The federal bureaucracy played a critical role in implementing most aspects of the Affordable Care Act's private insurance coverage expansion. Through brief case studies, the authors review three dimensions of this role: the development of the Center for Consumer Information and Insurance Oversight, rulemaking in the formulation of the essential health benefits package, and the implementation of the federal website. They relate these to themes in the public administration literature. Politics—both through state decisions and through continuing congressional action (and inaction)—pervaded the implementation process. The challenges of staffing and situating the new bureaucracy effectively changed vertical boundaries within the Department of Health and Human Services, with long-lasting consequences. Finally, the complex design of the policy itself made passage of the legislation easier but implementation much more difficult. Ultimately, however, implementation was remarkably successful, achieving improvements in coverage consistent with the Congressional Budget Office's projections.
The Affordable Care Act (ACA) passed on March 23, 2010. By 2015, the number of Americans with health insurance had increased by over 20 million (Zammitti, Cohen, and Martinez 2016). Tens of millions more had seen improvements in the scope and quality of their coverage. Through the ACA, Congress appropriated the funds, authorized the taxes, and described the rules that led to these changes. But the ACA also required substantial executive action, in the form of new regulations governing the behavior of others and in operationalizing new programs, to have actual effect. The ACA reads “The Secretary shall” 1,563 times (Conover and Ellig 2012). Each of those “shalls” required someone in the executive branch of government to do something. The ACA appropriated $1 billion to cover internal government costs for all this “doing.”
Many important elements of the ACA were modifications of existing programs. The ACA made important changes to Medicare payment methodologies and benefits. Much of the coverage expansion in the law occurred through Medicaid. These Medicare and Medicaid changes required enormous effort and often substantial creativity on the part of the federal bureaucracy, but they fit into established structures. Medicare had changed payment methods and expanded benefits before. Medicaid eligibility had been expanded over time, and the rules governing eligibility had changed. While these reforms and expansions touched thousands of providers and millions of people, it is testimony to the strengths of the federal administrative system that modifications of this scope and scale can be considered part of the “normal science” of federal programs.
Many of the elements of the ACA, however, involved incursions of the federal government into areas where there had been little prior federal presence. Most significant of these were changes made to the rules governing private insurance, the design of nongroup insurance benefits, and the establishment of health insurance exchanges. As Brown (1983) described in the case of the implementation of the 1973 HMO Act, this early implementation process can be decomposed into discrete organizational tasks: the development of a new bureaucracy, rulemaking, and implementation. In the next section of this article, we briefly describe these three sets of activities. In the final section, we consider broader questions they raise about policy implementation.
Organization: The Evolution of the Center for Consumer Information and Insurance Oversight
The ACA's regulation of private health insurance had two sets of components. One group of components was intended to be permanent and to lay the groundwork for the transformed health insurance market that would become fully operational with the opening of the Exchanges (later Marketplaces) in 2014. These included regulations governing medical loss ratios, coverage of dependents, elimination of annual and lifetime limits on covered expenses, and risk adjustment rules. The second group of components were intended as stop-gap measures that would remain in place only until the exchanges opened. These components—now virtually forgotten—included the Pre-existing Condition Insurance Plan, a set of high-risk pools operating in each state; elimination of preexisting condition exclusions for children; an early retiree reinsurance program; and an informational website that displayed current insurance prices in the nongroup market.
Most of these activities were federal extensions of activities that had long been undertaken by state insurance regulators. The ACA's rules in this sphere were largely intended to be implemented through state insurance commissioners, with the goal of increasing the consistency of insurance regulation across the country.
The Secretary of the Department of Health and Human Services (HHS), Kathleen Sebelius, had extensive experience in private insurance (having served as insurance commissioner prior to becoming governor of Kansas). Outside the secretary's office, however, HHS's familiarity with private insurance was limited to the Medicare Advantage and Part D drug programs, which posed their own challenges but were relatively circumscribed and addressed a known population. Because of the secretary's expertise, and the expectation that HHS's role would be largely limited to working with states to implement a set of consistent regulations, HHS initially assigned responsibility for private insurance regulation to a new office reporting to the secretary, the Office of Consumer Information and Insurance Oversight (OCIIO), and hired another former insurance commissioner to lead this office (GAO 2013).
OCIIO, unlike a typical state insurance regulator, would, in many cases, be adopting regulations that governed both state insurance markets (plans not exempt from the Employee Retirement Income Security Act of 1974 [ERISA]) and parts of the health insurance market that fell outside those bounds. Operationally, this meant that, in addition to HHS, the Department of Labor, which regulates self-insured/ERISA plans, and the Department of the Treasury, which regulates tax-favored health programs (such as health savings accounts), would need to work together to develop all regulations. Given the political salience of the ACA and the need to coordinate efforts across departments, the White House held meetings twice a week in the Eisenhower Executive Office Building with the principals from each of the agencies responsible for implementing the ACA, to ensure this coordination. The $1 billion in implementation funding would need to be divided fairly across all these agencies.
This might have been a recipe for serious interdepartmental strain. In practice, while there were certainly vigorous debates in the Eisenhower Building, there was remarkable consensus about the need to move forward, partly in response to mounting partisan opposition. Notably, over the entire period between passage of the law and the opening of the marketplaces, there were no significant media leaks about the content of upcoming regulations or about internal disagreements. Nonetheless, the need to work collaboratively across three departments absorbed considerable time and energy.
The second set of challenges emerged in the regulation of new markets, where the evidence base was limited. For example, available surveys of annual limits in health insurance plans asked only whether such limits were below $250,000 (National Archives 2010). The promulgation of regulations governing these annual limits revealed the existence of millions of “bare-bones” insurance policies, with annual limits in the single- or double-digit thousands of dollars. OCIIO had to develop and implement a contentious emergency waiver process to hold together these shabby policies until 2014.
The first year of regulatory activities made it clear to HHS that the human and fiscal resources needed to implement the ACA would be far greater than anticipated. The programmatic challenges were magnified by partisan opposition to the ACA at the federal, state, and public level, which hardened over time. Congress had appropriated substantial funding for states to implement insurance regulations and to build their own state-based exchanges. While initially most states cooperated with OCIIO in standardizing regulations, such as rate review, and all but one had accepted funding to begin building exchanges, over time cooperation waned, with many states choosing not to build their own exchanges and returning grant funding (Fulton et al. 2015; Gluck and Huberfeld 2018; Jones 2017; Jones, Bradley, and Oberlander 2014).
The election of a Republican House majority in 2010 meant that further funding for implementation of the ACA was unlikely, though estimated implementation costs were 5–10 times higher than the $1 billion appropriated (Levinson 2016). In the face of these challenges, in early 2011 the secretary made the decision to move OCIIO into the Centers for Medicare and Medicaid Services (CMS), where it was renamed the Center for Consumer Information and Insurance Oversight (CCIIO). By moving OCIIO into CMS, the secretary could achieve economies of scale using CMS's expertise in the regulatory process, contracting, management, resources, and technical support (Gogan, Davidson, and Proudfoot 2016).
Writing the Rules: Essential Health Benefits
Section 1302(b) of the ACA calls on the secretary of HHS to define the scope of essential health benefits (EHBs) that shall be made available to those obtaining insurance coverage in the new marketplaces. The law specifies 10 broad categories of benefits that must be included in the EHBs; it limits the scope of these benefits, requiring that they equal those in the “typical employer plan”; and it adds a grab-bag list of considerations for the secretary in developing the EHBs. In addition, section 1311(d)(3) of the ACA allows states to mandate the inclusion of benefits in addition to those in the federal EHBs but requires that states defray the cost of any such benefits. Beyond its vaguely worded limitations and requirements, the ACA offers little in the way of specificity around these benefits.
The EHBs would form the basis of insurer's bids in the new marketplaces. The ACA's provision calling on states to pay for the costs of mandates that exceeded the scope of the EHB contemplated further state legislative action to repeal such mandates. The regulatory process for developing and releasing a final regulation would take at least 6 months. To give states and insurers time to act in time for the January 1, 2014, start date for the marketplaces, HHS would need to have a well-defined idea of the scope of the EHB before late 2011 (Bagley and Levy 2014).
The administration proceeded along three parallel paths to meet this timetable and objective. First, in the fall of 2010 the Office of the Secretary contracted with the Institute of Medicine (IOM) to convene an expert panel and deliver a consensus report recommending a process for defining the initial benefit package and for updating the benefit package over time. Second, the Department of Labor began its analysis of data from the National Compensation Survey on the scope of benefits in “typical” plans. Finally, the HHS heard from various interest groups with views on the regulations.
The IOM released its initial report in early October 2011 (see Ulmer et al. 2012). It recommended that the secretary begin with a small employer plan, add benefits to round out the 10 categories required by the law, and then trim benefits to hold the overall cost of the package to the cost of a typical plan. It suggested that HHS engage in a structured, deliberative process of soliciting public input, through a series of small-group meetings, to further prioritize services within the plan. Finally, the IOM proposed that a national board update the benefits over time, adding and subtracting benefits to hold costs to a premium target.
This proposal raised serious political and logistical hurdles. The administration was struggling with vehement political opposition to the ACA's Independent Payment Advisory Board, and the idea of proposing another board that could be tarred as a “death panel” to determine benefits before the critical 2012 elections was unrealistic (Spatz 2011). In addition, the proposal's target premium idea, designed to encourage states to drop costly mandates, would put tremendous pressure on states at a time when the administration was trying to encourage them to expand Medicaid and establish marketplaces. Finally, the small-group deliberative process was incompatible with the timetable for releasing benefits.
The Department of Labor's analysis of its existing survey data provided even less granularity than expected about the scope of benefits. Examination of plan documents specifying benefits was similarly unilluminating. There turned out to be relatively little variation across markets in the principle elements of plans. Differences among plans emerged in access to a limited number of specific services: bariatric surgery, infertility treatment, chiropractic, hearing aids, and autism treatment (Skopec et al. 2011; Uberoi 2015).
The three paths yielded very little insight about how to proceed. Fortunately, after the ACA passed, one of the lead finance committee staffers who had subsequently joined the administration suggested a way out: build on the model used for the State Children's Health Insurance Program. Under this program, states could choose a benefit package from a menu of existing choices offered to state and federal employees. This approach avoided the messy work of weighing one benefit against another. Given the law's requirement that benefits reflect a typical plan, and the IOM's recommendation that benefits reflect a small, rather than large, employer plan, HHS added the three largest small employer plans in the state to the menu of options. This extension had the added feature that most state mandates applied to small employer plans. That meant that if a state chose a small employer plan as the basis of its EHB, all existing state mandates would fall within the scope of the EHB, and the state would neither have to pay the cost of these mandates nor repeal them. Given the partisan rancor around the ACA, the administration anticipated that some states would not make an EHB selection and designated the largest small employer plan in the state (which would generally incorporate state mandates) as the default (CCIIO 2011).
The “menu” idea solved many problems, but it was completely outside the bounds of what anyone had expected, especially because there had been no press leaks signaling that the menu approach was a direction under consideration (Bagley and Levy 2014). The requirements of the federal rulemaking process meant that if HHS released the proposal as a Notice of Proposed Rulemaking (NPRM), there would need to be an extended comment period and a formal process of revision and response. The administration worried that there might be a fundamental flaw to the idea, in which case following the NPRM process would be fatal to the exchange launch timetable. After extensive consultation with the Office of the General Counsel, the HHS decided to take the unorthodox step of releasing a bulletin describing the idea in December 2011 (Bagley and Levy 2014).
Reactions to the bulletin were generally positive, praising the choice as “a deft political move” (Levey 2012) or grudgingly accepting, given the political constraints (Sage 2011). The HHS released an NPRM governing the collection of data to establish the menu in June 2012; an NPRM on the scope of EHBs in late November 2012, after the 2012 election; and a final rule on February 20, 2013, well after state and insurer decisions on exchanges had been made.
Building the Federal Marketplace
A component of the ACA legislation required the development of marketplaces—consumer-facing website(s) that would allow qualified individuals to “sign up” for insurance electronically. The federal website, HealthCare.gov, went live on October 1, 2013, to worldwide media attention, in the midst of a government shutdown. On that first day, only six people were able to enroll, as the website crashed. Only by early December, after a harrowing, costly, and massive intervention, did the website work properly (Levinson 2016). While the initial failure of the website had only modest impacts on the number of people who signed up for coverage in the first enrollment period, it took a considerable political toll (GAO 2015; Levinson 2016; Light 2014; Polsky, Weiner, and Becker 2014a).
The website comprised two quite distinct information technology (IT) platforms. First, on the front end, there would need to be a consumer-friendly website that allowed an individual to shop for Medicaid or private insurance. Second, on the back end, this consumer-facing website would need to access a great deal of highly confidential federal data (citizenship or immigration status, tax returns, identity through social security, and eligibility for other health care programs) to verify eligibility for private coverage, subsidies, and Medicaid. This back end became known as the “Hub.”
The ACA had contemplated that each state would establish its own consumer-facing website, integrated with its Medicaid programs, and Congress had appropriated funds for this purpose. CCIIO would need to develop regulations governing the functioning of these state websites, including the application form for insurance, and protocols for the interaction between these state-level sites and the Hub.
There would be a single Hub for the federal and state-based marketplaces, operated by CCIIO in collaboration with the various agencies to be queried. Funding for the Hub was intended to come out of the $1 billion appropriation. The front-end, consumer-facing websites would need to collect the extensive information needed to verify eligibility through the Hub.
Most observers had presumed that states, which had regulatory authority over insurance, would want to manage their own exchanges (Levinson 2016). Initially, all but one sought and received an exchange grant for that purpose, with $3.6 billion in exchange planning grants dispersed (Gluck and Huberfeld 2018). As the ACA litigation progressed, several Republican-led states pulled back for political reasons, returning their planning grants (Gluck and Huberfeld 2018). The technical challenges of building a consumer-facing website deterred others (Vestal and Ollove 2013). Thus, by the time states were required to certify the operation of the website in early 2013 (a date that was delayed by HHS to encourage more states to participate), only 14 chose to proceed with their own designs, leaving HHS with responsibility for 37 consumer-facing websites, each linked to that individual state's Medicaid infrastructure (Dinan 2014; Gluck and Huberfeld 2018; Gogan, Davidson, and Proudfoot 2016; Noh 2016).
The development of the Hub seemed the more complex build, as it would require interfacing with varied IT systems across the federal government (GAO 2013), but the Hub has actually been a great unsung hero of federal technology development, as it has functioned nearly flawlessly since open enrollment in 2013. The front-end, consumer-facing websites ended up being the challenge, both for the federal government and for the states that chose to build their own sites. Ultimately, a total of 15 consumer-facing websites were built: one federal Marketplace and 14 state marketplaces (Vestal and Ollove 2013). Their varied experiences help shed light on the key problems. Ultimately, about one-third the state marketplaces ran smoothly from the outset; about one-third, including the federal Marketplace, had a mixed performance in the first enrollment period; and about one-third had significant and enduring problems (Polsky, Weiner, and Becker 2014a, 2014b; Vestal and Ollove 2013). This variation in performance can be traced back to both design and contracting decisions (Fagnot, Ye, and Desouza 2018).
The first key design decision, which challenged the design of all the consumer-facing websites, was the application process. In its effort to exclude undocumented residents and to minimize crowd-out of existing insurance coverage, the law required a great deal of information about enrollees. Individuals seeking insurance could not be eligible for other forms of health care coverage (employer-sponsored insurance, unless the firm's offer of coverage was unaffordable; Medicare; Medicaid; or coverage through other federal programs) and had to be legal residents of the United States (GAO 2013). Income status and family size would further determine eligibility for subsidies or Medicaid.
Despite considerable efforts to simplify the collection of this information through the application process, it remained very complicated. The diversity of American families and their situations—new citizenship, adoptions, income fluctuations, employment shifts, and other variations—created a vast number of permutations to eligibility verification and subsidy calculation. These complexities could, in principle, be overcome, and the IT services company CGI Federal did conduct two successful demonstrations of the federal website in the weeks before launch (Levinson 2016). But such packaged demonstrations only showed that the system could handle complexity when inputs fell within expected parameters and the volume of queries was modest. At actual launch, multiple unanticipated permutations arose at high volume, applications were aborted, and, depending on system design, the strain on the website infrastructure became overwhelming.
The second key element was the contracting choice. In the case of the federal website, the set of potential vendors for the consumer-facing Marketplace contract was constrained by federal requirements; effectively choice was limited to contractors who had been prequalified years earlier. Ultimately, the contract was awarded to CGI Federal, which already had work under way within CMS (Gogan, Davidson, and Proudfoot 2016; Levinson 2016). In retrospect, CGI Federal's approach to the technical problem of building an exchange was likely flawed—the company was also the lead contractor for three of the least successful among the state exchanges: Hawaii, Massachusetts, and Vermont (Polsky et al. 2014a; Vestal and Ollove 2013). By contrast, the best-performing state exchanges (Connecticut, Kentucky, Rhode Island and Washington) were all contracted with Deloitte, which made a different set of web architecture and management decisions.
Finally, CMS (and some states), in the interest of getting the website finished in time, decided to delay the implementation of a “shopping” function beyond the open enrollment start. That decision meant that the hordes of people who were simply curious about the website but had no intention of buying coverage had to initiate a “fake” application to see the site. The flood of applications overwhelmed the site and led to the famous crash. Had shopping been available, visitors to the site would not have initiated an application process and there would have been less stress on the system.
The dramatic failure of the system led CMS, HHS, and the White House to mount a substantial rescue operation, which included shifting the technical work on the website from CGI Federal to QSSI, the contractor that had successfully built the Hub. CMS faced tremendous pressure to solve the problems. Contractors, IT experts, and staff worked 24/7 to solve the issues, even as a government shutdown limited the staff and resources available. The environment of 24/7 scrutiny by the media, Congress, advocates, and the White House unnerved everyone.
The key first step to righting the ship was introducing a shopping function, which took some burden off the application. Next, as the application support team gained experience, they were better able to help families with more complicated situations get through the process. Finally, CMS went through the painful process of identifying and solving IT bottlenecks and working a punch list. By January 2014 the number of people enrolled through the site had exceeded 1 million and the pressure eased.
Several common themes flow across these three elements of ACA implementation, which had both short-term effects and long-term consequences.
Politics and Progress
The politics of the ACA clouded internal implementation efforts in several ways. Most obviously, state-level partisanship around Medicaid expansion and exchange establishment left the federal government with an unexpectedly large regulatory and implementation burden (Gluck and Huberfeld 2018; Oberlander 2016). But congressional politics also mattered. For example, exchange rules had to be in place in time for states to make decisions before January 2013. But the possibility that the Democrats might lose the Senate and presidency in 2012 pushed deadlines even further forward, out of fear that a new Congress would use the Congressional Review Act to nullify new regulations (Conover and Ellig 2012; Copeland and Beth 2008). In consequence, no substantial new HHS regulations were promulgated between late May 2012 and the elections that November, when the EHB NPRM was finally released.
The partisan shift in the House also interfered with the ability to manage the complex website contracts and regulatory processes, as congressional oversight intensified, with frequent requests for the secretary and CMS administrator to testify and enormous documentation requests, entangling the same staff tasked with regulatory development and coordination with insurers and states.
The ACA politics of 2010–14 continues to exert substantial influence on health care policy making. Most notably, a less partisan state environment might have led to the implementation of successful, state-managed, consumer-facing websites in more states, reinforcing the role of the states in future health reforms (Gluck and Huberfeld 2018).
Building a Bureaucracy
The short- and long-term success of a new public program depends on effectively drawing the vertical lines establishing and situating a new bureaucracy (Brown 1983; Kettl 2006). In sharp contrast to the evolution of a private-sector business, government bureaucracies do not arise spontaneously and grow incrementally. They must emerge from existing structures at scale, ready to serve tens of millions of customers on day one.
At the moment the ACA passed, HHS was already stretched thin managing ongoing programs and administering the rapid dispersal of over $17 billion in new American Recovery and Reinvestment Act Funds (Redhead et al. 2009). The day after the law passed, the Office of Personnel Management provided the HHS with direct hire authority, to facilitate hiring up to 1,480 people (Levinson 2016). However, relatively few people anywhere had the expertise needed to write the required new regulations; even fewer had experience working within the federal bureaucracy; still fewer had the ability to evaluate candidates for such positions. It had taken some 6–9 months to put staffing in place for the implementation of Medicare Part D back in 2003, and that was a much more bounded program within an existing infrastructure (Serafini 2010). The shortage of funding; endless oversight by the White House, Congress, and the media; and lucrative opportunities to work on state- and private-sector implementation of the law made it particularly challenging to recruit and retain effective senior-level managers within OCIIO. These funding and staffing challenges ultimately led to the move of OCIIO into the existing bureaucracy of CMS.
Integrating OCIIO into CCIIO stretched the role of the CMS administrator. Rather than acting as a quasi-apolitical manager administering well-established programs, the administrator's responsibilities now extended across the entire US health care system into the private sector. In this new structure, the recent clash between HHS Secretary Alex Azar and CMS Administrator Seema Verma over the future of the ACA may presage a series of battles to come (Pradhan, Cancryn, and Diamond 2019).
Policy Design and Execution
The design of the ACA itself posed significant challenges for implementation. The decision to generate immediate wins by starting transitional programs immediately (often with start dates of January 1, 2010, before the law had even passed) and then closing them out as new, permanent programs took their place proved to be very costly in terms of human resources (Conover and Ellig 2012). The transitory programs sprouted challenges of their own, requiring management and attracting congressional oversight.
The effort to constrain the costs of the program through the complex design of eligibility criteria meant that the website design became much more fraught. This wasn't building Travelocity—there was a sophisticated back-end eligibility system that had to accommodate permutations of the American family difficult to imagine. Ultimately, technical fixes saved the website. Computer designs today are quite capable of managing complex algorithms that include financial (means) testing and choices among dozens of insurance plan. But even the best computers don't eliminate the tensions inherent in the concurrent complexities of people's living situations and of choice-based program designs. The complexity of the ACA's design itself may be one source of the current interest in a conceptually simpler, non-means-tested, Medicare-for-all structure.
Implementation and Success
At the end of the day, the proof is in the pudding. Implementation of the ACA was far from perfect. Bugs big and small infected execution. The image of the HealthCare.gov website crashing is seared into the eyes of many health policy and management observers. But by the end of that first year, the US uninsurance rate had fallen by 7 percentage points, consistent with the Congressional Budget Office's projections at the time of enactment (Glied, Arora, and Solís-Román 2015; Zammitti, Cohen, and Martinez 2016). And as advocates had promised, Americans had better access to care and lower exposure to medical debt and bankruptcy. This tremendous historical success is owed, in large measure, to that much maligned institution, the federal bureaucracy.
We thank Aggie Tang for very useful research assistance. We thank the Commonwealth Fund for research support.