Association Between Assistance With Medicaid Enrollment and Use of Health Care After Incarceration Among Adults With a History of Substance Use

Key Points Question Is prison-based Medicaid enrollment assistance associated with increased use of health care within 30 days of prison release among adults with a history of substance use? Findings In this cohort study of 16 307 individuals, the availability of prerelease Medicaid enrollment assistance was associated with large absolute increases in the likelihood of any outpatient visit and small or no absolute increases for substance use–associated and hospital-based care use. Meaning This study found that the addition of Medicaid enrollment assistance to discharge planning in correctional settings was associated with increased outpatient health care use for individuals with substance use disorders during the immediate reentry period.

This supplementary material has been provided by the authors to give readers additional information about their work.

eAppendix 2. History of Substance Use Definition
Identifying "highly probable" need for substance use treatment The underlying function of the COMPAS instrument is to assess risk of recidivism including potentially modifiable correlates of recidivism including substance use. 2,3 Available assessments of the validity of the COMPAS substance use score concern the degree to which this score is associated with recidivism rather than a clinical diagnosis of substance use disorder. 2 During our study period, the Wisconsin Department of Corrections (WI DOC) was adopting the COMPAS with the eventual goal of collecting two COMPAS assessments per person: one using the COMPAS Core instrument at intake; and one using the COMPAS Reentry instrument close to the time of release. During this implementation process, it was frequently the case that individuals completed just one assessmenteither Core or Reentrydepending on the time of administration. Thus, for each subject we obtained from the WI DOC the most recently completed COMPAS assessment relative to the individual's release date, and no more than 120 days after their release. An assessment may have a date after the release if it was conducted through the community supervision program.
There are some differences in the Core and Reentry instruments with respect to the substance use history questions although the WI DOC generates the same 3-category score indicating a need for treatment from each instrument: highly probable, probable, and unlikely. The specific questions on which this score is based for each instrument are noted below. We do not have access to the proprietary algorithm used to generate the score. However, in our internal analysis the vast majority of individuals identified as "highly probable" using the Core instrument had three or more positive responses to the substance use history questions. Using the Reentry instrument, the vast majority of individuals identified as highly probable had five or more positive response to the substance use history questions.
CORE Instrument Substance Use History Questions 1.Do you think your current/past legal problems are partly because of alcohol or drugs? 2.Were you using alcohol when arrested for your current offense? 3.Were you using drugs when arrested for your current offense? 4.Are you currently in formal treatment for alcohol or drugs such as counseling, outpatient, inpatient, residential? 5.Have you ever been in formal treatment for alcohol such as counseling, outpatient, inpatient, residential? 6.Have you ever been in formal treatment for drugs such as counseling, outpatient, inpatient, residential? 7.Do you think you would benefit from getting treatment for alcohol? 8.Do you think you would benefit from getting treatment for drugs? 9.Did you use heroin, cocaine, crack or methamphetamines as a juvenile? COMPAS Reentry Instrument Substance Use History Questions 1.Committed Offenses while high/drunk? 2.Prior drug charges/convictions? 3.History of drug problems? 4.History of alcohol problems? 5.Prior treatments for drug/alcohol abuse? 6.Any history of failed drug/urine analysis test? 7. Is the inmate at risk for substance abuse problems?

Identifying individuals with self-reported opioid use
The COMPAS Substance Abuse Module, which is not used in the above algorithm, asks individuals what substances they have a history of using. From the list of self-reported substances, the following are identified as opiates: buprenorphine, codeine, fentanyl, heroin, methadone, morphine, opiates, and Vicodin. Staff at the WI DOC, which include staff from WI DOC community corrections programs, are required to enter data into this assessment at intake. When an individual is reincarcerated or placed under community supervision, information may be added to this assessment. Thus, data on substances ever used likely became more complete over time (i.e. data are cumulative).
WI DOC staff members collecting the COMPAS data, record all substances the individual self-reports having used. Although the purpose of this data point is to understand problematic drug use (e.g., misuse of prescription drugs and illicit drug use), it is possible that respondents interpret the question differently and report opioids used for pain management. We believe this scenario is uncommon because DOC staff collecting the data are aware of the goal of the question. However, we cannot guarantee that people who used opioids strictly for pain management have been entirely excluded. We adopt the diagnosis and procedure codes published by the Medicaid Outcomes Distributed Research Network 4 to define visits for OUD and SUD, as well as medications for opioid use disorder. An outpatient visit is considered OUD-or SUD-related based on the presence of one of the relevant diagnoses shown below in any position on the claim.

eAppendix 4. Empirical Model Specification
Our main regression model takes the following general form: Yit= α0+ α1PHpara + α2PHnopara +α3Fullimp + Xit + Pit + εit Y is the outcome, i indexes a release from prison in month-year t, adjusts for a vector of Xit individual characteristics, a vector of Pit control variables that are specific to the prison-release including duration of incarceration, and type of release, and ϵ it represents a random error. Three policy variables characterize the enrollment assistance program; PHpara = 1 during the phase-in period (Jan 2015 -March 2015) for facilities in which a part-time benefits specialist is present; PHnopara =1 during the phase-in period (Jan 2015-March 2015 for facilities in which no part-time benefit specialist is present; and Fullimp =1 for all facilities during the after the program is fully implemented (>=April 2015). The coefficient of interest is α3. It reflects the average change in the outcome after implementation of enrollment assistance compared to the baseline period.
To obtain separate predictions for fully implemented enrollment assistance programs with and without the inclusion of a part-time benefits specialist, we modified the specification to take the following form: Yit= α0+ α1PHpara + α2PHnopara +α3Fullimppara + α4Fullimpnopara + Xit + Pit + εit Y is the outcome, i indexes a release from prison in month-year t, adjusts for a vector of Xit individual characteristics, a vector of Pit control variables that are specific to the prison-release including duration of incarceration, and type of release, and ϵ it represents a random error. Three policy variables characterize the enrollment assistance program; PHpara = 1 during the phase-in period (Jan 2015 -March 2015) for facilities in which a benefits specialist is present; PHnopara =1 during the phase-in period (Jan 2015-March 2015 for facilities in which no benefit specialist is present; and Fullimppara =1 during the full implementation period (>=April 2015) for all facilities in which a part-time benefits specialist was present; Fullimpnopara=1 during the full implementation period (>=2015) for all facilities in which no part-time benefits specialist was present.