Data source and study population
We utilized data from the 2011–2019 Behavioral Risk Factor Surveillance System (BRFSS) survey, an ongoing, state-based, random-digit-dialed telephone survey of non-institutionalized U.S. adults (≥ 18 years old). BRFSS collects state data about health-related risk behaviors, chronic health conditions, and use of preventive services. The questionnaire consists of core questions asked in all 50 states, the District of Columbia, and U.S. territories. Data are weighted to reflect the age, sex, and racial/ethnic distribution of the state’s estimated population during each survey year. Data prior to 2011 was not included due to changes in weighting methodology and the addition of the cell phone sampling frame that occurred that year [17]. All participants that self-identified as being Hispanic or Latino ethnicity were included. Due to data limitations, we were unable to distinguish between Latinos who are undocumented, legal residents, and U.S. citizens.
Measuring perceived physical and mental health among Latinos
Perceived physical and mental health were captured using two items which asked, “for how many days during the past 30 days was your physical (mental) health not good?” Perceived general health status was assessed using the question “Would you say that in general your health is?” Response options included excellent (1), very good (2), good (3), fair (4), and poor (5). Scoring for this question was reversed for analyses, so that higher scores indicated better perceived general health.
Other sociodemographic characteristics included state of residence, age, gender, marital status, education, employment status, annual household income, interview language, health insurance status, inability to seek care because of cost, and amount of time since last routine physical exam checkup.
Identifying states that have immigrant-inclusive license policies between 2011 and 2019
Information on each state’s immigrant driver’s license policies were captured using reports from the National Conference of State Legislatures [8] and National Immigration Law Center [9]. Legislation was verified with corresponding state bills. States that have enacted immigrant-inclusive driver’s license polices (defined as state legislature that issues a license if an applicant provides certain documentation, such as foreign birth certificate, foreign passport, or consular card and evidence of current residency in the state) that allow undocumented immigrants to obtain licenses include: California, Colorado, Connecticut, Delaware, District of Columbia, Hawaii, Illinois, Maryland, New Mexico, Nevada, Oregon, Utah, Vermont, Virginia, and Washington. For our analyses, we excluded Latino participants living in Hawaii, New Mexico, Utah, and Washington because they had enacted their immigrant-inclusive license policies prior to 2011. Latino participants from Delaware were excluded due to small sample size (unweighted n = 2,072).
Because all license-expansion policies for the included intervention states were implemented between November 2013 and January 2015, time was stratified into three periods: pre-implementation (January 2011 – June 2013), during implementation (July 2013 – June 2015), and post-implementation (July 2015 – December 2019).
In order to remove the potential bias caused by the ACA and Medicaid expansion, we restricted our analysis to include only Latino participants living in states which opted into the expansion in 2014. This meant dropping Oregon and Virginia from the intervention state list, leaving our final group of included intervention states as: California, Colorado, Connecticut, Illinois, Maryland, Nevada, Vermont, and the District of Columbia (66,805 Latino adults).
We similarly restricted our control states (i.e., states that did not enact (and had not enacted previously) immigrant-inclusive license policies between 2011 and 2019) to those that participated in Medicaid expansion in 2014 as part of the ACA. Our final list of control states included: Arizona, Arkansas, Iowa, Kentucky, Massachusetts, Michigan, New Hampshire, New Jersey, New York, North Dakota, Ohio, Rhode Island, and West Virginia (57,002 Latino adults). Of note, New Jersey and New York enacted immigrant-inclusive license policies after the end of the study period (New Jersey: effective January 2021; New York: effective December 2019).
Statistical analyses
Average number of perceived poor physical health days per month, perceived poor mental health days per month, and average perceived general health score were estimated at three-month intervals (i.e., quarterly or four data points per year) between 2011 and 2019 using data from the BRFSS survey. Descriptive statistics were used to compare BRFSS participant characteristics across these three time periods, stratified by intervention and control group status.
In order to assess the immediate and gradual effects of enacting statewide immigrant-inclusive license policies on Latino health, we conducted an interrupted time-series analysis; this quasi-experimental approach is commonly used to assess well-defined population-level changes (e.g., new laws or policies) when randomization is not possible [18, 19] Using segmented linear regression, we estimated rates and linear trends before and after immigrant-inclusive policies were implemented in the intervention states [19]. We compared both the change in slope (gradual change) and intercept (immediate change) during the post-implementation period (July 2015 – December 2019) to pre-implementation time (January 2011 – June 2013). A similar analysis comparing these two time periods was performed in the control states; if similar changes were observed among states that did not enact license-expansion policies it would suggest that differences were due to other secular policies or trends that impacted Latino health.
Based on our hypotheses, we expected the average number of perceived poor physical and mental health days per month to decrease and perceived general health to improve in the post-intervention period in states where license policies were expanded. We also expected that perceived physical and mental health would remain relatively consistent among states that did not implement immigrant-inclusive license policies.
We also performed two sensitivity analyses. First, we dichotomized our outcomes and modeled the proportion of adults reporting having any (≥ 1 versus none) perceived poor physical or poor mental health days per month, as well as the proportion of those with perceived poor general health (poor/fair versus good/very good/excellent). For these dichotomized analyses, logistic regression was used. Second, we restricted the sample to Latino participants who experienced at least 1 (i.e., any) poor physical or mental health days, treating the number of perceived poor health days as continuous. The hypothesis for these analyses was that license expansion may not necessarily reduce the number of adults (objective of first sensitivity analysis) with perceived poor health, but that among adults who reported any poor health days, living in states with immigrant inclusive driver’s license policies would be associated with fewer perceived poor physical and mental health days per month (objective of second sensitivity analysis).
Descriptive statistics were estimated using SAS version 9.4 (SAS Inc., Cary, NC) and segmented linear regression was performed using SUDAAN release 11.0.3 (Research Triangle Institute International, Research Triangle Park, North Carolina). All analyses accounted for the complex survey design of BRFSS and were weighted to obtain national estimates. Variances in the regression models were computed using the Taylor Linearization Method, assuming a with-replacement design, in order to account for the complex survey weights.