Institutional Review Boards deemed this public program evaluation exempt from review.
Data Sources
Healthy Michigan Voices enrollee survey
We conducted a telephone survey of a longitudinal panel of Medicaid expansion enrollees, ages 19–64 with incomes up to 133% of the federal poverty level (FPL), effectively up to 138% FPL after application of the Medicaid 5% income disregard. Details of the Healthy Michigan Voices Survey, including sampling and design, have been described in previously published work [2,3,4]. The 2016 survey response rate was 53.7% (N = 4,090) and follow-up response rates among those who agreed to be recontacted were 83.4% in 2017 (N = 3,104) and 89.4% in 2018 (N = 2,608), with an overall response rate of 39.5% across the three survey waves [5]. The present study population included in analysis was only individuals who responded to all three waves of the survey in 2016, 2017 and 2018. While all individuals sampled in the 2016 survey were currently enrolled in HMP, individuals remained in our survey cohort whether or not they were enrolled in HMP in follow-up years, 2017 and 2018; thus, all 2017 and 2018 survey estimates reflect a combination of current and previous enrollees in the program.
Census data
To examine whether observed trends among HMP respondents reflected changes attributable to the program or statewide secular trends, we benchmarked HMP enrollee survey data against data from low-income adults ages 19–64 in Michigan overall. Data were obtained from serial cross-sectional surveys in the U.S. Census Bureau Current Population Survey (CPS) Annual Social and Economic Supplement from 2016, 2017 and 2018. Low income was defined as persons with household income up to 138% FPL, which approximates HMP income eligibility of 133% FPL after applying the Medicaid 5% income disregard. We also conducted sensitivity analyses using the U.S. Census Bureau’s American Community Survey (ACS), using the same definitions as noted for the main CPS analyses.
Measures
Enrollee survey data
Survey items measured employment (“Are you currently employed or self-employed?”, with response options yes/no) and student status (“Are you currently in school?”, with response options yes/no). For those not currently employed or in school full-time, a follow-up question asked, “Would you say you are: retired, unable to work, or out of work?”. Employed enrollees were asked their working hours (“Including overtime, about how many hours per week do you usually work on all paid jobs?”), while enrollees who were not employed were asked about job searching (“Are you currently looking for a job?”) or job training (“In the last 2 years, have you completed or are you currently enrolled in training, programs, or courses to increase your chances of getting or keeping a job?”). Respondents who reported inability to work were asked about reasons for being unable to work (“What are the reasons you are unable to work?”).
Health status was assessed in the enrollee survey by the items, “In general, would you say your health is…?” (with response options: excellent, very good, good, fair, or poor), “For how many days during the past 30 days was your physical health not good?”, “For how many days during the past 30 days was your mental health not good?”, and “During the past 30 days, for how many days did poor physical or mental health keep you from doing your usual activities, such as self-care, work, or recreation?”. Enrollee demographic characteristics were obtained from Medicaid enrollment files and survey items. Chronic conditions and mental health/substance use disorders were identified from survey and claims diagnosis data.
CPS data
Individuals who were civilians and reported any of the following were considered employed: work as paid employees, self-employed business/farm, worked at least 15 h as unpaid worker at business/farm owned by a family member, or had a job but were not currently working because of illness, weather, vacation, time off, or labor-management disputes. Individuals who reported they were enrolled in school full-time or part-time and had attended within the last week were considered students.
Data analysis
For HMP survey analyses, we used mixed models with individual random effects to assess changes in the proportion of HMP respondents who were employed or students by year, incorporating weight adjustments for sample design, initial nonresponse, and post-stratification adjustment for non-response to the follow-up. The mixed effects regression allowed us to examine longitudinal changes in the same sample of HMP respondents across years. For the benchmarking analyses of low-income Michigan adults ages 19–64 in the CPS data, we used survey-weighted logistic regression to assess changes in the proportion of Michigan residents who were employed or students by year.
The mixed effects regression model assumes a linear relationship between the natural log of the probability of the outcome (employed or student) and the predictor variable (year) and no outliers. With only three years of data we cannot definitively demonstrate a linear relationship, but there were no outliers. The mixed effects model was deemed appropriate to assess change over time in the repeated measures of employment or student status for each individual person, as we assume employment or student status in one time may be linked to employment or student status at another time.