For most of us, the essence population health management is focusing on patients who have already experienced serious adverse health events. But what if that doesn’t work? At least one writer suggests that though it may seem counterintuitive, the best way to reduce needless admissions and other costly problems is to focus on patients identified by predictive health data rather than “gut feelings” or chasing frequent flyers.
Shantanu Phatakwala, managing director of research and development for Evolent Health, argues that focusing on particularly sick patients won’t reduce costs nearly as much as hospital leaders expect, as their assumptions don’t withstand statistical scrutiny.
Today, physicians and care management teams typically target patients with a standard set of characteristics, including recent acute events, signs of health and stability such as recent inpatient admissions and chronic conditions such as diabetes, COPD and heart disease. These metrics come from a treatment mindset rather than a predictive one, according to Phatakwala.
This approach may make sense intellectually, but in reality, it may not have the desired effect. “The reality is that patients who have already had major acute events tend to stabilize, and their future utilization is not as high,” he writes. Meanwhile, health leaders are missing the chance to prevent serious illness in an almost completely different cohort of patients.
To illustrate his point, he tells the story of a commercial entity managing 19,000 lives which began a population health management project. In the beginning, health leaders worked with the data science team, which identified 353 people whose behavior suggested that they were headed for trouble.
The entity then focused its efforts on 253 of the targeted cohort for short-term personal attention, including both personal goals (such as walking their daughter down the aisle at her wedding later that year) and health goals (such as losing 25 pounds). Care managers and nurses helped them develop plans to achieve these goals through self-management.
Meanwhile, the care team overrode data analytics recommendations regarding the remaining 100 patients and did not offer them specialized care interventions during the six-month program. Lo and behold, care for the patients who didn’t get enrolled in health management programs cost 75% more than for patients who were targeted, at a total cost of $1.4 million. Whew!
None of this is to suggest that intuition is useless. However, this case illustrates the need for trusting data over intuition in some situations. As Phatakwala notes, this can call for a leap of faith, as on the surface it makes more sense to focus on patients who are already sick. But until clinicians feel comfortable working with predictive analytics data, health systems may never achieve the population health management results they seek, he contends. And he seems to have a good point.