Finding the Right Patients for Disease Management: They're in the Middle!

The Disease Management Care Blog was a speaker at a recent health care conference where it discussed the "targeting" of care management programs. It doesn't believe that every patient with a chronic condition will benefit from disease management. The audience agreed, so the DMCB will try it out here.

Apostasy you say? Read on.

This Kaiser study showed that patients with low risk heart failure didn't benefit from "telephone mediated nurse management." In the meantime, this Medicaid study said that patients with high risk heart failure didn't benefit from disease management. What gives?

From a pure economic perspective, if a typical health insurance plan arrays its membership by cost from low to high on the X axis and examines the numbers of persons in each cost category (the Y axis), the curve will be low on the left (persons who are never never seen have low cost and are relatively few in number), high in the left-of-middle (low to modest charges are typical of most persons with health insurance), and then have a low tail going off the right (persons with very high claims are few in number).

Stare at the graph for awhile and it will make sense:

While the absolute costs will vary, the same graph applies to most subpopulations in a health insurance plan, such as children, persons living in rural areas or even elderly duck hunters. It also applies to subpopulations defined by the presence of diabetes mellitus, chronic heart failure, or asthma, which will have the same pattern of low cost-low numbers (left), moderate cost-high numbers (toward the middle-left) and some will be very high (way off to the right)

So, if the idea is to save money, which patients should you aim your care management programs at?

Check out the next graph:

Patients who are in left side of the population have low baseline costs, so it's unlikely that care management will lower their costs any more than they already are. On the other hand, patients who are way on the right are unlikely to have illness that is meaningfully amenable to education, engagement, behavior modification, self-care, counseling, more frequent primary care visits or care plans. They're going to need specialists, they're likely to have to go to the emergency room frequently and they're going to have to be hospitalized a lot.

Rather, the DMCB thinks that there is a population toward the middle that are sick enough to warrant care management and have baseline costs than can be reduced. They're the sweet middle and that's where disease management programs should be directed. The Kaiser study focused on patients too far to the left, while the Medicaid study focused on patients off to the right.

Three additional points:

1) While the graphs above portray a health plan's current population, the purpose of predictive modeling is to array the future risk of a population using the same logic.

2) This also demonstrates how an insurance benefit - like access to disease management programs - shouldn't necessary be uniformly applied to everyone that would nominally qualify. It also shows why many of the quality measures in use today are imperfect because their binary "black/white" methodologies don't distinguish between populations that are likely to benefit the most.

3) This is another Achilles heel for the Patient Centered Medical Home (PCMH). While physicians may be able to intuit which patients are most like to benefit from primary care based, team-delivered and patient centered care, the same logic described above applies to a population assigned to a primary care site. Not all patients will detectably benefit from the PCMH, and it's important to know who will and who won't.

P.S. The powerpoint slide is available to anyone who asks. Just email (the address is off to the right somewheres).