Outreach and Maintenance of Medicaid Enrollment: Evidence from Wisconsin’s Navigator Program

This field experiment will test methods to increase maintenance of Medicaid enrollment by connecting beneficiaries with navigators, a group of professionals publicly funded since 2014 to help consumers enroll in coverage. The intervention will be implemented by Covering Wisconsin, a navigator grantee with a staff of more than a dozen professional assisters. Covering Wisconsin was contracted by the Wisconsin Department of Health Services (DHS) to conduct outreach to fee-for-service Medicaid beneficiaries after the end of the public health emergency.
The intervention arms will vary the modality of outreach (postcard, text message, and/or outbound call), content of the outreach (encouraging a call to a hotline vs. encouraging a text message that connects them with a chatbot) offered, and the number of outreach messages (1 vs. 2). These interventions will be repeated after the renewal window closes for people who lost their coverage. All beneficiaries will also be mailed standard letters by DHS.

2023-05-15

2024-06-10

The primary outcome is maintenance of Medicaid enrollment, measured on the individual level over the 12 months following each person’s redetermination deadline (measured at 1, 3, 6, 9, and 12 months following each person’s redetermination deadline).

The first set of secondary outcomes are proxies for whether the outreach was received. These measures include vendor data on text bouncebacks, invalid addresses, and feedback from an address and phone number validation service on whether the intended recipient was no longer reachable at the address or phone number provided by DHS. We will measure whether individuals who were sent a text message replied STOP, or DONE, the latter being a self-reported indicator of having taken action to renew their coverage. We will also measure whether recipients sent any reply message to the text, as a measure of engagement with the intervention.
The second set of secondary outcomes focus on application activities and outcomes, measured using administrative data. We will use the following categories: no application on record; application started but not filed; application filed but denied due to lack of eligibility; application filed but denied due to administrative reasons; application accepted. These will be measured 15, 30, and 45 days after the information was sent to vendors, and 15, 30, and 45 days after the vendor sent the messages to recipients.
Exploratory outcomes include SNAP enrollment, employment, and logged quarterly wages (+$1, to avoid taking the log of zero for those who are unemployed). These outcomes will also be measured at 1, 3, 6, 9, and 12 months after each person’s redetermination deadline.
When analyzing data from the second round of outreach (i.e., for people who had already lost their coverage), outcomes will be measured at 1, 3, 6, 9, and 12 months after the outreach messages were sent.

The study population includes all fee-for-service Medicaid beneficiaries in Wisconsin, except those who prefer a language other than English or Spanish, lack any phone number, or are missing a mailing address.

Assignment to treatment arms will occur monthly over the 12-month period after the end of the continuous coverage provision, to ensure each beneficiary receives outreach at the beginning of their two-month redetermination window assigned by the state. Each month, Wisconsin DHS will provide Covering Wisconsin a list of the beneficiaries whose redetermination window begins. Randomization for people in each group of monthly renewals will be clustered by case (e.g., members of a household who applied for Medicaid together) to address potential spillovers.

40% of cases will be assigned to be sent a postcard, and the remaining 60% of cases will be assigned to the text message arms, with 30% receiving a message with a text message call to action and 30% receiving a message with a hotline call to action. In addition, half of the full sample will receive an outbound call.

There will be a second round of randomization to determine outreach strategy for people who lost their Medicaid coverage during the redetermination process. As in the first round of randomization, cases will be assigned to receive outreach messages that vary the modality of outreach (postcard, text message, or outbound call), content of the outreach (encouraging a call to a hotline vs. encouraging a text message that connects them with a chatbot) offered, and the number of outreach messages (1 vs. 2).

Randomization will be stratified by the following factors: the preferred language of the primary person on the case, key Medicaid eligibility categories particularly relevant to ease of renewal, defined at the case level at baseline (being aged or disabled, being part of the “transitional medical assistance” population, or having eligibility due to pregnancy), whether there are enrolled children in the case at baseline, whether there are enrolled people over age 50 in the case at baseline, rural/urban residency of the primary person on the case, Black race and tribal membership of the primary person on the case.

Randomization will be done in office using a computer.

Randomization for people in each group of monthly renewals will be clustered by case (e.g., members of a household who applied for Medicaid together) to address potential spillovers.

Yes

168,000

224,000

33,600 clusters in Arms A and B, and 12,600 clusters in arms C through J

The study has 80% power to detect a 0.69 percentage point change in enrollment rates due to changes in outreach repetition or suggested modality for contact.

Pre-analysis plan V13 05152023.docx