Bundled Payment Systems: Can They be More Successful this Time

The success of health reform, the stability (and equity) of the health care system, and even the fiscal health of the nation all require the gap between health care spending growth and income growth to shrink. The recently passed Patient Protection and Affordable Care Act incorporated many provisions intended to accomplish this goal. Initiatives that bundle payments to health care providers are among the most touted of these strategies.

Several types of bundled payment approaches (not mutually exclusive) are under consideration. One approach, episode payment systems, bundles care across services and providers within an episode of care. Diagnostic-related groups (DRGs) represent an early, limited version of such systems, bundling services within an inpatient admission. Payment systems under consideration now go further, for example, bundling hospital, physician, and postacute care.

A more comprehensive approach, global payment systems, provides a fixed payment per person for a defined period of time, known in the past as capitation. Modern versions incorporate quality measures to guard against stinting on care. Gain sharing models, such as those commonly associated with accountable care organizations, are a version of global payments, but typically reduce or eliminate downside risk and allow providers to share in only a portion of the savings (McClellan et al. 2010).

While much of the attention surrounding global payments focuses on their desirable incentive effects, from a payer's perspective a primary appeal of global payment is that it allows for more complete control of spending. In a fee-for-service (FFS) environment, spending may rise even if fees are reduced due to increased utilization (provided that services remain profitable at the margin). Global payments control both price and quantity if there is sufficient downside risk. Episode-based payment systems provide less comprehensive spending control than global payments because of the potential for increases in the number of episodes (and difficulty in assigning all care into an episode). In both episode-based and global payment models some risk is transferred to providers, depending on the structure of the program.

Bundled payment systems are not new. Capitated contracts proliferated in many parts of the country during the managed care era over a decade ago. Yet, for a variety of reasons the capitated payment systems did not transform the health care system and the popularity of capitation waned in subsequent years. For example, the percent of physicians refusing all capitation rose from 37.4 percent in 1996/1997 to 60.7 percent in 2008 (Centers for Studying Health Systems Change 1997, 2008). Others have provided a postmortem for managed care of the 1990s, but as we design new bundled payment systems some attention to past lessons is useful (Mechanic 2001, 2004).

Existing, often dated, research provides some insight about the potential success of bundled payment systems and key factors that will be related to success. This literature focuses on global payment, as opposed to episode payment, because episode-based bundling was not prevalent in past eras beyond limited systems such as DRGs.

One strand of relevant literature examines the impact of managed care plans, generally HMOs (health maintenance organizations), on health care spending and utilization. Like all at-risk plans, HMOs receive a capitated payment. When the plans are tightly integrated with providers, the impact may be similar to that which would arise when the physician group is capitated directly. Perhaps the best evidence of the impact of integrated, at-risk health plans on use comes from the RAND Health Insurance Experiment, which randomized beneficiaries and found that an HMO that was tightly integrated with providers reduced use by 28 percent (Manning et al. 1987). More recent work summarized by Glied (2000) confirms this result, finding that HMOs generally reduce cost, though there is a wide range of findings in the literature. These reductions appear to have moderated since the late 1990s, in part due to pressure by patients demanding unconstrained choice of physicians (Glied 2003).

A smaller literature examines the more relevant question of how providers respond when they—as opposed to the health plan—are capitated. This literature indicates that capitated physicians utilize fewer health care services. For example, Kralewski et al. (2010) found that practices under capitation used significantly fewer services than FFSs or salaried practices. Other research suggests that capitation is associated with a reduction in hospitalizations (Hillman et al. 1989; Stearns et al. 1992;). Shrank et al. (2005) find that physicians operating under capitation were half as likely to perform cataract extraction, and Shafrin (2009) finds that outpatient surgery rates among specialists paid FFS were 78 percent greater than under capitation. Chernew et al. (2000) find spending on drugs was lower in capitated settings in the year after capitation was implemented.

The success of capitation requires not only a reduction in the level of spending, but also a reduction in the rate of spending growth. The factors related to spending growth (e.g., development and diffusion of new medical technology) may differ from those leading to high spending (e.g., high prices, over use of health care services) (Chernew et al. 2010).

The impact of capitation on spending growth depends on the rate at which the capitation rate rises as well as the political sustainability of the system. A literature review from the late 1990s suggests that managed care plans have similar rates of growth as nonmanaged care plans, but that markets with more managed care have somewhat slower spending growth. However, the spending growth in high-managed care markets was still significantly above income growth (Chernew et al. 1998). While we can afford some gap between health spending growth and income growth for many years, the size of that gap must drop considerably and it seems managed care of the 1990s did not have a large enough impact. Chernew et al. (2004) attribute the relatively modest impact of managed care on spending growth, in part, on the inability of health plans to control physician practice patterns when services are new, profitable, and evidence is weak. Provider norms that transcend health plan structures seem to dominate.

A system with widespread use of bundled payments may be more successful than studies of individual plans suggest. Certainly one would expect spillovers across the market if bundled payment systems become widespread. Yet if bundled payment is to successfully constrain spending growth, updates of the bundled payment rates must be restrained and the whole system must be politically sustainable. Chernew and colleagues report that during a period of rapid increase in prescription drug spending, the capitation rates in a capitated prescription drug program increased rapidly, thus the system failed to slow spending growth. Moreover, the whole initiative collapsed after several years. This is consistent with the demise of many of capitation programs adopted in the 1990s.

Ultimately the success of bundled payment systems at slowing the rate of spending growth and their sustainability will depend on the extent to which they slow the rate of introduction and diffusion of new services (which does not necessarily imply foregoing quality improvements). If the past is a reasonable guide, this will be challenging unless there is a fundamental change in provider culture, which perhaps widespread global payments can generate. Otherwise, pressure from medical professions and industry will continue to drive practice styles toward more expensive care patterns. It seems difficult for individual provider groups to deviate too much from the standards of care regardless of their incentives.

Several features of global program systems may facilitate their ability to affect care. First, the size of the provider group is important. The reward any given provider reaps if they practice conservatively is diluted in large groups. In smaller groups the incentives of global capitation are closer to the point of care because the share of any savings accruing to any one provider is generally greater. However, if the effects of capitation stem from group-level initiatives, larger groups may be able to devote more resources to efficiency promoting initiatives (Robinson et al. 2009). Thus, the optimal size of a group receiving global payments remains an empirical question.

Second, the extent to which groups pass incentives on to their physicians may influence the impact of capitation. Group size may be an important determinant of how they do that. Groups of all sizes must devise ways to distribute funds and encourage productivity. Many capitated groups may in fact use relative value units (RVUs) or other FFS-type bonuses to encourage productivity and as a mechanism to allocate the global payment across participating providers. This may particularly salient in large groups where other productivity monitoring strategies are more difficult.

Services or RVU-based payment within a group may create incentives similar to FFS that groups must address. Existing research suggests that group payment incentives have greater overall utilization effects than physician incentives (Conrad et al. 1998). This is consistent with the view that the effects of capitation arise because capitation encourages group level initiatives to manage spending and a group culture that influences utilization as opposed to because physicians make direct financial calculations on a case-by-case basis. However, there is substantial evidence that group incentives and individual incentives interact significantly, making it difficult to disentangle the two (Hillman et al. 1992; Robinson et al. 2009;). Thus understanding the impact of group- and physician-level incentives will be an important topic for future research.

A third factor that will affect the effectiveness of a global payment model is nature of the transfer of risk. There are many strategies that may affect risk transfer. Risk adjustment can control for some of the risk, though risk adjustment is imperfect and some random variation will remain. The random variation will fall as the size of the group rises, but systematic shortcomings of risk adjustment will still be a concern. For example, sociodemographic factors may be important determinants of spending and are poorly captured in risk adjustment systems. Moreover, issues such as coding practices must be examined more closely as evidence emerges that the coding may reflect practice styles with more intensive practice uncovering more risk factors (Song et al. 2010). Reinsurance may also reduce risk and in some payment models provider groups take only limited (or no) downside risk or may be responsible for only a subset of services. Partial capitation, in which providers receive some of their payment in the form of a bundled payment and some in the form of FFS payment (at a lower rate) could also mitigate risk and improve incentives, though this strategy has received less attention (Ellis and McGuire 1986; Newhouse 1996;). All of these strategies to mitigate risk could be important. Yet while reducing downside risk borne by providers may encourage participation in bundled payment systems, it may also reduce the incentive for providers to practice conservatively.

A fourth factor, which will be a central determinant of the impact of global payment on spending growth from the payers' perspective, is the rules for updating the global payment level. If rates are reset each year based on the trends in each group, any savings in a given year will lead to lower updates (or even decreases) in the payment rates in subsequent years. This may discourage groups from investing in efficiencies because the savings will be largely captured by payers. This effect is mitigated if the update is based on the collective trend in all groups under global payment. If there are many groups receiving bundled payment, use in any one group will have a small impact on the update and incentives for individual groups to practice conservatively will be stronger, depending on how the collective incentives facing all groups influence individual physician groups.

If the updates are based on spending trends outside of the capitated groups, the ability of capitation to slow spending growth is dampened because presumably spending growth would be faster in the noncapitated sector (and such a system requires a substantial population outside of the bundled payment scheme). A final approach would set updates based on normative assessment of what payment update is needed or sustainable. This approach may be successful at slowing spending but relies on our ability to assess appropriate normative updates, which may be easier in an episode as opposed to global bundled system. In all cases, our ability to politically sustain lower updates politically is crucial and eventually payers as well as providers will need to capture savings from more efficient practice. Longer contracting periods or targets that are not based on the past experience of individual or small numbers of capitated groups will permit providers to keep much of the efficiency savings generated by investments in improving care processes and may be more effective at promoting efficiency.

A final feature of global payment systems that does not receive a lot of attention is who the residual claimant is. In some models, primary care physicians may disproportionately reap savings and in others the hospitals may capture gains. In general, each organization will develop the internal allocation systems to determine which providers or groups of providers benefit from efficiency gains. Those decisions, shaped by the marketplace and the details of the design of the global payments will be central to determining how behaviors change as payment systems change.

Moving forward, bundled payments will likely be an important feature of the health care system. Momentum from large public payers and recognition of the well-known inefficiencies stemming from traditional FFS payment (and scheduled reductions in Medicare FFS rates) will likely lead to greater diffusion of these systems than in the past. Yet existing literature indicates that we should not underestimate the design challenges posed by global (or episode-based) payment. Choices and challenges will inevitably arise. The challenge faced by the field of Health Services Research is to provide theory and evidence to guide the choices and surmount the challenges.