Editor’s note: This is one of several posts Health Affairs Blog will publish stemming from sessions at the June 2015 AcademyHealth Annual Research Meeting (ARM) in Minneapolis. Watch Health Affairs Blog for additional posts on topics raised at the ARM.
We’re still looking for the right levers to drive what we all seek from our health care system — a system that helps people optimize their health and is affordable. Fundamentally, if we could better influence the health-related decisions that take place between a clinician and a patient, we’d be making significant strides towards the triple aim. Every day billions of such decisions take place, which translate into the use of various services and health impacts. In the aggregate, these decisions eventually add up to our national health profile, our health care system quality scorecard, and our national health care expenditures.
A number of forces influence these decisions: regulations affecting who can deliver care; which providers and patients are covered by insurance and what is covered; payment policies affecting what is reimbursed and how; and public transparency efforts. The organizational context in which care is delivered also has a huge influence through the culture, norms, processes, and resources available. And individual providers and patients have their own biases and preferences that factor in as well.
Over the past decade, policy has focused and experimented extensively on two strategies for improving provider performance: increasing the availability of performance data and tying financial incentives to performance. That makes sense; according to well-validated models of behavior change, people are motivated to act by factors including knowledge (data), the ability to access the resources necessary to act, and the right incentives to support their actions (financial incentives). This policy focus also aligns with the rational choice theory of decision making, in which providing people with information and incentives to act should induce more of the desired behaviors. So why have we failed to make real progress in our use of these drivers of change?
One reason is that most of our approaches to transparency and payment incentives have ignored the fact that people are not “purely” rational decision makers. That is, people may be unlikely to respond to simplistic “carrot and stick” reductionist models, or may respond to them in unintended ways.
This is where the science of behavioral economics could inform health care practice and policy going forward. With his best-selling book, Predictably Irrational, Dan Ariely has piqued our collective interest by pointing to decades of work on cognitive biases that affect human beings. In fact, behavioral economic principles are already catching on in the public policy sphere; in 2013, the White House launched its Social and Behavioral Sciences Team (SBST), a group of experts charged with harnessing knowledge from behavioral economics to inform policy. The UK and Australia also have behavioral insight or “Nudge” units.
But not much attention has been paid to using behavioral economics to improve clinician and patient decision making, thereby motivating and supporting behaviors that would lead to improved health outcomes. To date, what little has been done has largely focused on consumer and patient behaviors, particularly through participation in wellness programs or treatment adherence. We are just starting to see the use of behavioral economics to influence clinician behaviors.
We thought it would be revealing to review our past and current approaches to transparency and financial incentives with a behavioral economics lens, uncover why many of these initiatives have failed to reach the goals pursued, and then discuss how applying behavioral economics principles could inform more effective designs in the future.
Limits Of Data Alone
Despite evidence to the contrary, we’ve continued to assume that the simple act of providing data will drive action. This has led to a proliferation of reports about health system performance on quality and costs at all levels—national, state, regional, and organizational—either in print, or more recently online and even with sophisticated interactive functionalities. Yet behavioral economics teaches us that how that information is presented is critical in determining its effects: there has to be a compelling story, a narrative that draws the user in enough to galvanize action. A dictionary may have millions of words, yet it can never tell a story. Furthermore, how much information is presented, and when the information is presented, are also important.
Today there are over 1,700 measures of quality, and although we do have a mechanism to endorse and harmonize measures, they keep proliferating and there is still no clear narrative around them. Their sheer number creates choice overload. Faced with so many measures about their performance, clinicians are hard pressed to know where to focus; scorecards, report cards, and practice profiles end up paralyzing clinicians rather than motivating them, or they just get ignored as clinicians try to keep up with more salient patient and practice concerns.
The good news is that we’re now seeing greater attention to simplification, and to alignment of key measures of performance. The Core Quality Measures Collaborative (members include America’s Health Insurance Plans, Centers for Medicare and Medicaid Services, and the National Quality Forum, as well as national physician associations) was formed in 2014 with a goal of reducing, refining, and relating core measure sets to those that matter to clinicians, patients, and payers. The Institute of Medicine recently released its recommendation for a core set of 15 standard metrics. Many state payment and delivery transformation initiatives also feature multi-stakeholder alignment and parsimony as key principles.
There is good evidence that public reporting has been effective at influencing providers to improve. This could be explained by the role of social comparison in effecting change. Behavioral economics research suggests that people are concerned about how they compare to their peers, and that comparative data can motivate change. For example, sending letters to heads of households to inform them about their energy consumption compared to that of their neighbor led to reduction in energy use. The operative concepts here are peer groups and social context.
Peer comparative reports have been found to be four times more effective compared to profit incentives in improving mortality rates by cardiac surgeons performing coronary artery bypass grafts in Pennsylvania. Other approaches to exploit social relatedness effects are being tested. For example, Minneapolis’ Fairview Health Services is linking provider incentives to group performance; in an evaluation of this compensation model, most primary care physicians reported a shift in orientation toward quality and working more collaboratively with their colleagues. The majority reported that their quality of care had improved as a result of the new incentive model, as had that of their colleagues.
With the proliferation of public reporting around the country and the Medicare pay-for-performance initiative that will deliver comparative report cards to every Medicare physician by 2017, there will be substantial opportunities to design effective reports on performance that exploit the power of social comparisons.
Goal Gradient And Threshold Effects
The way incentive programs set goals can be a critical determinant of their success. Research suggests that people will work harder as they get closer to a reward. Following this logic, low-baseline performers may not bother to engage if the bar is set too high. On the other hand, evidence also suggests that people’s motivation may decrease after they reap a reward.
The Alternative Quality Contract (Blue Cross Blue Shield of Massachusetts) has been successful at titrating goals and rewards. The quality goals for physician practices are set in absolute, as opposed to relative, terms. Quality and efficiency are linked so that clinicians are motivated to focus on both aspects of care delivery. There are five performance gates, and the relationship between quality scores and rewards is S-shaped: a one-unit increase in aggregate score generates a bigger quality bonus increase for groups around the middle of the performance range, as opposed to the top or bottom.
Other programs based on relative targets, or changing benchmarks (tournament style, in which many providers are essentially competing against each other), such as the Medicare Shared Savings Program and the Physician Value-Based Modifier Program, may create unintended behaviors based on the principle of loss aversion. Faced with uncertainty, people overestimate their losses and adopt inappropriate behaviors in the attempt to minimize loss.
Timing, or when information or incentives are provided, is also important. Behavioral economics points to the fact that people value rewards significantly more when they are obtained immediately. The clear implication of immediacy in designing physician incentives is to eliminate the time between behaviors and rewards. Today, this is still a huge challenge given the significant data lags that link performance to rewards or penalties.
For example, the Hospital Readmission Reduction Program penalizes hospitals with relatively higher rates of Medicare readmissions. Yet the penalties in 2015 are based on readmission rates for 2010-2013. Such lags are the rule rather than the exception and point to a clear need for solutions for more timely data. Many organizations are developing data analytics capacity and experimenting with clinician access to their own performance data as close to just in time as possible. Advocate Healthcare (Pankaj Patel, Advocate Healthcare, panel presentation, Academy Health Research Meeting, June 14th 2015, Minneapolis) has created a smart registry that allows immediate feedback on performance, and clinicians can easily view performance relative to their targets, and can then drill down to individual patient levels to set in place management plans that are responsive and fit patients’ clinical situations.
People tend to work harder to avoid losses than they do to achieve gains of similar amounts. This behavioral economics principle is increasingly being leveraged by policymakers and delivery system leaders in the design of various provider incentive programs. For example, the Massachusetts General Physicians Organization implemented an “advanced payment program,” essentially pre-paying bonuses but placing the clinicians at risk of having to pay money back depending on their performance. Prepaying the bonuses takes advantage of loss aversion, as clinicians will not want to give up something they already have.
In the past few years, the nation has seen significant reductions in readmission rates, an area that had not seen any improvements over decades. One could hypothesize that loss aversion was at play here in changing behaviors. The Medicare Readmission Reduction Incentive Program is based on a penalty for hospitals with high readmission rates (as opposed to a bonus for hospitals with low rates), and that could have led to implementation of new management programs years ahead of the penalty being “felt” by hospitals — hence the impact on performance.
Mental accounting is a behavioral economics principle whereby people tend to treat money differently depending on where it comes from. For example, someone may be more likely to splurge on an expensive dinner after receiving a tax refund, even though that person might never consider doing that with earnings received from a usual paycheck. Delivery system leaders could take this bias into consideration when designing financial incentives.
Advocate Healthcare has been mailing performance-based bonus checks to clinicians’ homes or handing them out in front of peers, which not only decouples the incentive from usual pay but also adds social value effects. The Massachusetts General Hospital has also realized the importance of decoupling incentives payments from the usual payments providers receive for doing their work.
So, where does this leave us? Health care professionals are first and foremost people; by taking into consideration that we are all subject to a complex web of interacting influences and a multitude of cognitive biases, we could do much better at “nudging” decisions and actions to maximize our common goals — better health at affordable costs. Behavioral economics will not be simple to apply — with it comes intricate issues of human agency, as well as ethical considerations as to how decisions are influenced and by whom. But transparency in the applications of these principles by policymakers—be they at the national, state, regional, or organizational level—certainly has great promise, and it is time to test and evaluate new tools and advance our use of cognitive and behavioral sciences in health care delivery. We’re done with simple carrots and sticks.