Measuring What Matters in Health: Lessons from the Veterans Health Administration State of the Art Conference

Tracking performance lies at the center of any learning health care system’s efforts to improve patient health,¹ and is at the core of the reorganization of health care delivery resulting from the Affordable Care Act. Yet performance measurement is itself an intervention, sometimes with unintentional and unforeseen consequences.^{2 , 3} For example, the Veterans Health Administration (VHA) has long had a reputation as a leader in innovative clinical performance measurement,⁴ yet perverse performance incentives around performance on access-to-care measures may have exacerbated the VHA’s recent crisis of confidence.⁵ Even before that crisis, the VHA organized a state-of-the-art conference (SOTA) featuring thought leaders of national prominence to set goals for the next generation of clinical performance measures. The conference explored ways to make performance measurement more effective. This special issue of JGIM was motivated by the SOTA conference, so that innovations in performance measurement stemming from VA research activities could be brought together in one thematic issue. While other editorials will explore the meaning of the resulting manuscripts, here we report on the SOTA conference’s practical recommendations to enhance the value of measurement that systems across the country can consider as they ramp up their performance measurement activities.

We propose three key challenges for improving measurement: 1) making clinical measures more meaningful, 2) putting patient preferences at the center of the measurement effort, and 3) incorporating the value proposition into the measurement portfolio. Measures that are clinically meaningful are essential to improving the appropriateness of care that we deliver. Too often performance measures have been blunt instruments that were not tailored to individual patient circumstances, sometimes leading to overuse of unnecessary or even harmful services⁶ for some patients. One way to make measures more meaningful is to tailor them to the individual patient likelihood of benefit for the treatment or procedure being considered.⁷ Doing so puts the focus on appropriateness of both providing services when the benefit is likely to be high, and not providing services when the benefit is low (or negative). Such a “next-generation” clinical measure would incorporate the individual patient benefit of a service in deciding whether the service was indicated. For example, instead of basing performance for colorectal cancer screening on age alone, a new performance measure could base screening recommendations on life expectancy, risk from the screening procedure, and future cancer risk.⁸ Such new measures could examine both underuse of screening in patients who are likely to benefit, and overuse among those in whom the procedure could cause harm. Similarly, a measure focused on diabetes treatment could promote different glycemic control goals based on the probability that tight control could result in long-term benefit (i.e., prevention of future neuropathy in younger patients without current complications) or harm (i.e., hypoglycemia in older patients with multimorbidity).

Indeed, a common theme within the SOTA conference discussions was how best to put patients’ needs, perspectives, and opportunities for benefit at the center of performance measurement. Beyond tailoring, measures must be transparent and comprehensible to a wide range of stakeholders, including the patients themselves. This could mean annotating measures with an estimate of the likely magnitude of effect in terms that patients can understand, perhaps using standard number-needed- to-treat (NNT) statistics augmented with visual aids. True patient-centered measures should be paired with decision aids so that patients or their surrogates can decide the extent to which the measures apply to them when choosing among the measured entities. Conference participants also felt strongly that patient preferences and goals should be incorporated into the measures themselves, a sentiment echoed in a recent NEJM commentary.⁹ Perhaps most importantly, patients need to be included in the measure development process, as organizations like the National Quality Forum have already begun to do.

Patients, though paramount, are not the only stakeholders in performance measurement. As more and more care is delivered in cost-conscious contexts like accountable care organizations, the need for measures that incorporate value to payers and society will only increase. Incorporating value into performance measurement requires a common, or at least a transparent, definition of value. The Porter definition of incremental outcomes (numerator) divided by incremental resource use (denominator) is appealing,¹⁰ but outcomes are often remote from the processes most under the control of the health care system and providers. In these cases, using process quality in the numerator as a proxy for outcomes will often suffice if other evidence supports the link between the process and desired outcomes.¹¹ In either case, to be useful, measures of value should simultaneously assess both the numerator and denominator in the same population. For example, if outcomes within a health system are being assessed only for patients with diabetes, resource use should also be assessed for those same patients (rather than for the entire health system’s population). Value can be measured from the patient, system, or societal perspective; all are useful, and measures should be explicit as to which is intended. Implementation of value measures should guide resources away from low-value yet appropriate care, regardless of whether it is inexpensive (e.g., tracking glycosylated hemoglobin in older multi-morbid patients) or costly (e.g., MRI for low back pain without motor symptoms), and should drive resources to high-value care, even if it is expensive (e.g., antivirals in HCV).

Measurement is, and likely always will be, an imperfect science. However, there are principles that we can follow in developing and applying measures that will best promote the goal of fostering learning health care systems and minimize unintended consequences. SOTA participants felt strongly that better measures should be more clinically meaningful, personalized and patient-centered, and focused on value. Measures are not ends in themselves, though. They must also directly light the way to quality improvement, which means developers should link them to likely implementation strategies. Measures should also be linked and integrated with other interrelated measures, so that improvement bundles can have the greatest effect. Following these principles, and keeping the patient at the center of the task, can point the way to the next generation of performance measures.