Health IT, Startups

Can predictive analytics change how hospitals interact with patients?

Forecast Health recently raised $1.45 million from a super angel to grow its national sales team.

The Department of Health and Human Services is pushing for value-based reimbursement models to account for half of Medicare payments by 2018, per an announcement earlier this year. The move underscores the need for hospitals and accountable care organizations to have the tools to mitigate risk from readmissions to medication non adherence to reducing ER visits.

Although some hospitals and ACOs have developed this infrastructure internally, the health IT vendors developing the software to address this challenge reflect a broad range of startups and established companies. They differ not only in their strategy but the kind of datasets they tap to develop the predictive analytics tools to help institutions identify which patients need more attention when they’re discharged.

In a phone interview with Michael Cousins, the president and Chief Analytics Officer of predictive analytics startup Forecast Health, he talked about its approach.

Cousins previously worked for Evolent Health and served as vice president of analytics at Cigna. He has also worked in health informatics at Wellpoint. Referring to ACOs, Cousins said: “Providers are becoming more like payers and need to better understand and execute managed risk. Essentially, we are trying to take a lot of the learning we have had with health plan analytics and construct those analytics to maximize and optimize what providers need today.”

Forecast Health uses roughly 100 consumer-oriented, socio-economic datasets that can point to different types of risks for hospitals such as medication non adherence, readmission, and other things. Cousins noted that several companies tend to focus on a zip code or other census data to gauge risk but he finds that insufficient.

For example, a patient could experience financial stress even if they live in a zipcode associated with wealthy households.

If patients don’t have access to a car and lack reliable public transportation, that can reduce the likelihood of making an appointment. Patients with financial distress are more likely to be readmitted because they may have difficulty affording medication. If they live alone, that can significantly ramp up the chances of readmission because when they are discharged from the hospital, they’re in a fragile state and may lack caregiver support. But it’s critical to do those kinds of assessments before patients are discharged when hospitals can intervene easily, Cousins said.

Its predictive analytics tools are embedded in electronic medical records. The ability to access this information when it’s needed can open up the way for a talk with patients about free medication or getting them into a discount program.

Admittedly, Cousins said the company is still working on pilots and was only willing disclose one client — University of North Carolina Health Care. It recently raised $1.45 million from a super angel to grow its national sales team. The investor backed the company because he believed in what the company is doing and thinks its approach could save lives.

He views many of the players in the predictive analytics market as being in a quandry. It is divided into two types of businesses with shortcomings in each camp. There are the established companies such as Optum, Verisk and Milliman which have extremely well-developed analytics tools, but as Cousins sees it, their analytics are as old as their respective companies. On the other hand, younger companies are more nimble and have a fresher approach to using predictive analytics, but they often lack a nuanced understanding of clinical workflows.

“Talking about predictive analytics to reduce readmissions is easy, but few are actually doing it,” he said. “What matters is not just having a big pot of data on a supercharged technology platform; you need to be careful about constructing the variables to predict impactable patient readmissions.”

Shares0
Shares0