Disease management outcomes measurement requires mathematical proof

The Disease Management Purchasing Consortium (DMPC) and Zenger Analytics have completed a mathematical proof to address the question of valid outcomes methodologies for disease management, wellness, medical home and other population-based care management programs for which outcomes are calculated on a pre/post basis.

Math is not a consensus-based discipline. It is a proof-based discipline, as it should be for outcomes measurement. One plus one equals two not because a committee or some benefits consultants or actuaries agree it does, but because it’s a proven mathematical fact.

The customary pre/post measurement approach used when no prospective control study is feasible counts savings by comparing participant or enrollee experience prior to program implementation against experience after program implementation. Identification for the program is usually triggered by a claims algorithm, or in the case of wellness program, a high score on a health risk assessment.

Using a population identified in that manner as the “pre” population will result in an overstated return on investment (ROI). However, it is used by 80% of health plans and probably 90% of self-insured employers. A small minority of payers use valid metrics and are able to capture true ROI.

Consider a payer with two known asthmatic members. Member 1 has a 2009 baseline of $2,000 in claims, then $1,000 in claims for 2010, under a contracted program. Member 2 has no claims in 2009, the baseline year, followed by $2,000 in claims for 2010 under the program.

Member 1 has claims in the baseline. Member 2 doesn’t. In this case, the average cost per asthmatic increases 50%, from an average of $1,000 in 2009 to an average of $1,500 in 2010.

Now remove the unrealistic assumption that the health plan has identified Member 2 who had no claims during the baseline period. Unless Member 2 triggers the claims algorithm with a diagnosis or multiple prescriptions during the baseline year, Member 2 will not be identified as an asthmatic in the baseline.

So asthmatics who are new to the plan, non-compliant, not diagnosed, diagnosed at their previous health plan, or fail to fill enough prescriptions to be counted in the population are not included in the baseline.

Because the health plan has not identified Member 2-the member with no claims in the baseline year-the baseline average cost per asthmatic is calculated as $2,000. Thus the contract year-with both members identified-shows a 25% reduction in average cost per asthmatic, from $2,000 to $1,500. In this case, the costs of treating asthma actually rose by 50%.

In wellness, the same thing happens. Risk factors fluctuate, but often only the people whose risks were high enough to qualify for coaching are measured.

One carrier has even created a guarantee around this natural risk migration, promising that people who are high-risk will show improvement without taking into account that people with low risks may deteriorate. Using such math-essentially the same math as the second example above for the asthmatic members, but for wellness-gives the same invalid result.

Suppose your population smokes, but also quits in annual cycles. In any given year, half are smoking but half have quit. In the next year half still smoke, but the smokers have quit while the quitters resume smoking. This pre/post math, which focuses only on the outcomes of people identified as smokers, yields a 100% quit rate every year, even as the actual smoking rate remains constant.

The disease management industry has been haunted by the challenge of quantifying its value. Payers want easy-to-understand measures to show them that their disease management dollars are well spent.

As the American novelist Upton Sinclair said, it is impossible to prove something to someone whose salary depends on believing the opposite. Nonetheless, it does not matter that some people disagree. Math is not a consensus-based discipline.

Al Lewis is president of the Disease Management Purchasing Consortium.