Skip to main content

Predicting High-Cost Patients

By June 3, 2019Commentary

Okay, once again, a basic truth about health spending is that a small segment of the population accounts for most health spending.  The top 5% of people in health spending account for around 50% of all spending, and conversely, the bottom 50% in spending represent less than 5% of the total.  So, can we identify those people who are likely to be high spenders and manage their care in a way that reduces costs, or potentially even avoids acute disease episodes.  There are a lot of analytics companies out there claiming to do that.  One such approach was tested in a recent study.  The results were published in the Journal of Health Economics.   (JHE Article)   The study focused on Medicare Advantage members with congestive heart failure at a large health plan that had a proprietary algorithm that created a risk score for the members.  The higher the score the more likely that the member experienced gaps in care and was at higher risk of hospitalization.  A care management team then reached out to the members most at risk to intervene.  The primary goal of the program was to reduce ER visits and hospitalizations.  The study assessed both the accuracy of the algorithm and the effectiveness of the intervention program.

The algorithm used over 500 patient data points.  The intervention was primarily telephone outreach by nurse coaches, who then coordinated with the member’s physicians and worked on other issues affecting health and health care for the member.  The risk score calculated by the algorithm did seem to be an effective predictor of the likelihood of either hospitalization or an emergency room visit within several intervals in the year after the algorithm was run.  The care intervention program appeared to reduce ER visits during those intervals, with the probability of such a visit decreasing by 12 percentage points.  There was also a reduction in cardiologist visits of 14%.  But there was no statistically significant change in the likelihood of a hospitalization or of a primary care physician visit, although hospital admissions did appear to decline.  The authors did not estimate the cost of the program or the total cost savings, so we don’t know if had a net cost benefit.  The study does support the idea that finding general high-risk patients may be done effectively, but designing an effective intervention to improve care and avoid costly episodes is harder.  Telephonic interventions may not always be the best method.  And the results may be limited if patients aren’t truly motivated to pay better attention to their health needs.

Leave a comment