Lessons from Pandemic Medicaid Automation for Work Requirements

Small tech capacity investments offer big returns

In the next two years, new work requirements in the Medicaid program will strip health insurance from around 5 million people. But that massive coverage loss largely won’t be among ineligible people – most will lose their health insurance because they’ll be crushed by the paperwork.

Many states are now frantically preparing to minimize coverage loss, while facing the reality that work requirements are expensive to implement. Since Georgia added their work requirements in 2020, they have spent twice as much on administrative costs and consultants as they have on actual healthcare benefits.

States need tools – and they need them quickly – to keep their administrative costs in check, while ensuring eligible people don’t lose life saving coverage.

At the United States Digital Service (USDS), my team and I helped states automate Medicaid renewals after the COVID public health emergency ended, preventing coverage loss for millions of people. The speed, scale, and consistency of our interventions’ success offers lessons in how small investments in state capacity can dramatically improve benefits delivery – lessons which are especially urgent as states face the crisis of HR.1 implementation.

New research published in Health Affairs documents the impact that our work had in ensuring people did not lose their health care. Our team at Better Government Lab (Jeremy Barofsky, Pam Herd, Eric Giannella, Don Moynihan and myself) rigorously estimated the impact of a small USDS team on automatic renewals to protect Medicaid. While descriptive reports have already pointed to very large impacts of this USDS effort, our Health Affairs paper provides the first peer-reviewed estimates that account for state-specific policies and trends.

Our analysis focuses on the impact in four USDS partner states - South Carolina, New York, California, and Wisconsin.

We find that compared with other states, the USDS intervention led to:

• Automatic “ex parte” renewals increased 21.6 percentage points

• Overall renewal rates increased 7.7 percentage points

• Procedural denials dropped 8.3 percentage points

The USDS’ interventions roughly doubled the number of eligible people receiving automatic renewals and nearly halved the number of people denied for administrative reasons in the states analyzed. On the ground, these improvements seamlessly extended Medicaid for millions of eligible Americans who otherwise would have lost their health coverage because of paperwork burdens.

Figure 1: Medicaid ex parte renewal rates, overall renewal rates, and procedural denial rates before and after states received USDS assistance, for the 4 states analyzed.

Remarkably, the intervention worked across diverse contexts: red and blue states, expansion and non-expansion states, large and small, rural and urban. And it happened fast. Often within one month, states that had been lagging became some of the best performers in the nation.

Figure 2: Unadjusted Medicaid ex parte renewal rates for 4 states that received USDS assistance compared with states that did not. While most states expanded automatic renewals during the unwinding at the urging of the Biden administration, our new analysis shows that USDS interventions helped lagging states make rapid gains.

The post-pandemic administrative burden crisis

So what did we actually do to help states? In early 2023, my USDS colleagues and I piled into a van and began driving around the country to help State Medicaid Agencies after the pandemic. Our team was just six people with a simple idea: states had everything they needed to renew Medicaid coverage for millions more people, but lacked the capacity to execute. And the crisis they faced was urgent.

When the Public Health Emergency ended in March 2023, over 90 million Americans enrolled in Medicaid were thrust into a maze to renew their coverage for the first time in three years. State governments, depleted by the pandemic, scrambled to hire caseworkers, reboot software systems, and inform enrollees of fast approaching deadlines.

The early data was devastating. About 70% of people losing coverage in the first months weren’t ineligible – they just hadn’t completed their paperwork. Kids who qualified for Medicaid were disenrolled because their parents missed a form, couldn’t find the right documentation, or encountered a glitchy portal. We met families that had discovered they lost their coverage for the first time when they showed up to the doctor’s office or tried to fill their prescription.

We knew that states and enrollees were racing against the clock. Each month a new cohort of enrollees were moved through broken systems, and hundreds of thousands of eligible people were at risk of losing their coverage.

A tiger team model of federal-state collaboration

Our small USDS team - composed of engineers, designers, and product managers - coordinated with the Center for Medicaid and CHIP Services (CMCS) to rapidly deploy a new kind of federal-state technical assistance. We would not be writing policy documents or looking for compliance errors, but rather working side-by-side with state staff - from frontline eligibility workers and back office IT staff to vendor teams and senior leadership - to identify problems, design solutions, and implement fixes fast.

We focused on “ex parte” or automatic renewals as the most promising tool to help states and enrollees. This pathway uses data that states already have to automatically verify a person’s eligibility instead of making them fill out redundant forms. By law, state governments are required to attempt automated renewals for all enrollees. The automatic renewal pathway was attractive to states as it presented virtually zero burden for enrollees and state caseworkers while saving money on administrative overhead.

Despite the clear benefits of automatic renewals, very few states were successfully using them at scale. Even states with similar policies on paper had quite different rates of automatic renewals. For example, in July 2023 Pennsylvania renewed less than 5% of its population automatically, while Oregon renewed 86% automatically. As we began our work, ex parte rates ranged from less than 25 percent in eleven states to more than 50 percent in eighteen states.

A workshop with state staff and community based organizations in New York.

Our team bet that in the flurry of post-pandemic policymaking, the federal government could help with a new lever: improving on-the-ground implementation. Broken automatic eligibility software, confusing paper notices, years of accumulated vendor mistakes - these were likely responsible for the divergent outcomes across many states, not just their policy on paper. When we arrived in person to work with a state, the particulars of these problems varied, but our approach remained consistent:

• The code is the policy: We’d walk through a state’s automatic renewal rules as they were actually encoded in the software and use data to find where enrollees were falling out unnecessarily. Many states had years of policy miscommunications and vendor mistakes baked into their systems and never had the capacity to take a second look. We hosted marathon meetings that brought together everyone from state policy directors to back office IT staff which revealed dozens of opportunities to make improvements. “No, you should not be multiplying quarterly income by 12 to calculate annual income.” “Yes, you can use SNAP data.” “No, you should not deny children because their parents are ineligible.”

• Technical vendor management: Most states contract with private vendors to build and run their eligibility systems. Even when state staff find an error or want to make a change they often struggle to get vendors to make changes quickly, accurately, or cost effectively. Having USDS engineers work directly with the vendors altered the dynamic entirely. We could troubleshoot technical approaches, recommend more efficient solutions, and push back when vendors claimed something couldn’t be done. In many instances, this technical expertise enabled changes to be made in days instead of months.

• Human-centered design: We talked to the people actually using these systems every day, including frontline eligibility workers and community navigators. We convened sessions with Community Based Organizations (CBOs). What did they find confusing? Where were the bottlenecks? What errors did they have to manually override? These conversations became an invaluable asset for sourcing both problems and solutions – from software bugs to new paper envelope designs.

Figure 3: A sample renewal funnel that traces outcomes for one cohort of enrollees.

After four to six weeks of intense collaboration and one week driving around the state in a van meeting with state leaders, staff, vendors, and communities, we would have detailed fixes ready to go live. From updates to automatic eligibility software to new paper notices, our goal was always to implement improvements to impact the next month’s cohort of enrollees.

“..USDS highlighted that we had some unnecessary blockers that were preventing cases from passing through the ex-parte process. They quickly walked our business and system teams through the identified areas and helped press our system vendor to make the changes as quickly as possible by offering technical support and reassurance. We were able to implement the system change in two weeks and the next month, our ex-parte rate had increased by 15%!” - Deputy Medicaid Director, Hawaii

In just under a year, we conducted this intense rapid response technical assistance on the ground in eight states covering over 40% of the nation’s Medicaid enrollees: Michigan, Hawaii, Wisconsin, California, New York, South Carolina, Kansas, and New Jersey. In every state, it worked.

We share more details about the nitty gritty of our approach in the state medicaid renewals playbook.

What this means for the coming crisis

The success of this intervention reveals the severity of the underlying diagnosis. If our small teams of six people could have such a dramatic impact, then the gap in state capacity that we were filling must be profound. After decades of outsourcing to poorly performing vendors and failing to build robust capacity in-house, state Medicaid agencies were left with broken systems and little technical expertise in a crisis.

Now, with sweeping changes required by HR.1, states face a new implementation crisis. They will again need to build new software, create new workflows, craft new notices, and train staff on new processes - precisely the kinds of challenges where the capacity to implement makes all the difference.

Strategies to keep people enrolled will be similar to those used by our USDS team during the unwinding: optimizing ex parte renewals aggressively, connecting to new sources of data to verify eligibility, and using human centered design to find and reduce the friction on enrollees wherever possible.

If our experience offered an exciting new model of how to make federalism work – federal tiger teams working hand-in-hand with states to enhance their capacity – that model is less feasible today. The United States Digital Service no longer exists in the form that executed this work and CMCS’s capacity to help states with implementation is severely diminished. USDS was replaced by DOGE, and I and many other civic technologists resigned. Where we once had federal support to slash red tape in Medicaid, the federal government’s policy goal is now to add more.

In the absence of the federal government, the only path forward is for states to grow their in-house technology capacity. The model of total reliance on incumbent vendors has proven too risky, too slow, too costly, and too error prone to meet the moment. States need engineers, designers, data scientists, and product managers inside of government who can build systems that work, manage vendors, and translate policy requirements into user experiences. States can hire mission-driven technology talent in-house, they can use federal matching funds to pay for it, and they should be removing any in-state hurdles that stand in their way.

The results published in Health Affairs confirm that this model works. Today, millions of people are able to see their doctor because it worked in a moment of crisis after the pandemic. Now states must build this capacity in-house to be ready for HR.1, and whatever comes next.

---

Thanks to the United States Digital Service team who drove around the country in a van for a year doing the work of a lifetime and made all of this possible including Alicia Rouault, Alyssa Kropp, Chris Wren, Emily Mann, Greg Novick, Heather Myers, Izzie Zahorian, Max Mazzocchi, Megan Cage, and Navin Eluthesen.

The author, Luke Farrell, is a fellow at the Better Government Lab and a technology and policy executive who has built and led mission-driven technology teams across the government, nonprofit, and private sectors. He served as Senior Advisor for Technology and Delivery on the White House Domestic Policy Council and as a Product Lead for the U.S. Digital Service.