The Most Health Coverage for the Most People: Medicaid Unwinding Data & Decisions

Monday morning. Root canal. Low balance. Hopefully, the sight of these phrases will not raise the cortisol levels of readers simply for the sake of a compelling opening. These three word pairings know no borders—almost any human recoils at their mention. One last pair of words may evoke an even stronger reaction but only for individuals in the US: Medical debt.

Indeed, here is a Google search trend map for the two words—the more shaded the country, the more common the search term.

Benchmarked to the search term's dominant US traffic for the past five years, no country's volume of searches for “medical debt” is even one-fifth as high as that of the US.

A report from the US Consumer Financial Protection Bureau (CFPB) showed that from 2018 to 2022, medical debt constituted the majority of accounts in collections from a sample of 5,000,000 citizens' credit reports. But that only accounted for medical debt that has gone into collections and reported by credit rating agencies. A February 2024 report published by the Peterson-KFF Health System Tracker estimated the amount of Americans' collective medical debt to be about $220 billion, spanning more than 20 million Americans.

Who Are the Debtors?

A September 2022 investigation published in the Journal of the American Medical Association looked at numerous cross-sections of the population holding medical debt. The debtors' health insurance coverage corresponded closely with whether they had medical debt—the plurality of medical debtors have no health insurance, followed by those with high-deductible private plans.

The authors also conducted a prospective study of 51,872 adults between 2017 and 2019. Compared with the full sampled population, among those who at the end of 2019 had incurred medical debt, which they did not have at the beginning of the study, it was 1.63 times more common for them to have also lost their health insurance coverage for any time during these three years (95% CI, 1.23-2.14).

The February 2024 medical debt report from the Peterson-KFF Health System Tracker found that adults who were uninsured for just part of 2021 were more likely to report having medical debt (14%), than those who were insured for the full year (8%) or uninsured for the full year (11%).

How Does Medical Debt Relate to Health Outcomes?

Findings from a 2022 KFF poll suggest that people with unaffordable medical bills were more likely to delay or skip needed care to avoid incurring more medical debt, cut back on other basic household expenses, take money out of retirement or college savings or increase credit card debt.

The February 2024 medical debt report from the Peterson-KFF Health System Tracker found that those who acquired new medical debts between 2017 and 2019 had consistently higher odds of worsening social determinants of health (SDOH):

• 2.2 times more likely to become food insecure (95% CI, 1.58-3.05)
• 2.29 times more likely to become unable to pay rent or mortgage (95% CI, 1.38-6.31)
• 2.95 times more likely (95% CI, 1.38-6.31) to face eviction or foreclosure in 2019

It also found that adults living with a disability were more likely than those without a disability to report owing medical debt (13% vs. 6%), and that adults who reported their health status as “fair” or “poor” were more likely to say that they owe medical debt than those who said they were in “good” or “better” health.

Concerns over medical debt understandably influence how and if people choose to access medical care. In March 2024, KFF published updates to data originally collected in 2022 from a nationally representative survey of 2,375 adults. They found that one in four adults had skipped or postponed getting needed health care in the past 12 months due to cost. Half of the respondents did not have the means to pay an unexpected $500 medical bill without going into debt.

Among the uninsured respondents, six in ten reported skipping or postponing a needed healthcare service due to cost—three times the rate of insured respondents.

Medicaid enrollees were less likely than those with other coverage types to give their insurance negative ratings on these affordability measures.

Plugging Holes in the Safety Net

These recent studies support the notion that there is a correlation between the level of an American adult's health insurance coverage and the likelihood that they have some medical debt. This is a critical consideration for Medicaid agencies and policymakers, given that those with medical debt are often subject to pervasive, often catastrophic, consequences.

Efforts prioritizing the Medicaid beneficiaries' retention of their health insurance coverage are further warranted based on the public health data from states that provide more pathways through which people can be determined eligible for Medicaid.

The Affordable Care Act expanded Medicaid coverage to all Americans meeting household income thresholds alone, as opposed to the combination of income with other categorical qualifications (e.g. disability, pregnancy, etc.). In 2012, the US Supreme Court ruled that states could not be forced to expand their Medicaid programs, so it was up to each state to determine whether or not to participate. As of this writing, ten states have yet to adopt any degree of Medicaid expansion.

Residents of states that have refused the ACA Medicaid expansion were 40% more likely to have medical debts than those in Medicaid-expansion states. Further studies suggest a correlation between the mortality rate of individuals in a given state and whether that state has expanded Medicaid.

There is strong evidence that initiatives aimed at simplifying workstreams through which individuals can easily compare, obtain, retain and understand their health coverage can produce demonstrable benefits to public welfare.

Given this context, a long-foreseen regulatory change combined with the technological shortcomings of state Medicaid information systems resulted in millions of Medicaid beneficiaries losing their coverage based on eligibility decisions evaluated with incomplete data.

Prove This, Prove That, Prove This Again…

Prior to the COVID-19 pandemic, states were required to redetermine Medicaid beneficiaries' eligibility for coverage every twelve months. In response to the health emergency, the requirement of redetermining eligibility at this cadence was dropped—the federal government provided additional funding to cover the costs associated with providing continuous coverage to Medicaid recipients. On April 1, 2023, this provision ended—states could begin to shift back to annual eligibility redeterminations and evaluate the eligibility of all existing enrollees—a process referred to as “Medicaid unwinding.”

In the three years of the continuous coverage provision, the total number of Medicaid beneficiaries rose from approximately 23.3 million to nearly 95 million, according to KFF. This was attributable to the provision but also to Medicaid expansion going into effect in Nebraska, Missouri and Oklahoma, as well as the economic hardships brought on by the pandemic.

The period of unwinding, which will likely last through 2024, is expected to result in a net decrease of between 7.8 million to 24.4 million Medicaid enrollees. The imprecise timeline and process, different for every state, has reintroduced and even exacerbated challenges that already existed in the eligibility determination process pre-2020.

New Challenges Related to Awareness

Recent KFF surveys suggest that nearly one in three disenrolled adults discovered only when they sought health care—such as going to a doctor or a pharmacy—that they had been dropped from Medicaid. As much as 71% of those surveyed who had received a reenrollment notice took action while 29% did not recall receiving renewal information.

The challenges that came with restarting the enrollment process as it existed pre-COVID had a multiplier effect upon issues that predated COVID. These issues can be generally grouped into two buckets: issues of data and issues of how decisions are made from that data. They surface through an analysis of the proportion of those whose coverage was dropped for procedural reasons. In other words, they were not found to be definitively ineligible, but for reasons of data completeness or mistakes in the business logic, they were not found to be eligible either.

KFF surveys show that almost half of those who lost their government coverage re-enrolled just weeks or months later. This is a red flag, as it would be anomalous for large numbers of Medicaid beneficiaries to be definitively ineligible for coverage for a few weeks or months and then become eligible again. It is more likely that they should never have been dropped in the first place. Moreover, seven in ten adults who were disenrolled during the unwinding went without any health insurance, at least temporarily, rendering them vulnerable to amassing medical debt.

Ex Parte Renewals

One provision of the Affordable Care Act is to require states to first attempt to redetermine eligibility based on reliable data sources to make a determination without requiring any action or documentation from the beneficiary, a process known as an ex parte renewal. Only if there is insufficient information, the provision says, should a state request anything from the beneficiary.

In a 2019 webinar enumerating best practices for timely and accurate Medicaid coverage eligibility determinations, three of the steps identified as critical were:

1. Define and apply a logic to how and when data sources are called to manage duplication while maintaining accuracy and integrity of automated electronic verification.
2. Combine the use of federal and state data sources to enhance a state’s ability to efficiently verify applicant information electronically and to identify data inconsistencies that require resolution.
3. Automate systems’ rules engines, including a link to the master client index, when determining Medicaid and CHIP eligibility.

Expanding upon these points, the Centers for Medicare and Medicaid Services (CMS) published a presentation in October 2022 on strategies to maximize the proportion of renewals conducted on an ex parte basis, encouraging states to:

1. “Implement a strategic data hierarchy to support consistent and efficient application of data.”
2. “Review business rules, logic and operational procedures (e.g., using process mapping) to identify opportunities to expand verification data strategies and increase ex parte rates.”

Better Data

Individuals may be found eligible for Medicaid on a wide range of bases, varying from state to state. Common criteria include how a household's gross income compares to the federal poverty level for their state. State-specific criteria may capture whether the individual is part of a cancer treatment program or was formerly in foster care. Further, depending on the means by which an individual is found eligible, the level of coverage may vary—for example, some states offer an eligibility pathway called the Katie Beckett Waiver, which covers the costs of specific types of care for disabled children.

Depending on the basis of eligibility, the data that a Medicaid agency will need to determine eligibility ex parte are disparate and diverse. Data sources with overlapping data like street addresses must be ranked hierarchically (e.g. which is more reliable, the address stored in the US Postal Service change of address database, or that of the Department of Motor Vehicles?).

Each of these disparate data sources elucidates different portions of a person's life circumstances that may render them eligible for Medicaid coverage. The more fully agencies can incorporate, unify and curate this data to maximize its integrity, the more easily they can identify the eligibility pathway that provides the most beneficial coverage.

Better Decisions

The Medicaid unwinding process has served as a de facto report card on states' success in the accurate automation of eligibility determination processes. While states' automation workflows showed promise, audits of their eligibility decisions surfaced major problems in how the eligibility business rules were being implemented.

In August 2023, the CMS issued a letter to state Medicaid agencies enumerating concerns about the rules by which eligibility was being redetermined. Thirty states attested to having conducted ex parte reviews incorrectly—evaluating eligibility at the household level, not the individual level. Because each person in a household may be eligible for different programs, through different eligibility pathways and often by meeting different thresholds by demographic, a household-level determination is far from sufficient.

In other cases, many states have in frequent months been ineffective in their implementation of eligibility determination processes that evaluate all bases of eligibility before rendering the decision that someone is ineligible. A March 2024 CMS communique emphasized that states must immediately reinstate eligibility until a final determination has been made and the individual has been determined ineligible.

The advisory also instructed states that any ex parte determination resulting in a beneficiary being found eligible for Medicaid, but for less comprehensive coverage than they previously held, must offer beneficiaries the opportunity to provide additional documentation for their eligibility for the better coverage option.

Where Did You Learn That From?

The challenges not yet overcome by state Medicaid agencies in reverting to pre-pandemic eligibility determination workflows have, for better or worse, produced data that may be the basis of future eligibility decisions. One eligibility pathway mentioned earlier in this article is specifically for former foster children. Some states cover adults between 18 and 26 who were in a foster care program as children and still reside in the same state or in a different state where this eligibility pathway is available, as long as they were on Medicaid when they turned 18.

Convoluted? Certainly, but that comes with the territory of means-tested government assistance programs that differ by state. The bigger problem with such pathways is how individuals are determined eligible or not. An inescapable lesson from the period of the Medicaid unwinding is that humans are fallible, as are the automations we create. Mistakes in this sector are exceptionally consequential. Medicaid agencies can't easily remediate the problem of medical debt that results from periods in which Americans assumed they were covered by Medicaid but were not.

Consider the likelihood that there is a cohort of people who meet the following classifications:

1. They were in a foster care program at the time of their 18th birthday
2. They turned/will turn 18 in 2023 or 2024
3. They were on Medicaid at the time of their 18th birthday

Based on data from the Adoption and Foster Care Analysis and Reporting System (AFCARS), approximately 20,000 to 40,000 individuals fall into this cross section.

Their enrollment in Medicaid upon turning 18 is a necessary criterion for these individuals to be determined eligible in subsequent years, so the accuracy of the eligibility redetermination made during the Medicaid unwinding period is extremely consequential. If we follow this cohort for several years into the future to when they are between the ages of 18 and 26, there is likely a cross section of this population who will face hurdles obtaining coverage as former foster children due to past eligibility decisions which were invalid through no fault of their own.

The difference between a Medicaid agency being able to rapidly remediate this type of mistaken eligibility decision is contingent on the extent to which they can trace back why past eligibility decisions were made. What rules were evaluated in determining eligibility? What data was evaluated in reaching the decision? Where did that data originate from? To mitigate common issues with major government IT systems, it is essential that complex logic be explainable and available for review by a variety of stakeholders.

Given the grave human consequences of wrongly evaluated decisions, developing automations informed by eligibility determination data is rife with potential for rejecting coverage improperly. While the data is still accruing from the Medicaid unwinding, research reports suggest substantial numbers of beneficiaries have been incorrectly disenrolled.

Thus, beyond needing to be able to trace back the logic that led to an eligibility determination at a given point in time, if the decision that had been made at that time did not adhere to CMS eligibility determination guidelines, then the pool of Medicaid beneficiary data maintained by the state agency contained incorrect data. As a result, if this data were to be leveraged as training data for machine learning technologies, it can be expected that the insights produced may recreate issues in eligibility logic that would be disastrous to the lives of individuals in future years. And the more this logic is encoded in algorithmic black boxes of AI and ML platforms, the harder it will become to remediate.

Data and Decisioning Platform

Informed by the challenges faced and mistakes made throughout the Medicaid unwinding, the Progress teams focused on the Progress MarkLogic and Progress Corticon product lines have collaborated to untap the synergistic benefits of each tool.

The Progress MarkLogic Data Hub for Medicaid enables the finding, combining and searching of data, giving Medicaid agencies a comprehensive, consolidated view of member, provider and claims information—regardless of how many silos contain this data. Corticon provides a platform for business users to author, test and expansively document decision automations as business rules versus code. Once decision models are complete, they can be generated into a decision service that can run directly in MarkLogic as server-side JavaScript, providing high-performance and simplified deployments.

This solution enables all business rule definitions to be decoupled from the data and other modules, yet operationalized in the same place where the data is stored and curated. Decision services, when executed, can be configured to send verbose “—data documenting the nature and sequence of all changes made to the input data over the course of rule evaluation. This trace data, along with the actual output of the decision, can be persisted right back into the member record.

This cohesion of rules and data simplifies the task of incorporating eligibility determination data back into the master record of the beneficiary. The CMS has in recent months finalized new rules aimed at streamlining the process of transitioning children from Medicaid to the Children’s Health Insurance Program (CHIP) when their household income rises above the threshold set for Medicaid coverage.

One criterion for being determined eligible for CHIP is for the individual to have already been found ineligible for Medicaid coverage. This means that data must be evaluated and then reevaluated to adhere to CMS guidelines for ex parte renewals so individuals can receive the best coverage, not just any coverage and to check for other coverage options like CHIP if individuals are not eligible by any pathway for Medicaid.

There's no getting around the complexity of eligibility rules in the realm of Medicaid, and implementers cannot safely build systems with the assumption that all the components integrated into an eligibility system are error-free. With the Progress Data and Decisioning solution for Medicaid, though, implementers and all other stakeholders are given the tools to get the fullest picture of an individual’s data to evaluate all potential eligibility pathways and to understand, explain and document how eligibility decisions are made. By maximizing the number of people who can understand the data and decisions made from that data, stakeholders can be confident that the system works as intended and that issues can be quickly troubleshot when they arise.

Data and decisioning challenges are pervasive across industries, and building solutions for them is a core focus for our customers and us. Whether you're in the Medicaid enterprise systems domain or a technologist focused on improving data integrity and decision automation precision, we would love to hear from you. You can engage with the Progress team by reaching out to us.