Strong versus Weak Incentives: The Role of Policy, Management, and Theory in a New Research Agenda

The papers in this Special Issue provide an excellent overview of the current state of the research literature around improving the value of the U.S. health care system and offer a range of sensible new directions for future research. The papers address both payment incentives, noting the strengths and weaknesses in theory and practice of fee‐for‐service, capitation, and salary payments; and informational incentives, including pay‐for‐performance (which combines payment and performance measurement), public reporting, peer review, and the role of patient‐reported information. They point out that the most appropriate payment or information approach is likely to depend on the nature and setting of care. They also point to the layers and multiplicities of incentives that often exist in our fragmented health care system. Insurers may, for example, pay a group practice a capitation rate for services, while the practice pays its provider employees using salaries, or by fee‐for‐service. Payers may monitor hospital outcomes, while hospitals monitor the productivity of their employees. Medicare may offer hospitals bundled payments that include postacute care, while the dominant area private insurer may pay using hospital‐only diagnosis‐related groups (DRGs). The U.S. health care system is nothing if not complicated.

As the papers show, we have learned a lot, and could plausibly learn a lot more. Yet after reading the papers, it is clear that the challenge ahead remains daunting. There is an enduring mismatch between the very limited array of instruments at our disposal—money and information—and the nearly infinite and endlessly evolving set of contexts that those instruments must address. Research studies tend to examine specific elements of these enormously complicated interactions, but drawing more out of the research will require developing a framework that can be used to integrate the results obtained from specific studies. Below I focus on two elements of that future framework: distinguishing between policy and management, and identifying the process of change.

Policy versus Management

The awesome diversity of contexts within the health care system—from the radiologist reading images to the ICU team providing care to a very ill patient to the primary care practitioner encouraging a smoker to quit—poses immense difficulties for policy makers. Public policy is, by design, a blunt and inflexible tool. Health care analysts grumble about the clumsiness of policy makers, prompting enthusiastic efforts at reinvention and reimagining. But the sluggishness and one‐size‐fits‐all nature of policy is a feature, not a bug. Nor is this feature simply a consequence of the inherent inefficiency of government bureaucracy. The policies of large private payers are likely to be nearly equally graceless and intractable. Policy is neither agile nor tailor‐made, even when the process works perfectly.

The government's policy process is required to be transparent and open. Regulations must be reviewed with respect to their implications for costs and quality, and competition in the provider market, and potential effects on vulnerable populations and rural providers, and implications for other programs and potential regulations. The process must solicit, review, and respond to public comment. Congress may demand that regulators respond to questions, in writing or in oral testimony. The health care provider sector, in turn, demands rules that are predictable and stable, so that it can make appropriate investment decisions. These varied processes generate democratic accountability and policy legitimacy—delay and rigidity are unavoidable byproducts.

Private payers can bypass much of the accountability apparatus, but they must still implement new payment rules across their own bureaucracies. That too requires briefings and rulebooks and computer coding changes. Private payers too need to be perceived as offering arrangements that are predictable and stable, so that providers will be willing to participate. As many of the authors note, in our fragmented health care system, payment and information reforms should be coordinated across payers for maximum effect. This coordination exercise generates even more delays and complexities.

These problems go well beyond payment. The information provided on a public—or even a payer—website must meet standards of fairness and accuracy well beyond those on Yelp. A rant from an unsatisfied customer means something quite different when it is sanctioned by CMS and when it appears on an unmonitored bulletin board. A payer‐established provider review organization must meet standards around conflict of interest, transparency, and accountability that will surely impede its flexibility and curb the extremes of its judgment. Its determinations are subject to formal review and sometimes even to review under judicial processes.

The nature of the policy‐making process favors bright line, simple rules that do not account for differences in context, and bright line, simple sanctions that are insensitive to individual circumstances. But that is not altogether a bad thing. In the health care sector, these policy features may actually be better than more agile alternatives. Despite our best efforts to collect information, health care providers know much more about their patients, their own organizational structures, and the incentives they face than payers ever will. Complex, situation‐specific rules offer myriad opportunities for nimble and better‐informed private actors to exploit arbitrage opportunities. The most recent example of this tension between context and rule has been around the Medicare facilities fee, an apparently sensible accommodation to the added costs of operating a hospital—but one that has led hospitals to purchase physician offices to generate additional billings. Medicare's effort to move to a site‐neutral payment methodology is an implicit acknowledgment of the risks of molding payment policy to context.1

How should the nature and the limitations of the policy‐making process (whether public or private) affect the development of a research framework? The framework needs to draw a line between the scopes of policy making and management. Policy makers can only use very limited combinations of very simple tools. For example, under the Medicare shared savings program, one of the most complex policy designs in use, Medicare pays based on actual versus anticipated costs and a limited set of quality mechanisms. Managers can use quite sophisticated mechanisms, involving frequent informal monitoring of physician behavior, selection of participating physicians, and the collection of impressionistic information about the quality of care, as well as quantifiable evidence, to respond to these incentives. Researchers might consider not only what the best payment or information policy is but also where that mechanism might best be situated—at the level of policy or at the level of management. For payers, generating value is likely to come through varying payment levels within a very straightforward payment system and monitoring coarse outcomes within an equally simple information collection, analysis, and dissemination system. Managers within organizations, operating within these payment and information structures, can use much more complex and rapidly varying internal tools to achieve results. Management‐oriented research can help identify not only what the best tools are likely to be, but the nature of the organizations where they are likely to be deployed effectively. Such management‐oriented research is critically important: policy makers are much more likely to achieve their policy goals if managers have a good idea of how best to respond to the incentives presented.

This bifurcation of roles highlights the importance of policy‐oriented research on the measurement of appropriately broad and meaningful outcomes. Policy makers set incentives and goals to which managers respond. If those measures are not well designed, management effort will be focused in the wrong directions. Thus, a key imperative of policy research is to properly define these measures. The challenges of outcome measurement have led many programs to rely heavily on process measures, but these can stymie the very types of innovation that are best left to management practice. Research on patient‐generated, functionally focused outcomes, and on methods for risk adjustment of such measures, is particularly critical in generating appropriate incentives for quality improvement.

Process of Change

Drawing a brighter line between policy and management practice highlights a second gap in our research—identifying a theory of change. Most past efforts to ensure health care quality, such as licensure and the use of peer review organizations, have focused on eliminating the bottom of the quality distribution, either by preventing unqualified providers from entering practice or by improving the quality of less effective providers once they are in practice. Much of the improvement in the general quality of practice over time has come through shifting the entire distribution of practice, usually in response to new information about what does or does not work. Efforts to improve value through dissemination of information usually encourage patients (or providers) to aim for the very best—or highest value—providers, in the hope that competition will lead others to follow this desirable, high‐quality path.

All three of these mechanisms have the potential to improve value—but they are strikingly different from the process that generates greater value outside the health care sector. That process is much more ruthless and unstructured. Better value providers, say Walmart (which creates value by offering existing products at low prices) or Starbucks (which creates value by offering a better product at a premium price), earn profits from their operations in an initial market and expand into another. Once there, they drive their less value‐generating competitors out of business. They replace local, idiosyncratic practices with standardized, effective processes honed at central headquarters. The former owners of the local five‐and‐dime may take a job as cashiers at the superstore, or they may become unemployed or move out of town. Quality generally improves not by training the worst, moving the entire distribution, or encouraging competition toward excellence, but by replacing inefficient local producers with subsidiaries or franchises that hew to an unvarying corporate line.

For many reasons, ranging from the not‐for‐profit nature of most hospitals to the strong local character of care to the difficulties of recruiting replacement practitioners, this process would be unlikely to occur in health care. The political power of health care providers makes it even more dubious. Our theories of change in health care instead assume that, with proper training and incentives, everyone can be a winner. That is consoling, but it would be nice to see some evidence that it's also plausible. In addition to studies that examine the average effects of incentive systems on outcomes, it is also essential that we study how change occurs, both within an organization and, critically, across organizations. Research can help us understand more about the process through which value improves across the health care system. Studies of payment or information reforms should look not only at average effects and at case studies but also explore the mechanisms of system‐wide change, especially in ways that can be manipulated by policy. Do desirable responses to payment incentives happen faster in more competitive markets? Are responses faster among nonprofit rather than for‐profit hospitals? Does change spread across providers within health systems or across providers in neighboring markets? Do providers who cannot respond to incentives exit markets or merge with their successful counterparts?

A framework for unifying research on improving value needs to explicitly incorporate a theory of change. As we sharpen the distinction between policy and management, policy makers need a clearer understanding of how organizational change affects cost and quality outcomes. With that information, policy makers could monitor change at the organizational level—which providers are expanding, which are consolidating, which are contracting—as an early warning measure, an output in logic model parlance, of how value is changing in the system.

Next Steps

As a health policy researcher, the papers in this volume give me some hope. We have learned a lot—but there is lots more research to be done. The nature of research in this area has been centripetal, driving out toward very specific examples and situations. Progress will be accelerated if we can develop a framework to unify this literature—to point out patterns that can be generalized in future policy changes. That framework needs to be careful in specifying the roles of policy and management in achieving value, and in defining the process of change that will get us there.