Letter to the Editor: Quality Measurement for Seriously Ill and Dying—The Denominator Problem

Dear Editor:

As baby boomers age, concerns grow about the high costs of health care and the solvency of the Medicare Trust Fund. Federal policy has focused on alternative payment models that promote value through constraining costs and maintaining or improving care quality. This is an important change from the incentives of traditional Medicare, where profits are generated by doing more, to a system where profits are generated by doing less. The shift to value-based care increases the importance of quality measures to ensure that appropriate care is delivered, particularly to those with serious illnesses who have high care needs.

Despite their importance, the population with serious illness is relatively small: a challenge for quality measurement that we term the denominator problem. While value-based organizations like Medicare Advantage (MA) plans now care for the majority with serious illness, those with serious illness comprise a small portion of beneficiaries. Therefore, measurements of the “average” beneficiary may not reflect the experience of those with serious illness. Compared with those with serious illness, healthier beneficiaries do not have adequate interactions with the health care system to test the ability to provide person-centered care. Efforts to focus on sicker subpopulations are stymied by the requirement that measurements have a high interunit reliability (IUR). The IUR, a statistic that measures the true difference between programs, is highly sensitive to population size. For example, despite a congressional mandate and the growing enrollment of people in MA dual-eligible Special Needs Plans (D-SNPs), the Medicare Payment Advisory Commission (MedPAC) ¹ has been unable to report on quality because of the small size of D-SNPs.

Table 1 proposes solutions to improve the use of the IUR and address the serious illness denominator problem. In part, these recommendations are drawn from the National Quality Forum report on rural health ² and Elliott et al. ³ The IUR could be revised by using a statistic called the exceedance probability to contextualize the level of certainty for measurements without sufficient IUR. Similar to the weather forecast reporting the probability of rain, it reports the probability that a program performs lower than average. While imprecise, this could provide important information. Two potential solutions address concerns over the “average” Medicare beneficiary dominating quality measures: the use of weighting to increase the contribution of the seriously ill population to overall scores, as has been proposed in equity weights ⁴ and oversampling those with serious illness to ensure their representation. Disenrollment from lower-quality programs can further reduce the denominator. For hospice, this could be addressed by a brief live discharge survey on the experience of those disenrolling from hospice, while for MA plans, a flag could highlight high rates of disenrollment to indicate to consumers potential quality concerns and a biased denominator.

As we change the financial incentives of payment models, we must address the denominator problem. While those with serious illnesses are the minority, they best test the ability of our health care systems to provide high-quality care and they are the most at risk for adverse effects of changing financial incentives.