Cropped_new_image

Insights from GNS Healthcare

Deciphering Exchange Populations with Machine Learning

Posted by Colin Hill on Aug 5, 2015 12:30:00 PM

Miscalculating the exchange risk equation has cost health plans dearly.One insurer posted losses of more than $140 million on individual market policies when its medical costs exceeded premiums by nearly 15 percent as Reed Abelson reported in The New York Times.

Those that underestimated medical costs or overestimated the number of new enrollees now are at risk of overcompensating. Proposed premium increases, some as high as 85 percent, are based on the same guesswork that caused miscalculations to begin with.

Health plans recognize that not every member who enrolls via the exchange will incur high medical costs, such as those driven by emergency room visits and hospital admissions. Despite substantial investments in data resources and technology, identifying the new or returning members whose healthcare needs will contribute disproportionately to total cost eludes even the most sophisticated analytics teams.

Why is it so difficult to identify high-cost members?

One factor is churn. As Abelson points out, nearly one in three consumers (29 percent) who re-enrolled in 2015 selected a different policy. The challenge is magnified by the nature of claims data, which is a lagging indicator of risk and cost. The effect is especially profound for Medicare and Medicaid members since enhanced reimbursement under Hierarchical Condition Categories (HCC) classification can be delayed.

Lack of information, or delayed information, carries greater risk as the number and cost of interventions continues to rise. Abelson’s article points to the high cost of new therapeutic drugs, such as for hepatitis C, as an example. Indeed, health plans are spending heavily on many types of interventions, including population health and disease management programs.

How can these knowledge blind spots be eliminated?

Health plans have tremendous resources and an ongoing opportunity to learn about new members. Even when claims data is lacking, such as in the new member onboarding process, other forms of data enable actionable insights when coupled with a unique and highly sophisticated form of analytics called “machine learning.”

Within a month of enrollment, for example, it is possible to identify which individual members are most likely to have a chronic disease. This knowledge allows population health leaders to select members for intervention on a prioritized basis, giving health plans an immediate opportunity to affect individual behavior and begin to expand their understanding of new members. As time goes on, more data becomes available on claims and engagement patterns, driving more refined insights from the machine learning platform.

This approach leverages principles of precision medicine to eliminate uncertainty and guesswork from health plan operational decisions, giving payers the confidence to price products strategically in support of market growth strategies.

GNS is a pioneer in the precision medicine movement and has been applying its patented machine learning technology for more than 15 years. In that time, we have helped health plans, healthcare providers, pharmaceutical companies, researchers, and others turn knowledge blind spots into actionable intelligence, improving health and delivering a competitive advantage.

 New Call-to-action

Topics: Pop Health, Health Plan, Payer, New Member Exchange