Series 15: Human Dimensions of Work Requirements
Theory is useful. Templates are more useful. Behavioral science has generated decades of research demonstrating that good intentions do not automatically produce completed forms, that the gap between wanting to maintain coverage and actually maintaining it is not a motivation problem but a design problem, and that systems can be constructed to bridge this gap rather than widen it. Article 15C laid out the theoretical framework. This article translates those principles into concrete interventions that states, managed care organizations, and navigation organizations can deploy starting now.
The interventions that follow are not speculative. Each has been tested, refined, and in many cases scaled across populations far larger than Medicaid expansion enrollment. The question is not whether these approaches work. The question is whether work requirement systems will incorporate what we already know about helping people succeed.
The Cost of Not Trying
Before examining specific interventions, consider the baseline. Arkansas’s work requirements produced 18,000 coverage losses in ten months before court injunction halted the program. Subsequent research found that most who lost coverage were actually working or qualified for exemptions but failed to navigate the verification system. Georgia’s Pathways to Coverage enrolled only 5,500 of an estimated 240,000 eligible residents in its first eighteen months, with monthly reporting requirements creating monthly opportunities for administrative failure. These are not programs that tried behavioral approaches and found them wanting. These are programs that never tried.
The alternative is documented across domains. Oklahoma’s SoonerSelect program doubled plan selection rates from 10 to 20 percent using targeted SMS outreach, with enrollment jumping from 6,000 to 14,000 per day following text messages. Louisiana’s pilot program increased Medicaid renewal rates by 10 percentage points and SNAP renewals by 19 percentage points through text reminders. Michigan’s form redesign achieved a 96 percent completion rate compared to 73 percent for the legacy forms while reducing errors by 60 percent. The interventions exist. The evidence exists. The implementation gap remains.
SMS Reminder Strategies
Text messaging has become the most extensively tested behavioral intervention in benefits administration. The evidence base is substantial, the costs are minimal, and the outcomes are measurable. The Federal Communications Commission’s 2021 ruling cleared the path for state agencies to deploy automated texting campaigns without the cumbersome consent requirements that previously deterred adoption.
The effectiveness of SMS interventions depends less on whether messages are sent than on how they are constructed and timed. Research consistently shows that personal tone outperforms bureaucratic language. A message reading “Don’t lose your health coverage. Submit your work hours by Friday” generates meaningfully higher response rates than “Failure to comply with reporting requirements may result in coverage termination.” The information content is identical. The psychological activation is different.
Timing matters more than frequency. Messages sent too early become background noise, filed away with good intentions that never translate to action. Messages at deadline create panic without sufficient time for response. The optimal window varies by population and task complexity, but research consistently supports the pattern that the State Health Value Strategies consortium identified: reminder sequences at 30 days, 14 days, 7 days, and 48 hours before deadline, with escalating urgency signals that match the approaching consequence.
Two-way messaging enables immediate action. A text that says “Reply YES to confirm your work hours” and then asks three follow-up questions can complete verification in under a minute. Compare this to the alternative: logging into a portal, resetting a forgotten password, navigating to the correct form, entering information, uploading documentation, and confirming submission. Both accomplish the same administrative function. One takes sixty seconds from a phone that is already in someone’s hand. The other requires sustained attention, multiple steps, and cognitive resources that stressed populations may not have available.
GSA’s Notify.gov initiative with Norfolk, Virginia demonstrates another critical element: establishing credibility before asking for action. Many Medicaid enrollees report uncertainty about whether text messages are legitimate or scams. Norfolk’s intervention includes a preliminary “credibility text” informing enrollees that the city will text them from a consistent number about renewals and will never ask for personal information via text. This priming message increases trust in subsequent action-oriented messages. The intervention addresses a barrier that system designers often overlook because it does not occur to people who expect government communications to be legitimate.
The channel optimization research from recent Medicaid unwinding efforts reveals important demographic variation. Message framing emphasizing keeping coverage and benefits outperforms framing emphasizing the need to reapply, with a 4 percentage point lift compared to placebo. This holds across most subgroups. However, Latino respondents react negatively to loss-framed messages, while Republican-identifying respondents show stronger response to loss framing. Young adults respond most strongly to access-to-care framing. There is no single optimal message. There is optimal matching of message to population.
Form Redesign Principles
The form itself is an intervention. How questions are worded, how information is organized, how much cognitive load each page demands, whether progress is visible, whether errors are caught before submission or after, all of these design choices produce measurable differences in completion rates. The research from Civilla’s work with Michigan’s Department of Health and Human Services demonstrates what is possible: a 23 percentage point improvement in completion rates from form redesign alone.
Plain language is not optional. Research on Spanish-language Medicaid applications found that most state forms were written far above the recommended sixth-grade reading level, with document complexity requiring the equivalent of fifteen years of schooling. Applicants are not failing because they are unwilling to comply. They are failing because they cannot understand what is being asked. The Centers for Medicare and Medicaid Services has made plain language a priority in its own materials, with customer satisfaction scores increasing measurably following readability improvements.
Visual hierarchy guides attention to critical elements. When everything on a form appears equally important, nothing is important. Effective redesign uses size, color, and positioning to ensure that the most essential information, specifically deadlines, required actions, and consequences, receives cognitive priority. Progress indicators reduce abandonment by showing how much has been completed and how much remains. Error prevention through field validation catches problems before submission, when they can still be corrected, rather than after, when correction requires starting over.
Mobile-first design has become essential for populations that access the internet primarily through smartphones. The ideas42 analysis of benefits access notes that many people who struggle with desktop portals can complete the same tasks easily on mobile interfaces designed for their devices. This is not about accommodation for a minority. Ninety-seven percent of adults with income under $30,000 own a cell phone, and seventy-one percent own a smartphone. For many Medicaid expansion adults, the phone is not an alternative access point. It is the only access point.
Pre-population transforms the task from data entry to data verification. When a form arrives with name, address, employer, and prior work hours already filled in based on existing records, the member reviews and confirms rather than remembering, locating, and entering. This is not merely more convenient. It is categorically different in cognitive demand. Pre-population leverages the power of defaults, making confirmation the path of least resistance and requiring active effort only when information has changed.
Deadline Architecture
The psychology of deadline response is well documented and consistently ignored in benefits administration. Most action happens in the final 48 hours regardless of how long the runway is. A 45-day deadline does not produce 45 days of distributed effort. It produces 43 days of intention and 2 days of frantic activity, with predictable attrition among those who encounter any friction in that compressed window.
This pattern has implications for deadline structure. Shorter deadlines with more frequent touchpoints may outperform long deadlines that rely on prospective memory. A monthly verification with a 14-day window and reminders at days 10, 7, and 2 activates different psychological mechanisms than an annual verification with a 60-day window that most people will ignore until the final week. The former creates regular habits. The latter creates irregular crises.
Rolling deadlines distribute administrative burden more effectively than fixed deadlines that create system-wide peaks. When everyone’s deadline falls on the same date, portal crashes, call center overload, and processing backlogs compound individual compliance failures. When deadlines are distributed by birth month or enrollment date, the system can handle volume smoothly and provide meaningful support to those who need assistance.
Soft deadlines with cure periods fundamentally change the compliance dynamic. A system that terminates coverage immediately upon missed deadline treats administrative failure as conclusive evidence of non-compliance. A system that sends a warning, provides 15 additional days, offers navigation support, and terminates only after multiple failures treats administrative failure as a problem to be solved. Both systems can claim to enforce work requirements. Only one system can claim to be designed for human beings.
The color coding and urgency signals in deadline communications matter more than administrators typically recognize. The Behavioural Insights Team’s EAST framework emphasizes making messages attractive through attention-capturing design elements. A notice that looks like every other piece of mail gets treated like every other piece of mail, which is to say, set aside for later attention that may never arrive. A notice with urgent coloring, prominent deadline display, and clear consequence statement receives cognitive priority.
Navigator Nudge Deployment
Frontline workers can incorporate behavioral techniques without behavioral science training. The key is making the techniques procedural rather than theoretical. Navigators do not need to understand implementation intentions as a psychological construct. They need a script that says, “Let’s schedule exactly when you’ll complete this form. What day works? What time? Where will you be?”
Implementation intention protocols transform vague plans into specific commitments. Research consistently shows that completion rates approximately double when people specify the when, where, and how of intended action compared to when they simply state intent. A navigator who asks “Will you submit your verification?” gets an affirmative response that predicts little. A navigator who asks “When will you submit it? Where will you be when you do? What might get in the way?” creates a mental representation that activates when the specified conditions arise.
Commitment devices created during navigation encounters leverage social accountability. Having stated a specific plan to another person, the member experiences psychological pressure to follow through. This is not manipulation. It is recognition that human beings are social creatures who respond to interpersonal expectations. The navigator who asks “Should I check back with you on Thursday to make sure it went smoothly?” creates an accountability structure that increases completion likelihood.
Follow-up sequences using optimal timing require navigation organizations to build systems that track commitments and deliver appropriately timed outreach. If a member commits to submitting on Saturday morning, a Friday afternoon text saying “Ready for tomorrow?” and a Saturday midday text saying “Did you get it done?” provides scaffolding without surveillance. The technology is trivial. The organizational discipline to use it consistently is the real investment.
Social proof messaging normalizes compliance and establishes that the task is achievable. A navigator who says “I helped someone in a similar situation yesterday, and they completed it in 20 minutes” accomplishes several things simultaneously: establishing that people like the member successfully complete verification, providing a time estimate that makes the task feel manageable, and positioning the navigator as someone with relevant experience. Research from the Behavioural Insights Team shows that social norm statements can shift behavior by 5-15 percentage points depending on context.
Digital Intervention Design
Digital systems can incorporate behavioral principles throughout the user experience. The choice between push notifications and pull information determines whether the system reaches people or waits for people to reach it. Medicaid populations with competing demands will often not proactively seek information. Systems that push relevant information at appropriate moments intercept people before they fall into compliance gaps.
In-app guidance and completion assistance transforms portals from passive interfaces into active support systems. Rather than presenting a blank form and waiting for input, effective digital design walks users through each field, explains what is being asked and why, catches errors before they accumulate, and provides contextual help at the moment of confusion. This is standard practice in commercial software. It remains rare in government benefits systems.
Chatbot interventions have shown mixed results in health contexts, with effectiveness depending heavily on design quality and appropriate use cases. Simple, transactional interactions, such as checking deadline status, confirming submission receipt, or answering common questions, work well. Complex situations requiring judgment, such as determining exemption eligibility or navigating unusual circumstances, require human escalation. The technology works when deployed for tasks within its capability.
Auto-save and return functionality addresses the reality that people do not complete forms in single sessions. A system that loses all entered data when a phone call interrupts the session creates unnecessary restart friction. A system that saves progress automatically and enables return to the exact stopping point removes this barrier. The technical implementation is straightforward. The impact on completion rates is substantial.
Email subject line optimization matters because most emails are never opened. Research on benefits outreach shows that subject lines containing specific deadlines, dollar amounts, or clear action requests generate higher open rates than generic subject lines. “Your health coverage expires March 15” outperforms “Information about your Medicaid eligibility.” The message inside can be identical. The wrapper determines whether anyone reads it.
Cost-Effectiveness of Behavioral Interventions
The return on investment calculation for behavioral intervention versus enforcement investment is not close. The cost of sending text message reminders is measured in cents per member. The cost of processing coverage terminations and subsequent re-enrollments is measured in hundreds of dollars per case. The cost of emergency department utilization by formerly covered populations who deferred care due to coverage uncertainty is measured in thousands of dollars per event.
Consider the arithmetic. If SMS outreach costs $0.10 per member per message and a four-message reminder sequence reaches 100,000 members at $40,000 total cost, and if that intervention retains 3 percent of members who would otherwise have lost coverage, and if the administrative cost of processing termination and re-enrollment is $200 per case, the intervention prevents $600,000 in administrative costs for a $40,000 investment. This calculation excludes the healthcare costs of coverage gaps, the member burden of navigating re-enrollment, and the MCO revenue loss from member churn.
Form redesign requires upfront investment but generates sustained returns. Michigan’s investment in the Civilla partnership produced forms that will serve millions of interactions over years of use. The per-transaction cost of better design approaches zero as volume accumulates. The per-transaction cost of processing errors, return mail, incomplete submissions, and customer service calls for the legacy system accumulated indefinitely.
Navigation support has the highest per-member cost but produces the highest retention rates among high-complexity populations. For members with serious mental illness, substance use disorders, housing instability, or limited English proficiency, the return on navigation investment reflects not just retained enrollment but retained risk adjustment revenue. As Article 12C documents, the MCO economics strongly favor navigation investment for populations whose administrative complexity correlates with medical complexity.
The Gap Between Knowing and Doing
The interventions described in this article are not novel. Text message outreach has been standard practice in commercial marketing for decades. Form design research has been refining user experience principles since the 1990s. Implementation intentions have been studied since Gollwitzer’s foundational work in 1999. The fresh start effect was documented by Dai, Milkman, and Riis in 2014. The EAST framework has been public since its original publication in 2012 and was updated in 2024 to reflect a decade of additional evidence.
The gap between behavioral science research and benefits administration practice represents a choice, not a constraint. The same governments that struggle to retain Medicaid enrollment have successfully deployed behavioral interventions for tax compliance, retirement savings, and vaccination uptake. The same technology platforms that lose members to portal friction manage billions of commercial transactions daily without losing users to identical technical challenges. The tools work. The question is whether systems will use them.
Creating organizational capacity for behavioral intervention requires recognizing that design is not peripheral to program administration. It is program administration. The form is not a neutral vehicle for collecting information. It is an intervention that either supports completion or impedes it. The deadline is not a neutral boundary for distinguishing compliers from non-compliers. It is a psychological trigger that interacts with human cognitive limitations in predictable ways. The reminder is not an optional courtesy. It is infrastructure for bridging the intention-action gap that separates people who want to maintain coverage from people who actually maintain it.
A toolkit is only useful if used. The nudge infrastructure described here could be deployed by any state, MCO, or navigation organization with modest investment and reasonable timeline. The evidence base is established. The implementation path is clear. The remaining variable is whether systems designed to administer work requirements will also be designed to help people comply with them.
For 18.5 million Medicaid expansion adults facing work requirements beginning December 2026, the answer to that question will determine whether compliance statistics reflect actual work activity or administrative navigation ability, whether coverage loss concentrates among those who cannot prove what they are doing rather than those who are not doing it, and whether a policy intended to connect healthcare with labor market participation instead disconnects healthcare from the populations who need it most.
The tools exist. The evidence exists. The choice remains.