Dr. Christopher Dale on better ways of deploying quality metrics to improve the value of health care.
A physician in our medical group recently complained about the quality-based component of his pay: “I spent 2 1/2 hours last night looking at my compensation statement, and I found 12 patients that shouldn’t have been included [on my list],” he lamented. He’d narrowly missed his payout threshold and was convinced that his recorded quality performance was inaccurate. Such sentiments are new to Swedish Medical Group, a Seattle-based, 1200-provider multispecialty group that is part of Providence St. Joseph Health. In 2015, we began redesigning compensation to include performance on a specialty-specific quality metric in clinicians’ contracts. The process uncovered a latent hunger for compensation strategies that reflect clinicians’ complex impact on health and high-value care. It also revealed a threat to the joy and meaning of practicing medicine — and a potential path forward.
The traditional U.S. physician-compensation model provides incentives for the generation of visits and procedures, not better health outcomes, and has contributed to unsustainable growth in health care spending. The limits of such compensation are a driving force behind payment reforms being embraced by Medicare and commercial insurers. But rather than ceding reimbursement redesign to insurers, provider groups can lead the way with value-based clinician compensation.
The truism that “you get what you measure” appeals to anyone attempting to improve compensation. Measurement seems to be a way to align clinician, payer, and purchaser interests and increase efficient production of desired health outcomes. For example, Swedish began including in primary care providers’ compensation roughly 5% for improving performance or obtaining a threshold performance level in screening for breast, cervical, and colorectal cancers — and cancer screening increased (see table).
Yet the focus on metrics felt stifling to many clinicians, who found themselves scouring lists of potentially overdue patients, tracking down old records, and scanning them into the chart to get “credit” for tests and satisfy the compensation metric. Although more guideline-concordant care was being delivered (or documented), such improvement may stand on a continuum with “you get what you measure” at one end; at the other end lies the reality that “not everything that counts can be counted.”1
Measuring quality in terms of task-based care can diminish the value of clinicians’ essential role of deciphering medical complexity and building relationships. That clearly wasn’t our goal; our clinicians yearn to excel at their jobs and improve patients’ health. Unfortunately, current quality metrics assess the system’s ability to reliably deliver certain processes of care — a good thing, but not reflective of individual clinicians’ worth as compassionate diagnosticians and healers.
Our performance improved on the metrics we emphasized, but those metrics didn’t reflect the core, joy-producing essence of caring for patients. Provider compensation strategies built on current quality metrics cannot transform our health care system. So how can we chart a path forward?
First, we can measure what’s important to patients. The closer we get to measuring value — patient-centered outcomes divided by the costs of producing them — the more rational and efficient a health care marketplace we’ll create.2 Functional status, depression-free days, and total costs of managing a condition are all excellent value metrics. Organizations such as the International Consortium for Health Outcomes Measurement can help us develop more robust, patient-centered value measurements. Swedish tracks more than 300 quality metrics, most of which are based on what was measurable 10 or 20 years ago, not on what we can measure now. For example, when Washington State created the Common Measure Set for Health Care Quality and Cost in 2014, it used only currently available metrics.3 This recycling of metrics stifles innovation.
As payers and purchasers demand metrics more reflective of value, delivery systems will need to keep investing in data infrastructure to measure the health produced in an episode of care. Translating value metrics into clinician compensation is complicated, since medical services are responsible for only 10 to 15% of health, so many outcomes that are important to patients are beyond clinicians’ direct control.
We’ve found that recognizing clinicians’ indirect control over some outcomes and those outcomes’ importance to patients can lead to a compensation system that respects both parties’ interests. A successful strategy is to set achievable performance goals, acknowledge the importance of clinician professionalism, and use metrics to create a culture of excellence in patient-centered care. We’ve taken to saying that “the choice of metric determines the focus of attention; the threshold determines money in the clinician’s pocket.” Despite inherent complexity, aligning payers and providers on the goal of producing high-value care as measured by patient-centered outcomes can foster a new focus on the production of health.
Second, organizations can optimize their use of existing metrics by prioritizing those more tightly linked with producing health. For example, we considered the health effects of cancer-screening metrics and estimated that improving performance on cervical-cancer screening to the 90th percentile and maintaining that performance for a decade would save about 1 life; doing the same for breast-cancer screening would save about 2. But doing so for colorectal-cancer screening would save about 24. These estimates are rough, and producing health means more than averting cancer-related deaths, yet recognition of these effect sizes highlighted the opportunity to use such estimates to focus attention on what matters to patients. Not everything assessed by quality metrics has the same impact on health.
So when the local health department approached Swedish about running a human papillomavirus (HPV) vaccination quality-improvement program, the organization calculated the number of lives it could save. We could have agreed because it seemed like the right thing to do, while assigning the work moderate priority because HPV-screening rates aren’t a commonly used metric, but we calculated that improving screening rates from 40% to 80% would prevent at most four cancer-related deaths. Even this rudimentary approach to effect estimation is informing a richer conversation about allocation of resources and bandwidth.
Third, clinical groups can embrace quality-improvement science. Clinicians often feel that process metrics are a fairer basis for compensation than outcome measures because they reflect actions under clinicians’ direct control. On the continuum from “things that can be counted well” to “things that matter but are hard to count,” metrics such as number of patients screened per 100 opportunities lie at one end — closer to being under clinicians’ direct control and reflecting the whole team’s effectiveness. Balanced inclusion of such metrics along with outcome metrics can more fully capture delivery-system quality, but deploying them alone is not transformative and risks perpetuating a system that doesn’t produce health.
The Institute for Healthcare Improvement and others advocate a more sophisticated understanding of the production and improvement of health care. For example, bar charts of colorectal-cancer screening over time are nearly useless in identifying change and monitoring the production of care. Control charts, which use upper and lower control limits to identify boundaries for statistically significant changes over time, are better suited to this task for process measures of quality, yet they are infrequently used. We need continued investment in quality-improvement infrastructure and in our capacity to interpret and use quality data to drive meaningful change.
Fourth, we can acknowledge quality metrics’ limitations for compensation. Paying clinicians to improve quality scores rarely improves performance significantly and sustainably. But evidence suggests that behavioral-economics tools, including audit and feedback, transparent display of data, and making it “ridiculously easy to do the right thing,” can powerfully affect care delivery.4 Sustained focus on removing barriers to reliable production of high-value care, optimizing data analysis and use, and reducing workflow waste will improve health more than pay-for-performance schemes.
As practices shift from fee for service to a focus on person-centered, high-value health care, we can avoid creating systems that rob clinicians of the joy and meaning of partnering with patients to create health. Instead, we can use quality metrics wisely to create a better clinician-compensation program to drive the delivery of health. It’s not that we don’t want to manage what we can measure; we simply know that not everything that counts can currently be counted. And it’s time to learn how to count better.
1. Cameron WB. Informal sociology: a casual introduction to sociological thinking. New York: Random House, 1963.
2. Porter ME. What is value in health care? N Engl J Med 2010;363:2477-2481. Free Full Text | Web of Science | Medline
3. Washington State Common Measure Set for Health Care Quality and Cost. Olympia: Washington State Health Care Authority, December 2014 (http://www.hca.wa.gov/hw/Documents/pmcc_final_core_measure_set_approved_121714.pdf).
4. Meeker D, Linder JA, Fox CR, et al. Effect of behavioral interventions on inappropriate antibiotic prescribing among primary care practices: a randomized clinical trial. JAMA 2016;315:562-570. CrossRef | Web of Science | Medline
This Perspective article originally appeared in The New England Journal of Medicine.