Moving Toward Paying for Outcomes in Medicaid.

Medicaid can improve beneficiary health and help sustain its own future by embracing payment for outcomes. Good precedents exist from states such as Florida, Maryland, Minnesota, New York, Ohio, Pennsylvania, and Texas. Medicaid outcome measures include preventable admissions, readmissions, emergency department visits, and inpatient complications; early elective deliveries; infant and child mortality; patient-reported outcomes, satisfaction, and confidence; and reduction in low-value care. Criteria to prioritize initiatives include potential savings, availability of established models, impact on health status, and Medicaid's ability to effect change. We offer 5 principles for success, emphasizing clinically credible initiatives that generate actionable information for clinicians.

MEDICAID—the nation's largest health insurance program, covering more than 70 million Americans—can improve beneficiary health and help sustain its own future by embracing payment for outcomes (P4O). Good precedents exist from states such as Florida, Maryland, Minnesota, New York, Ohio, Pennsylvania, and Texas. State efforts will be propelled by better data when the Transformed Medicaid Statistical Information System goes live in 2018 (Centers for Medicare & Medicaid Services, [CMS] n.d.b).

P4O will help states meet federal regulatory requirements for closer monitoring of managed care quality and for tighter ties between payment and utilization, quality, and outcomes (CMS, 2016). At the same time, the very worth of Medicaid coverage continues to be debated (Sommers et al., 2017). In short, Medicaid must respond to political and budget pressure to save money and demonstrate its value (Brooks, 2017).

We list key P4O questions facing Medicaid policymakers and suggest answers based on a research review and our own decades of work designing and implementing methods to measure outcomes and improve payment incentives.

WHY MEDICAID P4O?

Why consider Medicaid separately?

That Medicaid is not Medicare often bears repeating. Medicaid policy matters far more for maternal and child health, mental health, developmental disabilities, and for specific conditions such as HIV/AIDS, cystic fibrosis, and sickle cell anemia (Kaiser Commission on Medicaid and the Uninsured, 2013). Medicaid also dominates policy and payment for people who need institutional or community-based care for the remainder of their lives. Managed care covers more than 60% of Medicaid beneficiaries, or approximately twice the proportion of Medicare.

While the CMS has led the national push toward paying for quality, it focuses on Medicare. Some of its Medicaid initiatives, such as Provider Preventable Conditions, have simply extended Medicare approaches, with negligible adjustment for Medicaid. Although the CMS does compile Medicaid-specific quality results by state, only a few of the 57 adult and child measures address outcomes (CMS, n.d.a). And while Medicare's success in reducing readmissions for heart failure and similar conditions has been widely noted (Wasfy et al., 2017; Zuckerman et al., 2016), there has been little recognition that psychiatric patients are the major readmission challenge facing Medicaid (Trudnak et al., 2014).

In the eyes of some evaluators, the impact to date of Medicare quality initiatives has been “mixed,” “far from convincing,” and “discouraging” (Jha, 2017; Ryan et al., 2017). In addition to designing quality programs more suited to their populations, Medicaid agencies can learn from Medicare's experience, especially on questions of process versus outcomes, focus, “actionable” information for clinicians, and the structure of incentives.

Why focus on outcomes?

Of the famous 3 dimensions of quality—structure, process, and outcomes—outcomes have attracted the least fealty. But patients care about outcomes. We therefore start with a predisposition toward outcomes, which include mortality, readmissions, functional status, patient confidence, and healthy childbirth, among others.

Focusing on outcomes applies a basic management principle: set the goal and get out of the way. Unlike measurement based on rigid adherence to specific processes, a model based on outcomes encourages innovation, rewards collaboration, and does not intrude upon the patient-physician relationship.

Providers often criticize the sheer number of quality measures (Schuster et al., 2017). Aside from the administrative burden, the plethora results in diluted efforts. The Medicare Hospital Value Purchasing Program, for example, awards 1 to 5 stars to hospitals based on 62 very different measures across 7 weighted groups, resulting in a single composite performance measure. Commentators suggest that a more focused program would be more effective (Jha, 2017; Ryan et al., 2017). A more focused approach would also avoid the difficulties of creating a composite measure (Shwartz et al., 2015).

Emphasizing outcomes also reinforces other goals that have been identified for the US health care system: weeding out ineffective care (Morgan et al., 2018), recognizing the value of primary care in achieving good outcomes (Koller & Khullar, 2017), quicker adoption of evidence-based medicine (Rennie & Chalmers, 2009), more appropriate use of expensive pharmaceuticals (Fuller & Goldfield, 2016), and greater integration among providers (Bodenheimer, 2008).

Can P4O reduce cost?

Yes. Many structural and process measures, despite their benefits, often translate into doing more—more inspections, more tests, and more treatments. Better outcomes, on the other hand, can mean fewer low-birth-weight babies, hospital admissions, readmissions, emergency department (ED) visits, infections, etc. A few examples:

In Maryland, inpatients experienced 36 466 fewer complications per year—a 68% decrease—between 2010 and 2015 (Maryland Health Services Cost Review Commission, 2016). The list, more comprehensive than the Medicare Hospital Acquired Condition list, included potentially preventable complications such as renal failure, diabetic ketoacidosis, septicemia, and shock (Patel et al., 2015).

In Minnesota, an award-winning program led by hospitals resulted in more than 7000 fewer readmissions over 2 years (McCoy et al., 2014).

In Texas, early elective delivery rates fell by as much as 14% after a Medicaid payment policy change (Dahlen et al., 2017).

Also in Texas, Medicaid managed care beneficiaries in 2015 experienced 9600 fewer potentially preventable admissions than expected, for conditions such as pneumonia and asthma (Millwee et al., 2017).

For Medicaid, savings from better outcomes are far preferable to tighter eligibility rules, narrowed benefits, and reduced payment rates (McConnell & Chernew, 2017).

OUTCOMES APPLICABLE TO MEDICAID

What outcome measures are especially important for Medicaid?

The list should reflect Medicaid's share of the many “markets” for health care, with particular focus on neonatology, pediatrics, obstetrics, mental health, developmental disabilities, and long-term services and supports (Quinn & Kitchener, 2007). Medicaid P4O programs in place today, applied to providers or health plans, include preventable admissions, readmissions, ED visits, and inpatient complications; early elective deliveries; mortality; and patient satisfaction. Other patient-centered measures, such as avoiding low-value care (Kerr et al., 2017), patient-reported outcomes (Lavallee et al., 2016), and patient confidence (Wasson & Coleman, 2014), hold promise. We know, for example, that confident patients experience better outcomes such as fewer avoidable ED visits (Hibbard et al., 2013). New York Medicaid, among other states, monitors patient confidence (New York Department of Health, 2014).

In maternal and child health, the growth of Medicaid managed care offers opportunity. Both infant and maternal mortality have relatively high prevalence in the United States, disproportionately affect populations served by Medicaid, and require approaches more holistic than the traditional medical model (Lu & Johnson, 2014; Molina & Pace, 2017). For children, a parsimonious set of quality and outcome measures is under development (Gardner & Kelleher, 2017).

In long-term services and supports, where Medicaid has particular responsibility and influence, states such as Arizona, Minnesota, and Ohio have P4O programs for nursing facilities that target avoidable hospital admissions, readmissions, and ED visits as well as pressure ulcers, infections, and use of physical restraints (Libersky et al., 2017). Outcome measures in home and community-based settings, which serve 3 million Medicaid beneficiaries a year, are notable for their scarcity (National Quality Forum, 2016). In both institutional and community settings, we need to better understand how P4O strategies can incorporate patient and family engagement and confidence (Forum on Aging, Disability, and Independence, 2016).

How should a state prioritize initiatives?

A limited number of P4O initiatives—10 or fewer—helps maintain focus. Criteria would include initiatives (a) likely to save money (b), that can follow established models, (c) have substantial impact on health status, and (d) where Medicaid has enough presence, on its own or in partnership with others, to exert influence (Averill et al., 2011).

HOW CAN MEDICAID PROGRAMS MOVE FORWARD?

What are the keys to success?

We suggest 5 principles: availability of clinically meaningful and transparent results, focus on population rates rather than individual incidents, careful risk adjustment, appropriate incentive structure, and collaborative phased implementation. We consider these in turn.

How can payers help providers and plans effect change?

“The ultimate objective of any payment reform is to motivate behavioral change that leads to better quality and lower costs” (Millwee et al., 2013). The importance of this principle is often overlooked. If clinicians and managers perceive a payment reform as clinically unsound, opaque, or irrelevant to daily practice, they will ignore or undermine it. The opposite holds when the reform is clinically sensible, transparent, and delivers “actionable” information for providers or plans. For this reason, we prefer categorical approaches over regression-based approaches (Fuller et al., 2016). When Texas implemented readmission measurement, for example, each hospital could download a spreadsheet showing which specific readmissions were considered potentially preventable.

Should states focus on individual incidents or population rates?

The traditional quality paradigm centered on individual incidents of bad quality (Quinn et al., 2016). An example is the initial campaign against “hospital-acquired conditions,” which affected a mere 0.02% of Medicare inpatient payments (Fuller et al., 2009). The alternative approach, now reflected in most Medicare quality programs and in our own work with states, recognizes that less-than-excellent quality often stems from “good people in bad systems” (Clifton, 2009). Instead of looking just at “never events” such as blood transfusion errors, the focus widens to hospital complications and other potentially preventable problems that cannot be completely eliminated. The paradigm shifts from “This should never happen” to “This has happened too often” (Fuller et al., 2011).

The typical population-based approach involves calculating performance for each provider or plan, and then comparing it to a population-wide benchmark on a risk-adjusted basis. Benchmark options include the population mean or a “best practice” rate achieved by top performers (Averill et al., 2011).

How important is risk adjustment?

In a word, essential. Inadequate risk adjustment undermined Medicare's 1980s publication of mortality rates by hospital, a debacle that should remind quality advocates to step carefully (General Accounting Office, 2008). Because of the concentration of health spending—just 5% of people account for 50% of spending (Cohen & Uberoi, 2013)—providers and plans are unfairly penalized and access to care jeopardized when risk adjustment algorithms fail to recognize that some patients start off much sicker than others. Indeed, much of medicine's historical reluctance to compare outcomes reflects the challenges of risk adjustment. But the science has advanced significantly, thanks to better data, better software, and deeper understanding (Iezzoni, 2012).

Even after clinical risk adjustment, poor outcomes are often associated with low socioeconomic status (SES) (Joynt et al., 2017; National Quality Forum, 2014). For example, risk-adjusted readmission rates in Rhode Island in 2010 were about 23% worse for Medicaid than for Medicare or commercial insurers (Quinn & Davies, 2014). The question is what to do about it. We suggest that, where possible, the SES impact should be calculated and analyzed separately. This approach enables separate consideration of policy actions. For example, SES probably affects premature birth rates more than it does hospital perinatal complication rates. Over time, SES adjustment often can be phased out, thus addressing concerns about payers rewarding poor quality. In the Medicare readmission program, for example, improvement has been most marked for the lowest-performing hospitals (Wasfy et al., 2017). We also note that socioeconomic differences matter less within a Medicaid-only population than when P4O initiatives cover multiple populations.

How should financial incentives be set?

In our view, the failure of financial incentives to effect change often reflects the failure of incentives to be substantive, clear, and credible (Jha, 2017). The exemplar of payment reform changing health care delivery continues to be Medicare's 1983 implementation of diagnosis related groups, which met all 3 criteria and has been widely emulated (Quinn, 2014). In P4O, payers also need to pay close attention to the esoterica of incentives: benchmark selection, rewarding improvement versus achievement, incentive “cliffs” and “slopes,” time lags, economic externalities, and the choice of penalties versus bonuses (Conrad & Perry, 2009, L&M Policy Research, 2013).

How important is implementation?

In a recent report on value-based purchasing for nursing homes, all state interviewees emphasized consultation and technical assistance (Libersky et al., 2017). The experience in Texas provides an example of a phased approach (Millwee et al., 2013). The analysis began in 2009, and the state first shared results on a confidential basis with hospitals in 2011. The results were explained in many venues. With this information in hand, performance began affecting payment to hospitals in 2014. After a delay, implementation of penalties to plans took effect January 1, 2018. Starting with 2011 data, hospital-specific results have been publicly posted on a learning collaborative Web site (https://thlcportal.com). In 2015, potentially preventable admissions, readmissions, and ED visits were lower than expected by 21%, 8%, and 3%, respectively (Millwee et al., 2017).

What is a realistic timetable for a Medicaid P4O strategy?

Inspiration to implementation of financial incentives can take 3 to 5 years. Yet, benefits can accrue almost immediately, as providers, plans, and policymakers increase focus on outcomes, report results, gain insight, and take action. In Texas, for example, reductions in admissions and ED visits were seen even though financial incentives to plans had not yet been implemented, which we attribute in part to awareness of planned P4O penalties.

We conclude that a focus on outcomes, driven by well-designed payment incentives, ranks among the best routes to health system delivery reform.