Newtown, CT – March 14, 2014
What's the difference between unpredictable variation and unwarranted variation: Solvency – Think of it as the tale of two tails. The distribution of total costs per year for an average commercially insured population includes some plan members with no claims experience, some with moderate to high, and some with a lot. Since the left side of the distribution is locked at the intercept of $0, all the unpredictable variation in costs is on the right side. This includes the occurrence of rare and complicated diseases and conditions that are expensive to treat and whose costs vary significantly from individual to individual, not because of provider preferences, but because of the inherent variability in the effects of the treatments on individual patients. It also includes babies born before full term who require long stays in the neo-natal intensive care unit. In the middle of the distribution lie common chronic conditions, as well as acute and procedural medical episodes of care. Our prior work has shown that there is a significant amount of unwarranted variation in the costs of those medical episodes and that much of it could be ascribed to deficiencies in patient management – what we've called potentially avoidable complications. The difficulty to-date in understanding health care cost variation, and therefore better explaining the opportunities for cost reduction, is that the tails get in the way of the tale.
What this means to you – If we exclude the low costs plan members and the ones above a certain percentile (say the 90th) we're left with the bulk of expenses and a better way to create accountability for unwarranted variation. It's one of the reasons we've referenced in past missives the warnings of some researchers about the necessity for very large sample sizes when contracting with health systems for total costs of care. While the warnings are real and need to be heeded, they can impede the potential for moving the dials on value based purchasing. And that's why we're suggesting an alternative: lop off the right tail. First off, you don't want to create incentives for rationing care for patients whose conditions are, by their very nature, difficult to treat and require significant resources. Second, the tail wags the rest of the distribution. Small swings in the frequency of these events, and their inherent variability (which cannot be adjusted by any technique because the underlying sample sizes are small to begin with) can shift total aggregate costs of care for a population by a few percentage points. That can be the difference between profit and loss for any provider taking on total cost of care risk. However, with the tail lopped off the focus of cost management can finally turn to the causes of unwarranted variation: price differences between providers for the same services; service mix differences usually caused by the occurrence of potentially avoidable complications; and volume of services differences often resulting from overuse of routine services. It becomes a tale of effectiveness of care management and efficiency in use of services, for which the accountability rests with providers and patients. At that point, profit or loss is no longer based on the random swings of a tail, but on the competence of providers. Of course, that also means their incentives (and that of plan members) can't get in the way, but that's a tale for another day.