How would the performance of hospitals, physicians and health systems compare if, hypothetically, they all had the same mix of patients?
That's a question that policymakers and many in the healthcare community raised when suggesting that performance measures would be more accurate if adjusted for the socio-demographic status of the patients being treated.
This type of risk adjustment involves a statistical approach that allows patient-related factors to be taken into account when computing scores on performance measures, thereby improving the ability to make fair and accurate conclusions about quality. Supporters of the idea point to a growing understanding throughout the healthcare community that social determinants significantly influence a person's health. We know that factors far outside the control of a doctor or hospital—patients' income, housing and education—can significantly affect patient health, healthcare and providers' performance scores.
The stakes of inaccurately assessing quality are raised, of course, when the results are used in pay-for-performance programs. With providers increasingly being paid based on the quality of their care, some say that those caring for the disadvantaged are being unfairly penalized. If measures are not adjusted to consider a patient's socio-demographic factors, they believe, we'll continue to create disincentives to care for the poor.
Opponents of adjusting measures for patient socio-demographic criteria, on the other hand, say it essentially sanctions delivering lower-quality care to already vulnerable patients. They worry that such adjustments could mask differences in quality and make meaningful information on social and economic disparities disappear. They say that adjusting measures in this way sets a different standard for providers who treat poorer patients and lowers expectations that they will improve.
At the center of this conversation is the National Quality Forum—which for more than 15 years has been the gold standard in endorsing measures. Reviewing and agreeing to measures through a multi-stakeholder process is not easy, and more often than not requires a critical blend of science and consensus. That was in evidence a year ago, when the NQF changed its rules to allow measures to be adjusted for patients who are poor, homeless, illiterate or have other socio-demographic risk indicators.
This change is significant, and it's in place for a two-year trial period. The trial was part of a compromise that the NQF brokered between providers—primarily hospitals—who said risk-adjustment was necessary for fairness, and others who worried it would disguise important gaps in quality.
The trial period was recommended by an expert panel composed of stakeholders with a variety of experiences related to outcome measurements and disparities. The recommendation was debated and approved by the NQF's board, which has a wide range of views represented among its directors.
Under the terms of the trial, all new measures submitted to NQF for endorsement after April 1 of this year are being assessed to determine if adjustment is appropriate. Measures endorsed prior to that date, but that are undergoing maintenance during the trial period, will also be considered fair game for adjustment.
There are other pathways for evaluating whether performance measures already endorsed should be reviewed for adjustment, including requests related to evidence of unintended consequences.
Some measures—including ones related to readmissions, as well as cost and resource use—are already being mandatorily reviewed as a condition of endorsement. If adjustment is determined to be appropriate in any of these cases, the NQF will endorse a measure with and without socio-demographic adjustment, as well as stratification for full transparency. We want the measurement process to be as flexible as possible for providers while also serving the best interests of patients.
After two years, we will evaluate the success of the trial and solicit feedback from stakeholders on its impact.
The National Quality Forum is, above all, a forum—so we take seriously our charge to listen to a full range of perspectives. Finding answers to difficult measurement-science issues such as risk adjustment, attribution and comparability will help us use outcomes when they are most needed to meet the needs of the healthcare delivery system. We believe the trial period enables us to move forward in a thoughtful way while producing data we can all learn from.