Can Alternative Payment Models And Value-Based Insurance Design Alter The Course Of Diabetes In The United States?

Abstract

Diabetes process and outcome measures are common quality measures in payment reform models, including Alternative Payment Models (APMs) and value-based insurance design (VBID). In this commentary we review evidence from selected research to examine whether these payment models can improve the value of diabetes care. We found that higher-risk APMs yielded greater improvements in diabetes process measures than lower-risk APMs, and that VBID models appeared to improve medication adherence but not other quality measures. We argue that these models are promising first steps in redesigning the payment system to improve diabetes care. However, greater coordination and alignment across models is needed to enhance their impact on providers’ behavior, diabetes care processes, and patient health outcomes.

Successful diabetes prevention and management require active participation from patients and coordination across health care providers. In an ideal world, care plans would be tailored to patients’ individual needs, with providers routinely screening for and addressing the social determinants of health that affect diabetes outcomes. Patients would have the time, training, and support to fully engage in self-care through evidence-based lifestyle interventions. Diabetes medications would be affordable, with education readily available for medications requiring special instructions. Importantly, patients would have continuity with a provider team between visits and over time.

The real world of diabetes care is far from this ideal.¹ Many patients are unaware of their diabetes status. Health care systems in the US are still developing screening protocols for social determinants of health and population health programs that support patients outside of the traditional clinical encounter. The health insurance system and health care delivery systems are highly fragmented and lack cohesion. Finally, traditional fee-for-service payment, which is still the predominant form of payment, rewards providers for the volume of service provided, not for the quality of care provided.

During the past two decades researchers and policy makers have looked to payment reform as a vehicle for health care system improvement. To incentivize care quality rather than volume, policy makers have promoted Alternative Payment Models (APMs).² APMs link physician and hospital reimbursement to the value of the care provided, as evaluated on a predetermined set of quality and utilization measurements. APMs have the potential to improve health outcomes for patients with diabetes by enhancing care co-ordination, focusing attention on outcomes, and incentivizing investments in social determinants of health.

Value-based insurance design (VBID) is another development in payment reform with the potential to improve diabetes outcomes. VBID is a benefit design strategy that reduces patients’ out-of-pocket spending requirements for high-value care and sometimes increases out-of-pocket spending for low-value care. As many VBID programs focus on reducing copayments for prescription medications, clinical management of diabetes is a natural application for VBID.³

In this commentary we examine evidence from selected research to address the following question: Can APMs and VBID improve the quality and value of care provided to the thirty-seven million people with diabetes in the US?

Alternative Payment Models

We examined thirteen studies evaluating APM models of varying risk to providers, as categorized by the Health Care Payment Learning and Action Network (HCP-LAN)⁴ (see online appendix exhibits 1–3).⁵ This framework categorizes payment models based on the level of financial risk assumed by the provider, with category 1 reflecting no risk (for example, fee-for-service Medicare) and category 4 reflecting full risk (for example, full capitated payment). Most studies evaluated process measures, such as glycated hemoglobin (HbA1c) testing, although several also reported on outcome measures such as blood pressure control.^6–8

FEE-FOR-SERVICE WITH A LINK TO QUALITY

Pay-for-performance models (HCP-LAN category 2), in which providers receive bonus payments for meeting quality and utilization targets, first began in the commercial sector in an effort to control health care costs. This payment approach then spread to the public sector in the early 2000s, hastened by an Institute of Medicine report calling for better alignment of payment incentives and quality improvement. We identified five prominent pay-for-performance models (appendix exhibit 1)⁵ across public and private payers.^7,9–12 These studies reported both positive and negative associations between bonus payments and diabetes process measures. One study that examined Blue Cross Blue Shield Michigan’s Physician Group Incentive Program reported an increase in HbA1c testing.¹¹ In contrast, two other studies, one examining a program implemented by a national managed care organization⁹ and another examining a multi-payer patient-centered medical home intervention,¹² both found decreases in HbA1c testing.

MODELS BUILT ON FEE-FOR-SERVICE

Seeking to further innovate in payment reform, the Centers for Medicare and Medicaid Services began to move beyond simple bonus payments and toward shared-savings and shared-risk models (HCP-LAN category 3). We evaluated five such models^13–17 (appendix exhibit 2).⁵

In the Medicare Physician Group Practice demonstration, the first shared-savings initiative under Medicare (2005–10), demonstration sites showed significant improvements in the frequency of HbA1c testing (3.55 percent) and eye exams (1.52 percent) compared with controls.¹³ In 2012 Medicare launched the Pioneer accountable care organization (ACO) program with thirty-two organizations. During the first year, participating organizations showed modest (<1 percent) improvements in glucose and cholesterol testing, as well as eye exams.¹⁵ During the same period, Medicare also launched the much larger Medicare Shared Savings Program ACOs. These 220 ACOs showed a significant but modest (1.1 percent) improvement in frequency of low-density lipoprotein (LDL) cholesterol testing compared with controls in the first year of the demonstration.¹⁶ The Comprehensive Primary Care Plus (CPC+) program was designed to improve primary care in ambulatory practices. Under track 1, which retained traditional fee-for-service payment, practices received care management fees as well as a performance-based incentive payment based in part on shared savings. The impact of CPC+ track 1 on diabetes outcomes was modest, with less than 1 percent improvement in the evaluated measures.¹⁷ Finally, the northeast Pennsylvania Chronic Care Initiative experiment identified the largest improvements in care among the models with shared savings payment designs.¹⁴ The initiative was a regional program that combined medical home practice support with shared-savings incentives. Practices saw significantly higher performance on all diabetes measures relative to controls (frequency of HbA1c testing increased by 8.3 percent, LDL testing by 8.5 percent, nephropathy monitoring by 15.5 percent, and eye exams by 12.0 percent).¹⁴

POPULATION-BASED PAYMENT MODELS

We identified three population-based payment models (HCP-LAN category 4), which represent the greatest departure from traditional fee-for-service (appendix exhibit 3).⁵ The first was the Blue Cross Blue Shield Massachusetts Alternative Quality Contract, a landmark global budget experiment in which providers received an annual per patient global budget covering the entire continuum of care for a defined population of enrollees.⁸ This initiative reported significant improvements in all process-related quality metrics, the largest of which was a 7.2 percent increase in eye examinations. The second was track 2 of the CPC+ initiative, in which practices received prospective lump sum quarterly payments to support selected primary care services with reduced fee-for-service billing.¹⁷ This track reported significant but modest (<1 percent) improvements in the frequency of HbA1c testing and eye exams. The third was Maryland’s Total Patient Revenue (TPR) program, in which hospitals received a flat annual amount for all hospital-based services.¹⁸ This program demonstrated reductions in emergency department admissions related to diabetes.

Value-Based Insurance Design

Since the mid-2000s, the collective experience with VBID has grown from a small number of early projects at private firms (Pitney Bowes) and public institutions (the University of Michigan) to a sufficient number of experiments to warrant systematic reviews.¹⁹ We discuss the findings of eleven recent evaluations of VBID programs published during the last decade,^20–30 focusing on the medication management of diabetes (appendix exhibit 4).⁵

Most of these studies evaluated VBID programs implemented by commercial health plans or employers. Study designs were typically pre-post comparisons, often with a concurrent control group of patients with diabetes who were not exposed to VBID. Two studies used an interrupted time series design with matched controls.^29,30 Although the specific benefit design varied across interventions, patients typically had copayments of $0–$10 for generic medications. The reduced out-of-pocket expenses for brand-name medications were more varied, ranging from $0–$30 copayments for preferred brands and 30–50 percent coinsurance for non-preferred brands (appendix exhibit 4).⁵

Among studies that included a control group, seven identified a small increase in adherence associated with VBID, although in a few cases, the difference between groups was primarily due to adherence declines over time in the control group. Patients with lower baseline adherence tended to have a greater increase in adherence. One study found a greater increase in adherence for low-income patients compared with high-income patients.²⁹ Another found that a zero-dollar “preventive drug list” in a high-deductible health plan was associated with a 16 percent decrease in emergency department visits for acute diabetes complications compared with the control group.³⁰

Several other notable studies provide additional insights into VBID and diabetes care. One used data from the Medicare Current Beneficiary Survey to simulate total costs with copayment caps of $1–$25 for statin medications. Smaller copayment caps were associated with reduced total medical spending, but only for people at higher cardiovascular risk.²² Another described a recent natural experiment among the Wisconsin adult Medicaid population²⁸ that demonstrated that small reductions in copayments were associated with increased medication adherence.

Finally, a comprehensive evaluation of thirty-two VBID plans introduced by a large pharmacy benefit manager examined medication adherence data from more than 75,000 patients with diabetes.²³ Several different design features were associated with improved adherence, including targeting high-risk patients, offering a concurrent wellness program, and giving patients incentives to use mail-order pharmacies to receive lower copayments.

Discussion

The past decade has been an intense period of experimentation with APM and VBID payment models. Identifying the most effective models, and understanding what features make them that way, is important for the future of diabetes care.

APMs vary in their design and impact on diabetes care. Models that place greater financial risk on providers, such as global payment and shared savings, tend to demonstrate greater improvements in diabetes quality metrics compared with lower-risk pay-for-performance models. This pattern is consistent with expectations that higher-risk models incentivize providers to make larger investments in program components. However, nearly all of the payment models we looked at assessed process measures alone, and improvements in these processes might not translate into improvements in outcomes.

It is not easy for providers to make the changes necessary to implement new payment models. New workflows must be designed and implemented, electronic records must be upgraded to provide timely performance reports, and practitioners must use the measures to change how they provide care. Some programmatic investments, such as developing a community health workforce, require up-front costs, whereas downstream savings might not materialize during the short length of APM contracts. Moreover, provider organizations are frequently juggling multiple APM contracts with differently calibrated quality measures while simultaneously maintaining other legacy fee-for-service payments. Given the considerable investment associated with implementing APMs, it should be no surprise that some practices and providers, particularly those that are part of integrated health systems, are more likely to participate in APMs than others.

Unlike APMs, VBID models put the onus of incentivizing quality on payers rather than providers. Some VBID models have reduced patients’ out-of-pocket spending and modestly improved medication adherence, but there is little evidence of improvement in outcome measures, such as HbA1c control. Recent work has shown a link between zero-dollar copayments and reduction in emergency department visits for acute diabetes complications; additional well-designed studies are needed to confirm this finding. Overall, the lack of VBID impact on clinical outcomes may be due to the fact that medication adherence may be necessary for, but not sufficient to achieve, optimal HbA1c control.

Overall, the fragmented US health care system, with its myriad payers and payment models, is structurally at odds with the need for care continuity for chronic conditions such as diabetes. The growth in payment experiments adds to this sense of misalignment: Many provider organizations manage a diverse portfolio of APMs across payers and categories of risk, posing challenges to care coordination and continuity. The individual experiments are frequently focused on specific subpopulations, which discourages development of comprehensive diabetes programming. Short of a larger overhaul of the US health insurance system, there are several areas in which stakeholders and policy makers can encourage greater alignment.

First, policy makers should consider increasing APM and VBID experimentation for the Medicaid population. This population includes a racially diverse group of low-income adults who are at risk for long-term complications of diabetes as they age. Current fragmentation within payment reform efforts across payers may exacerbate well-documented racial and socioeconomic disparities in diabetes outcomes.³¹ Experimentation in payment reform in this population is necessary to ensure that innovative strategies are promoting health equity. Population health programs offered by safety-net providers, focusing on social determinants of health,³² may be particularly beneficial for the Medicaid population.

Second, policy makers should do more to standardize diabetes quality measures across APMs. For glucose control, payment models differ in whether they emphasize reducing poor control (HbA1c >9.0 percent) versus increasing optimal control (HbA1c <8.0 percent). Standardizing diabetes quality metrics would encourage provider organizations to pursue more comprehensive programming across payer populations and reduce administrative burden from multiple measurement approaches. Further, expanding quality measures to include both processes and outcomes over time will help quantify the true impact of APMs and VBID on diabetes care.

Finally, given that the effects of APMs and VBID appear to be moderate at best, policy makers should consider integrating APMs and VBID under a single payment model, combining both provider and patient incentives with the hope of achieving larger effects on diabetes care outcomes. Broader alignment of payment reform efforts, combined with increased investments in the primary care safety net, is needed to maximize their collective impact.

In October 2021 the Center for Medicare and Medicaid Innovation launched a strategy refresh, outlining its goal to advance total-cost-of-care models while addressing health inequities. Now is a good time to take stock of how the experimental payment reform models of the past decade have fared. The evidence for early APMs and VBID is promising, but incomplete. With ongoing iteration, APMs and VBID may allow the US to move closer to the ideal world of diabetes care—one in which patients are afforded individually tailored care plans, training, and support for self-management; access to resources that address social determinants; and continuity across providers and time.