|
Field
Experimental Design (Public)
|
Before
The experimental population will include people in Wisconsin who lost Medicaid coverage for procedural reasons during an 18 month period. Exclusion criteria include preferring a language other than English, Spanish, or Hmong, and lacking a working phone number.
Assignment to treatment arms will occur monthly. Randomization will be clustered by household. Households will be assigned to treatment arms in equal numbers, stratified by key characteristics that can affect eligibility and enrollment processes.
|
After
The experimental population will include people in Wisconsin who lost Medicaid coverage for procedural reasons during an 18 month period. Exclusion criteria include preferring a language other than English, Spanish, or Hmong, and lacking a working phone number.
Assignment to treatment arms will occur monthly. Randomization will be clustered by household. Randomization will be stratified by key characteristics that can affect eligibility and enrollment processes.
To maximize power to detect an effect among non-English speakers, which are a small minority of the sample, we will randomize households whose head preferred a language other than English to the three treatment arms in equal numbers. English speakers will be randomized to remaining live outbound call spots until the capacity cap is hit, and remaining English-speaking households will be assigned to the remaining two treatment arms (pre-recorded call and control group) in equal proportions. To reflect this randomization strategy, we will separately analyze data for English and non-English speakers.
|
|
Field
Sample size (or number of clusters) by treatment arms
|
Before
Live outbound call from a Navigator 36,250 households
Pre-recorded outbound call (45% of the sample) ~42,500 households
No-outreach control arm ~52,000 households
|
After
Live outbound call from a Navigator ~36,000 households
Pre-recorded outbound call ~47,000 households
No-outreach control arm ~47,000 households
|