“Starting today, you and your friends and your family and your coworkers can get covered, too. Just visit healthcare.gov, and there you can compare insurance plans, side by side, the same way you’d shop for a plane ticket on Kayak or a TV on Amazon.” — US President Barack Obama, 1 October 2013

On 1 October 2013, the United States Department of Health and Human Services (HHS) launched HealthCare.gov, a health insurance marketplace exchange, as the digital centrepiece of the Affordable Care Act (ACA). The site was meant to facilitate dynamic online comparison-shopping of health insurance plans for individuals who could not receive coverage from their employer. In reality the site facilitated nothing. A litany of coding flaws, in combination with an underlying infrastructure completely overwhelmed by massive public interest, made the interface completely unusable. What’s more, the backend processes that provided insurers with critical enrolee information were also non-functional, stifling the paper and telephonic enrolment channels that were meant to supplement online enrolment. For the first two months of a six month open enrolment window, the website facilitated only 365,000 signups, with a projected 7 million enrolments needed to make viable the financing of the new law.

After initial claims that the website “glitches” were signals of high interest and “a great problem to have,” the Obama Administration eventually acknowledged the severity of the technical failures. In a curious mixed reference to policy clichés from both the Vietnam and Iraq Wars, on 20 October HHS announced that “our team is bringing in some of the best and brightest from both inside and outside government to scrub in with the team and help improve HealthCare.gov,” as part of a “Tech Surge.” Enlisting help from software engineers at Google, Red Hat and Oracle and redirecting additional millions of dollars towards fixing the site, President Obama appointed Jeffery Zients, a highly regarded management consultant, to oversee the rehabilitation of the digital infrastructure of what the President hoped would be his signature initiative. Under time pressure to save the open enrolment period, Zients declared, “by the end of November, HealthCare.gov will work smoothly for the vast majority of users.”

In spite of the disastrous rollout, by the end of the open enrolment period in April 2014, health insurance exchange signups exceeded 8 million. Millions of other people gained insurance coverage for the first time thanks to other ACA provisions and the general publicity created by its implementation. This surprising turnaround certainly speaks more to the demand for coverage than the brilliance of the e-portal. Nevertheless, HealthCare.gov’s dramatic improvement in capacity and functionality between October and December allowed millions more Americans to gain health insurance in 2014 than would otherwise be the case, potentially creating a political constituency that will ensure the law’s future survival.

Health insurance reform has been a key platform of progressive policy in the United States since Theodore Roosevelt proposed a “single national health service” in 1912. In 1965, President Johnson memorably cajoled and bullied congressional leaders into supporting both Medicare and Medicaid, anti-poverty measures that provided government health insurance to the very poor and the elderly. While Johnson’s programmes are considered politically untouchable ‘entitlements’ because of their massive popularity, the vast majority of American citizens are still reliant on their employers for access to affordable and comprehensive health insurance.

Sweeping victories in the 2008 elections enabled President Obama and congressional Democrats to enact the progressive health care agenda, passing a large scale restructuring and regulation of the individual insurance market in the face of unified opposition from congressional Republicans. More than four years after passage, the ACA, colloquially referred to both pejoratively and by supporters as ‘Obamacare’, remains a highly controversial and politically salient issue that is central to President Obama’s legacy. Re-litigation of nearly every aspect of the ACA is a crucial campaign issue for the upcoming 2014 congressional and gubernatorial elections. In the midst of this epic political struggle, operational implementation becomes merely the next battlefield of drawn-out political warfare. Successful implementation necessarily fulfils the original vision of the law. Achieving government-initiated service transformation within the otherwise private health insurance marketplace is both required for the law’s success and anathema to the law’s opponents, the very source of its intractable political debate. Therefore, high-ranking politicians demonstrate resistance to otherwise technocratic and operational minutia, lest they appear supportive of the law’s overarching reform agenda.

In assessing the promise, failures, and potential of HealthCare.gov as an eGovernment initiative, we can gain broader insight into the intersection between service quality and political saliency. As “eGovernment service quality plays an instrumental role in determining citizens’ acceptance of public e-services,” public acceptance and therefore political neutralisation of the ACA hinges on the service quality of HealthCare.gov and its ability to develop a satisfied political constituency, (Tan, Benbasat & Cenfetelli 78). This article investigates central components of HealthCare.gov’s original vision for an integrated management plan, the underlying causes of that plan’s initial failure, and possible future directions, based on academic and industry guidance, that can build on the site’s recent success. Key framing questions are identified in the right-hand margin and addressed throughout, grounding proposed future developments within the existing contextual narrative of established design choices.

How did we get here?

The charge of the ACA is to make individual coverage easily accessible and affordable to the millions of Americans that are ineligible for government healthcare services and locked out of the employer-provided coverage market for one reason or another. While individual coverage options existed prior to the ACA, insurance plans were notoriously hard to compare on price and benefits, and various exemptions and restrictions on coverage presented a lattice of obstacles for individual shoppers. The ACA vitiated common medical underwriting practices, where insurers set prices, caps and access based on pre-existing health conditions. Superseding regulations created minimum tiered standards for plans that would be offered in exchange markets. Bronze, silver, gold and platinum tiers enabled would-be shoppers to more rationally compare prices across plans with similar levels of benefits.

The vision for the online exchanges is transformative in both process and outcome, fundamentally changing both the government’s role in the health insurance market and in the consumer’s experience and leverage when acquiring coverage from private insurers. Substantially increasing the number of individuals with private insurance requires a radical disruption. In order to stabilise the exchange risk pools with both healthy and vulnerable populations, new purchasers needed to include not only to those previously shut out of the individual market, but also to those ‘invincibles’ — mostly young healthy males — who had voluntarily rejected coverage.

In conception, HealthCare.gov was meant to transform the health insurance shopping experience, facilitating e-commerce through easy, interactive browsing and comparison of otherwise complex and legalistic products. In addition, the site would provide the user a personalised portrait of new government subsidies available. Paper, telephonic, and in-person channels (reliant on the same website for their underlying processes) are employed solely to broaden the reach of the web interface, acting as mediators for those less comfortable with online services. Generous financial support and a small tax penalty levied against the still-uninsured act as carrots and sticks of the ACA. However, successfully penetrating target markets and substantively reducing the rate of uninsured required trustworthy, straightforward presentation and seamless online navigation, research and transaction.

What caused this mess?

HealthCare.gov’s near non-functionality for two months in late 2013 laid bare the poor development strategy underlying the site. The clear project management failure to rigorously test the system’s functionality before launch or scour for coding vulnerabilities fed impressions of general incompetence and lack of political accountability. Subsequent successes of Silicon Valley experts and imported crisis management leadership teams to correct the site reinforced the sense that these costly flaws were avoidable and unnecessary.

With cost outlays in excess of $650 million, no expense was spared in employing a constellation of 55 different contractors to assemble HealthCare.gov. This outsourced collaboration created a patchwork of architectures with fundamental inconsistencies across components. While parts of the front-end web interface were developed by QSSI under open source protocols (which allow for scrutiny of the underlying code), the backend infrastructure coordinated by lead contractor CGI Federal remains closed to outside investigation. As the flow chart to the left demonstrates, acute failures and emergent security flaws in systems have often stemmed from the failed interaction of one part of the site with another as it attempts to verify identities, calculate subsidy amounts and connect user information with hundreds of private insurers’ internal systems. In one instance, “hitting ‘apply’ on HealthCare.gov cause[d] 92 separate files, plug-ins and other mammoth swarms of data to stream between the user’s computer and the servers powering the government website,” (Begley). Lead contractors and HHS officials are naturally in dispute over who is to blame for the litany of site flaws, each citing various unheeded warnings they sent to the other. Such discord further demonstrates a comprehensive lack of operational and technical leadership during development and deployment of the site.

However, the website’s initial failure can also be seen as a direct product of the political circumstances under which it was developed. Many conscious development decisions were made to avoid industry ‘best practices’ (like open-sourced coding and transparent, measurable benchmarks), because peer review systems could be exploitable by ‘anti-stakeholders’ who actively wished for the law’s failure. The site’s content was closely guarded as a political asset, severely limiting the potential for any participatory stakeholder engagement from either outside web development experts or potential end users. The Obama Administration actively buried the site’s funding streams and governance arrangements within the backwater Centers for Medicare and Medicaid Services IT department, rather than housing their flagship initiative under the newly installed Government Chief Technology Officer — and healthcare startup expert — Todd Park, to insulate site development from retaliatory budget cuts. Even frontline ‘navigators,’ the low level staff deployed to assist individuals in accessing the website through in-person channels, were often recruited from the President’s former campaign apparatus, prioritising cause loyalty above all else. These cumulative design and implementation decisions by political actors consciously traded away assurances of site integrity for assurances of the site’s political survival.

Furthermore, ACA’s constrictive legislative framework governing the site’s construction stifled options for flexibility and innovation that are often taken for granted in the private sector. Open enrolment deadlines fixed by statute meant the many delays early in the development process had to come at the expense of time set aside for interrogation of the coding immediately prior to launch. These fixed nationwide launch dates also eliminated the possibility of beta testing or regional rollouts to probe for flaws, practices that are standard practice for private tech firms launching large-scale initiatives. When comparing Healthcare.gov’s authorising legislation to that of ConsumerFinance.gov, the recently launched face of the Consumer Financial Protection Bureau, the unusual extent of the ACA’s legislative control over web development becomes clear. The Dodd-Frank Wall Street Reform and Consumer Protection Act mentions the term “website” only once in reference to ConsumerFinance.gov, citing only the need to establish one. This allowed developers at a single, newly established agency the latitude necessary to innovate and scale up iteratively, without pre-defined congressional expectations of the final product. In contrast, the Affordable Care Act contains 118 “website” references, outlining complex governance arrangements and providing detailed instructions for HealthCare.gov’s content and development timeline. Since the website was the core facilitator of health exchanges, and the exchanges were a major tenet of the entire legislation, congressional legislators appeared far more invested in shaping HealthCare.gov than ConsumerFinance.gov. The ACA forced health insurance exchanges to mediate multi-jurisdictional collaborations between many state and federal political actors, multiplying decision veto points and constricting flexibility of implementation. As summarised by Ezra Klein, this micromanagement “handed control of the site’s development to lawyers, not engineers,” setting up the site for engineering failure, (3).

Did it have to be this way?

Iterative and incremental launch strategies, long favoured by both academic maturity models and industry standards, were not conducive to the legislative expectation of a nationwide, single-day opening of the enrolment period. Imagining substitute approaches, such as those demonstrated by success stories like ConsumerFinance.gov, must account for differing levels of political scrutiny and public expectations. Promoting agility, embracing risk and learning from failure are more feasible mantras outside of the political spotlight. A quick Google Trend search confirms the stark disparity in relative interest between the two sites and their underlying laws over the last eight months.

In spite of these constraints, we can develop useful alternative benchmarks for the site’s future development by understanding how its development departed from Layne and Lee’s widely used maturity model or from Capability Maturity Model Integration (CMMI) industry standards.

Layne and Lee’s four-stage maturity model adds utility in both its simplicity and optimistic vision of e-government initiatives. Though empirical studies question the applicability of the 2001 model as a predictive driver of government initiatives, Layne and Lee’s competing axes of Complexity and Integration continue to provide a helpful classification method for government e-portals. Classifying HealthCare.gov along the model’s stages quickly reveals the site’s ambitious scope in contrast to Layne and Lee’s gradual evolutionary framework. From an initial state of fragmented and incomplete cataloguing of insurance plans, HealthCare.gov attempts in one step to provide a comprehensive and straightforward menu of insurance options and full transaction capabilities, while vertically and horizontally integrating an untamed web of both governmental and non-governmental databases. Tellingly, the site’s most problematic features involve the horizontal integration of variegated information stores, which Layne and Lee recognise as the most complex phase of government e-portals at both an organisational and technical level.

HealthCare.gov’s lead contractor, CGI Federal, is one of only 10 enterprises with the highest level of CMMI certification. But the sweeping aims, shrinking timetables and unheeded warnings of HealthCare.gov are anathema to the cautious and measured approach preferred by CMMI processes. Originally designed to oversee software development for Department of Defense systems, CMMI focuses on “careful and deliberate development paths according to lower-risk, standardization-heavy and contractually-driven relationships between the developer and the customer,” (Glazer, et al. 5). The industry framework is designed to achieve successful delivery in the first instance given low-trust, contractual bonds and intensive process management. As detailed above, HealthCare.gov’s development processes were constrained by statute and shunted aside in favour of external political demands, even though low-trust relationships dominated. Any alternative approach based strictly on CMMI principles would demand early and continuous measurement of development practices (through appraisals such as the SCAMPI method) to document any process maturation against expected outcomes from the CMMI model. These benchmarks, once considered perilous vulnerabilities by HealthCare.gov’s political actors, should instead be conceived as learning devices that improve transparency and promote future site efficacy.

Eschewing these established, plan-driven models, HealthCare.gov’s emergency management overhaul in November 2013 instead implemented a variation of Agile, the value-driven process favoured by Silicon Valley startups and site subcontractor QSSI. By clustering groups of engineers into high-trust teams and reducing management layers, the new structures pushed technical fixes out quickly and iteratively with Zeints’ mantra—that “healthcare.gov will work smoothly for the vast majority of users,” by the end of November—as the only driving goal.

Whether this management model can continuously function within government bureaucracy is an open question. While acute crisis leadership enables temporary disruption and shortcutting of institutional channels, the end goal of that leadership should be a return to normalcy, and with it normal processes. In their crisis management model, Boin, ‘t Hart, Stern, and Sundelius stress that “governments — at least democratic ones — cannot afford to stay in crisis mode forever,” and leaders must transition from emergency to routine operations and provide full and accepted accounts of what happened in order to maintain legitimacy (14). Agile processes de-emphasise the collection of process documentation, preferring evaluation based solely on results and outcomes, which constrains the full accounting that CMMI institutionalises. Despite the demonstrated success of the “Tech Surge,” a more sustainable vision is required to propel the site through future open enrolment periods.

Where do we want to be?

In developing a sustainable plan that fully realises the potential of HealthCare.gov as a transformative eGovernment initiative, we must consider not only a repair of the service delivery quality (i.e. fixing technical flaws that affect site accessibility and security) but a citizen-centric upgrade of the service content quality as well. Confounding Tan, Benbasat and Cenfetelli’s rigid separation of these two concepts, HealthCare.gov’s initial failures saw them feed upon each other. Requirements that site visitors create detailed user accounts before they could browse available health plans swamped the registration system and crashed the site with unanticipated loads of curious but uncommitted shoppers. In terms of Tan et al.’s schema, the absence of a ‘Trying’ service content function directly mediated a ‘Navigability’ service delivery function. Integrating the interests and demands of variegated site users — in this case both committed insurance buyers and curious browsers — into the design will better facilitate both service delivery and content quality improvements.

In order to credibly match HealthCare.gov’s initial transformative vision, redesigned content and delivery mechanisms must speak directly to user expectations, often set by global eCommerce sites that have mastered the online shopping experience in other markets. Functions that allow casual browsing, modifiable “shopping carts” and side-by-side comparisons are essential to even the most remedial eCommerce sites, let alone the tech industry giants that President Obama regularly compared with HealthCare.gov. Matching these features with the near constant uptime, worry-free security mechanisms, and intuitive navigation structures that are standard on established sites, while considered a major feat for the “tech surge” engineers, will meet only the most basic expectations of online shoppers. As the health exchanges represent entirely new government services within a previously loosely-regulated commercial market, HealthCare.gov rightfully engenders these comparisons to commercial sites and needs to surpass them in order to be perceived as exceptional and transformative.

A long term consequence of the Government’s newly-active role in the individual health insurance market lies in the citizenry’s newly projected Government responsibility to adjudicate and moderate commercial transactions between individuals and private insurers. Before the ACA, American insurance shoppers could blame government officials for little more than non-involvement, but HealthCare.gov users now carry a reasonable expectation of government assistance and protection with regard to their privately purchased health insurance. Further, the press, academics and interested citizens have a much stronger claim to government involvement with widespread awareness of the number and demographic composition of insurance signups. Institutionalising user ownership over the transaction process and broad public transparency of the generated data, therefore, is vitally important to satisfying democratic expectations and accommodating the increased responsibilities associated with the government’s new role.

While Tan, et al.’s analysis suggest that such “ownership-oriented service content functions” only improve site content quality when transactions with government are frequent — not the case with HealthCare.gov’s once yearly open enrolment — the authors did not test for the complexity of transactions. The scale and intimacy of purchasing health insurance, particularly for the first-time buyers that are central to the site’s success, requires an intensity of shopping and comparison more applicable to leasing an automobile than Tan et al.’s benchmark of government websites accepting payment for traffic fines. HealthCare.gov’s unique role and massive scale require achieving a challenging design synthesis between advanced eCommerce applications and mature eGovernment goals in order to fully integrate user expectations and successfully deliver transformation.

How do we get there?

Improving the ownership and transparency service dimensions of HealthCare.gov requires a large scale move away from the proprietary models that continue to underly the site’s backend coding infrastructure. Allowing outsiders to scrutinise code will enable rapid repair of flaws and ultimately lead to more secure software platforms, though conferring some ownership of that security maintenance beyond government control will disrupt current bureaucratic models. Metrics on site signups and relative user interest in available plans need to be publicly accessible real-time through an automated IT dashboard, to build confidence and trust in the sustainability of the underlying reforms, even if this disrupts public relations schedules or raises political concerns. More broadly, the anonymised data and supporting code generated by government IT initiatives must be public by default in order to re-conceptualise the development of public IT infrastructure as public discourse. This frames eGovernment initiatives as openly evolving and always in need of improvement, rather than a proprietary product.

Further, to sustainably maintain the high-trust relationships offered by the recent Agile-oriented surge, government IT departments need to rebuild their own capacity, revamp their dismal reputations and actively recruit from the same knowledge pools as tech giants. The interoperability failures between different contractor platforms dramatically highlighted the need for IT projects to have a considered, in-house strategy, design and technical development in order to ensure a uniform vision and clear management direction. Industry engineers could then be engaged as peers and collaborators throughout the design process, avoiding the blame-shifting common with contractual relationships and expensive redundancies of future tech surges.

Finally, the governance structures authorising these massive, ICT-enabled service transformation initiatives need to encourage sites to start small and iteratively test and release applications. Legislative statute can offer both target outcomes and transformative visions for eGovernment initiatives without prescribing a litany of constrictive benchmarks and operational minutiae. Defining fixed resource streams and then encouraging a creative development process—driven by user demands and available capacity—will encourage better government websites while avoiding the ‘too big to fail’ catastrophes all too common among fixed requirement ‘waterfall’ projects. Successfully implementing open-source, in-house, and fixed-resource reforms can potentially de-politicise the implementation of even highly political eGovernment initiatives, allowing for less nihilistic political debate.