinical data. It also takes into consideration factors that drive health equity.
HCI: What are some new things you are excited about?
Saghbini: We are making some solid headway in the use of generative AI in our solutions. Some of it is for our own internal use inside the company to create efficiencies that can translate to a better member experience, for example. But we're also looking at applying some of these capabilities to extract further insights in the clinical space. We believe we're in a super exciting space because we have all of the technology capabilities around AI and now around generative AI, but we're also on the ground in our deployment model so that we can ensure that the human stays in the loop.
The other thing is more effectively using AI to further create that 360-degree view of the patient either for insights that were not available before, or it would have been required R&D investments and complicated software. At the same time, the ‘human in the loop’ means those insights don't translate directly to clinical action. There's a human in the loop and effectively AI is the assistant to the clinician.