Predictive algorithms fuel risk stratification efforts at Denver Health

To improve care coordination and deploy enhanced staff resources and tailor care to meet individual patient needs, a 21st Century Care project deployed by Denver Health involved applying predictive modeling, according to an article at eGEMS (Generating Evidence and Methods to improve patient outcomes).

The work was based on a $19.8 million federal grant to implement a "population health" approach into the delivery of primary care. The organization used algorithms to identify four levels of patient risk, then applied added services to support those at highest risk.

Another provider using algorithms that include risk stratification is Penn Medicine, which is using employing a strategy to apply data to clinical care. In addition, Covenant Health said real-time analytics and risk stratification has helped it save a projected $1.8 million, so far, in 2015.

All patients at the hospital are offered text message reminders about appointments and recommended preventive services. However, higher risk patients often need more frequent and comprehensive follow-up care, as well as substantial social and behavioral health support, according to the article.

To that end, the health system added patient navigators and clinical pharmacists, as well as social workers and behavioral health consultants. For the highest risk patients, it also funded three high intensity clinics with small patient panels.

The study found strong leadership essential to such a project, as well as involving a range of stakeholders, including clinical directors and senior management, health services researchers, finance experts and IT developers.

While the project initially examined costs each person generated, over time it found costs alone were not the only item to research for tier assignments, and sought better alignment with clinical interventions.

"Adopting an intentionally iterative process from the outset reduced the pressure of getting it right the first time," the authors said. "Through trial and error, we learned which strategies work best to identify population tiers and which work best to trigger specific clinic action."

To learn more:
- read the article (.pdf)