Authors: Catherine H. MacLean, M.D., Ph.D. et al
This article was published on April 18, 2018, at NEJM.org.
Performance measurement in the U.S. health care system has expanded dramatically over the past 30 years. The National Quality Measures Clearinghouse now lists more than 2500 performance measures. These measures are used in various quality-reporting, accountability, and payment programs sponsored by commercial payers, government agencies, and independent quality-assessment organizations. The Centers for Medicare and Medicaid Services (CMS) aims to base 90% of Medicare fee-for-service payments to clinicians on “value” by the end of 2018 by using performance scores.
Although most physicians view the delivery of high-quality care as a professional imperative,1 performance-measurement activities face increasing resistance from physicians and some policymakers who believe that current measures are not meaningful.2 In a recent survey, 63% of physicians said that current measures do not capture the quality of the care that physicians provide.3 Yet U.S. physician practices are spending $15.4 billion each year — about $40,000 per physician — to report on performance.3
In response to these concerns, the Performance Measurement Committee (PMC) of the American College of Physicians (ACP) developed criteria to assess the validity of performance measures. Using a modified version of the method developed at RAND and UCLA for evaluating the benefits and harms of a medical intervention, we applied the ACP criteria to the measures included in the Medicare Merit-based Incentive Payment System (MIPS)/Quality Payment Program (QPP). We hypothesized that if most of the MIPS/QPP measures assessed were deemed valid using this process, physicians could have more confidence that adherence to the measured practices would result in improved patient outcomes. Conversely, if some substantial proportion of the measures were deemed not valid, the results would suggest the need to change the process by which MIPS measures are developed and selected.
ACP MEASURE REVIEW CRITERIA.
DOMAIN 1. IMPORTANCE
- Meaningful clinical impact: Implementation of the measure will lead to a measurable and meaningful improvement in clinical outcomes.
- High impact: Measure addresses a clinical condition that is high-impact (e.g., high prevalence, high morbidity or mortality, high severity of illness, and major patient or societal consequences).
- Performance gap: Current performance does not meet best practices, and there is opportunity for improvement.
DOMAIN 2. APPROPRIATE CARE
- Overuse: Measure will promote stopping use of a test or treatment in general population or individuals where the potential harms outweigh the potential benefits.
- Underuse: Measure will encourage use of a test or treatment in general population or individuals in whom the potential benefits outweigh the potential harms.
- Time interval: Time interval to measure the intervention is evidence-based.
DOMAIN 3. CLINICAL EVIDENCE BASE
- Source: Evidence forming the basis of the measure is clearly defined with appropriate references.
- Evidence: Evidence is high-quality, high-quantity, and consistent and represents current clinical knowledge.
DOMAIN 4. MEASURE SPECIFICATIONS
- Clarity — numerator and denominator clearly defined:
- • For process measures, numerator includes a specific action that will benefit the patient, and denominator includes well-specified exclusions.
- • For outcome measures, numerators detail an outcome that is meaningful to the patient and under the influence of medical care.
- • Denominator includes well-specified and clinically appropriate exceptions to eligibility for the measure.
- Clarity — all components necessary to implement measure clearly defined
- Validity: The measure is correctly assessing what it is designed to measure, adequately distinguishing good and poor quality.
- Reliability: Measurement is repeatable and precise, including when data are extracted by different people.
- Risk adjustment: Risk adjustment is adequately specified for outcome measures.
DOMAIN 5. MEASURE FEASIBILITY AND APPLICABILITY
- Attribution: Level of attribution specified in the measure is appropriate (measure ties the outcomes to the appropriate unit of analysis) and is clearly stated.
- Physician’s control: Performance measure addresses an intervention that is under the influence of the physician being assessed.
- Usability: Results of the measure provide information that will help the physician to improve care.
- Burden: Data collection is feasible and burden is acceptable (low, moderate, or high)
Of 271 measures in the 2017 QPP measures list, we identified and rated the validity of 86 that the committee considered relevant to ambulatory general internal medicine. Among these, 32 (37%) were rated as valid by our method, 30 (35%) as not valid, and 24 (28%) as of uncertain validity. We also determined the proportion of the measures that had been developed by the National Committee for Quality Assurance (NCQA) or endorsed by the National Quality Forum (NQF) that were rated as valid by our method. As compared with measures that were not endorsed by these organizations, greater percentages of NCQA-developed and NQF-endorsed measures were deemed valid (59% and 48%, respectively, vs. 27% for nonendorsed measures), and smaller percentages were deemed not valid (7% and 22%, vs. 49% for nonendorsed measures).
For each measure, the committee rated validity with respect to five domains: importance, appropriateness, clinical evidence, specifications, and feasibility and applicability.
Notably, among the 30 measures rated as not valid, 19 were judged to have insufficient evidence to support them. For example, MIPS measure 181, “Elder Maltreatment Screen and Follow-Up,” requires the completion of the Maltreatment Screening tool on the date of an encounter and a documented follow-up plan for all patients 65 years of age or older. Although elder abuse is a serious problem that physicians should appropriately diagnose and report, the U.S. Preventive Services Task Force has found insufficient evidence to warrant routine screening. We believe the substantial resources required to screen large populations of elderly patients for maltreatment and to track follow-up would be better directed at care processes whose link to improved health is supported by more robust evidence.
Another characteristic of measures that were not rated as valid by our method was inadequately specified exclusions, resulting in a requirement that a process or outcome occur across broad groups of patients, including patients who might not benefit. MIPS measure 236, “Controlling High Blood Pressure,” for instance, requires that a blood pressure of 140/90 mm Hg or lower be achieved in the clinic setting for all patients. Forcing blood pressure down to this threshold could harm frail elderly adults and patients with certain coexisting conditions.
We also identified measures that were directed at important, evidence-based quality concepts but had poor specifications that might misclassify high-quality care as low-quality care. For example, MIPS measure 009, “Anti-depressant Medication Management,” assesses whether patients who started taking an antidepressant medication continued taking one at 3 and 6 months after initiation. This measure does not consider patients’ reasonable preferences for switching to alternative, evidence-based interventions such as psychotherapy or electroconvulsive therapy after experiencing side effects of antidepressants.
Our analysis identified troubling inconsistencies among leading U.S. organizations in judgments of the validity of measures of physician quality. Although the ACP assessment was limited to a defined set of measures, that set was large and included the vast majority of measures that will be applied to ambulatory care internists as part of the United States’ largest physician quality-assessment program for the purpose of accountability. Our findings are striking given that the criteria we used were similar to those used by NQF and CMS. Why the disconnect?
Possible explanations include the methods used to assess measures and the characteristics of the experts who did the assessing. The RAND–UCLA appropriateness method does not classify measures as valid when there are significant disagreements among the panelists. In contrast, the NQF threshold for endorsement is close to a simple majority of panelists (60%). The ACP method thus sets a higher standard for validity. In addition, we would argue that the RAND–UCLA method can be considered more evidence-based than other methods, since favorable clinical outcomes have been demonstrated for patients treated according to standards developed with this method.4,5
It is also possible that the perspectives of the groups doing the rating contribute to differences in validity ratings. Specifically, NQF convenes multistakeholder groups, whereas the ACP committee is composed exclusively of physicians with expertise in clinical medicine and research. However, analyses of the RAND–UCLA method in which multiple panels were convened to rate identical criteria have demonstrated high levels of agreement across panels for necessary care. Hence, although changing the panel composition might result in some differences in ratings, we would not expect the variation to be large enough to explain why so many NQF-endorsed measures were rated as not valid by the ACP committee.
The fact that only 37% of measures proposed for a national value-based purchasing program were found to be valid with a standardized method has implications for physician-level performance measurement. The use of flawed measures is not only frustrating to physicians but also potentially harmful to patients. Moreover, such activities introduce inefficiencies and administrative costs into a health system widely regarded as too expensive. If developers, assessors, and public and private payers adopted a more rigorous method of assessing measures’ validity, potential problems could be identified before the measures were launched. It makes sense for practicing clinicians to participate in the development and review of measures. At the same time, a single set of standards (like those put forth by the National Academy of Medicine for clinical practice guidelines) could be developed that would allow others to evaluate the trustworthiness of performance measures.
We believe that the next generation of performance measurement should not be limited by the use of easy-to-obtain (e.g., administrative) data or function as a stand-alone, retrospective exercise. Instead, it should be fully integrated into care delivery, where it would effectively and efficiently address the most pressing performance gaps and direct quality improvement. For now, we need a time-out during which to assess and revise our approach to physician performance measurement.
Qaseem A, Snow V, Gosfield A, et al. Pay for performance through the lens of medical professionalism. Ann Intern Med 2010;152:366–369.
Berwick DM. Era 3 for medicine and health care. JAMA 2016;315:1329–1330.
Casalino LP, Gans D, Weber R, et al. US physician practices spend more than $15.4 billion annually to report quality measures. Health Aff (Millwood) 2016;35:401–406.
Higashi T, Shekelle PG, Adams JL, et al. Quality of care is associated with survival in vulnerable older patients. Ann Intern Med 2005;143:274–281.
Hemingway H, Crook AM, Feder G, et al. Underuse of coronary revascularization procedures in patients considered appropriate candidates for revascularization. N Engl J Med 2001;344:645–654.