Support for evidence-informed opioid policies and interventions: The role of racial attitudes, political affiliation, and opioid stigma.

Political affiliation, racial attitudes, and opioid stigma influence public support for public health responses to address opioid use disorders (OUD). Prior studies suggest public perceptions of the opioid epidemic are less racialized and less politically polarized than were public perceptions of the crack cocaine epidemic. Analyzing a cross-sectional, nationally representative sample (n = 1161 U.S. adults) from the October 2020 AmeriSpeak survey, we explored how political affiliation, racial attitudes (as captured in the Color-Blind Racial Attitudes Scale [CoBRAS]), and OUD stigma were associated with respondents' expressed views regarding four critical domains. Respondents with unfavorable attitudes towards Black Americans were less likely to support expanding Medicaid funding, increasing government spending to provide services for people living with OUD, and distributing naloxone for overdose prevention. Democratic Party affiliation was associated with greater support for all three of the above measures, and increased support for mandatory treatment, which may be seen as a substitute for more punitive interventions. Black respondents were also less likely to support expanding Medicaid funding, increasing government spending to provide services for people living with OUD, and of distributing naloxone. Our finding suggest that negative attitudes towards African-Americans and political differences remain important factors of public opinion on responding to the OUD epidemic, even after controlling for opioid stigma. Our findings also suggest that culturally-competent dialogue within politically conservative and Black communities may be important to engage public support for evidence-informed treatment and prevention.

Introduction

Public support is an important political determinant of which policies can be enacted to address opioid use disorders (Saloner et al., 2018; Wakeman and Rich, 2018; Adams et al., 2021; Kennedy-Hendricks et al., 2017; Tsai et al., 2019; McGinty and Barry, 2020; Barry et al., 2014; Perry et al., 2020). Support for such public health responses is influenced by many factors; the impact of stigma towards opioid users have been well-established (Kennedy-Hendricks et al., 2017; Perry et al., 2020; Ezell et al., 2018). Race and political affiliation may also impact public opinion and thus support for evidence-informed policies and interventions, especially when use of specific substances is identified with particular socio-demographic groups.

A large literature documents the racialized nature of drug use epidemics, most strikingly the crack cocaine epidemic of the late 1980s and early 1990s, whose epidemiology was relatively concentrated in nonwhite populations (Alexander et al., 2012; Vega et al., 1993). In contrast, many observers have noted that the current opioid epidemic differs in in media imagery, policymaker, and public responses, and the racialized depictions of substance users. Netherland and Hansen performed a content analysis of 100 articles produced between 2001 and 2011 related to opioid use. Their analysis revealed “a consistent contrast between criminalized urban black and Latino heroin injectors with sympathetic portrayals of suburban white prescription opioid users” (Netherland and Hansen, 2016). Shachar and colleagues provide the most comprehensive recent review, analyzing media keywords and framing from the 2016–17 prescription opioid epidemic to those observed in 1988–89 for crack cocaine and methamphetamine, and with heroin media stories across the same three-decade period; they find that the response to the crack cocaine epidemic was more punitive, while the response to the current opioid epidemic was more explicitly medical – and that this difference could be related to race (Shachar et al., 2020). Public perceptions of the opioid epidemic as arising from pharmaceutical industry misconduct, and as disproportionately affecting non-Hispanic whites appears to have promoted a more empathetic, less-punitive response to individuals with opioid use disorders (Cohen and Jangro, 2015; Mendoza et al., 2016; Mendoza et al., 2019; Glanton, 2021). In contrast, published surveys, based on 2014 data, suggested that most Americans believe that opioid use disorders affect all races (80%), or that such disorders are especially common among whites (Kennedy-Hendricks et al., 2017). Therefore both individual race/ethnicity and racial views may be important factors.

Other epidemics, including crack cocaine and HIV, have seen political affiliation emerge as an important predictor of public health response, with Republican leaders often opposing harm reduction and supporting stringent criminal sanctions against drug users and drug sellers (Massing, 2000). Opposition to needle exchange and other harm reduction strategies continue today, often divided along partisan lines (Legan, 2021; Goldberg, 2021). Prior studies also indicate that both Democratic and Republican legislators are more likely to deploy medical rather than criminal justice language in describing opioid use disorders, a pattern which study authors suggest may once-again highlight the role of race in support for a public health response (Kim et al., 2020; McGinty et al., 2016; Dvorak, 1999).

In summary, previous epidemics found stigma towards and racial perceptions of those affected by the epidemic, as well as political affiliation, have all been associated with the public’s response to the epidemic – specifically, whether to favor a punitive, criminal-justice-oriented response or a compassionate, medical treatment response. However, the explicit role of racism, defined here specifically as unfavorable attitudes towards African-Americans, may also be an important predictor of support for public health responses. We used nationally-representative data from the October 2020 AmeriSpeak® survey to explore these questions. In particular, we examined how racist attitudes, political affiliation, OUD stigma, as well as individual race and ethnicity are associated with support for public health responses to the opioid epidemic.

Methods

We analyzed a cross-sectional random sample of 1161 U.S. participants who completed the survey drawn from AmeriSpeak®, a probability-based ongoing panel of over 35,000 households designed to be representative of the U.S. household population, from October 22–26, 2020. The study was approved by the data collection organization’s Institutional Review Board. For AmeriSpeak®, US households are selected and sampled using area probability and address-based sampling, with a known, nonzero probability of selection from the NORC at the University of Chicago (NORC) National Sampling Frame.

The AmeriSpeak panel provides sample coverage of approximately 97% of U.S. households, and leads to a sample comparable to the US Census American Community Survey (ACS) sample. AmeriSpeak contacts sampled households by U.S. mail, telephone, and field interviewers (face-to-face) to improve coverage by capturing harder-to-reach cases, and has an annual panel retention rate exceeding 80%. (Technical overview of the AMERISPEAK panel)

Informed consent for future surveys is obtained during panel recruitment and emails/texts were sent to a randomly-selected group of panelists describing the study and inviting them to participate in the survey. The survey was offered in English and Spanish. Participants who did not respond to the initial invitation were contacted multiple times by email, text, and phone. Participants received a small incentive ($4) for responding to the short survey. Of the 4358 individuals contacted, 1095 (25.12%) fully completed all survey items. An additional 66 participants completed surveys with item non-responses, bringing the sample size to n = 1161 for some analyses).

The survey consisted of fifty items. The opioid-pertinent subset of items (described below) took an average of fifteen minutes to complete. Of note, this data was collected during the COVID-19 pandemic, when there were increased OUD overdoses and limited resources; the full survey is available in Appendix B (Macmadu et al., 2021; Products, 2021; KFF, 2021).

We use several measures to quantify respondents’ personal and family exposures to the opioids and the criminal justice system, expressed opioid stigma, political affiliation, expressed support for public policies that address opioid use disorders, and expressed attitudes regarding racial inequality.

Opioid policy scale

Policy attitudes were assessed with four items that explored support for public health responses to the opioid epidemic. For the current analysis, respondents were queried using a five-point Likert scale (“Strongly disagree,” “Somewhat disagree,” “Neither agree nor disagree,” “Somewhat agree,” or “Strongly agree”) with the following statements:

“I favor expanding Medicaid insurance benefits for low-income families to provide coverage for treatment of opioid use disorders.”

“I favor increasing government spending to improve treatment of opioid use disorder.”

“I believe that making drug treatment mandatory is an effective way to help people with an opioid use disorder.”

“I favor making naloxone (also known as ‘Narcan’), a medication that can quickly reverse the effects of a person experiencing an opioid overdose, widely available and affordable without a prescription.”

Social stigma towards people with an OUD

We developed a 10-item scale (Cronbach’s α = 0.84) adapted from prior stigma survey research (Kennedy-Hendricks et al., 2017; Yang et al., 2019). Questions asked about willingness to have a person with a past history of OUD work with you or marry into your family and willingness to have a person with a current OUD work with you, marry into your family, their perceived dangerousness, and perceived trustworthiness. Also, four items covered persons currently living with OUD or who experienced past history of OUD and their likelihood of stealing or to be a high-risk employee. Respondents rated their agreement with each statement on a five-point Likert-type scale (1 = strongly disagree, 2 = somewhat disagree, 3 = neither disagree nor agree, 4 = somewhat agree, and 5 = strongly agree). A higher score reflects greater stigma towards individuals with an OUD.

Color-blind racial attitudes scale

Race-conservative attitudes regarding Black Americans was measured using a subscale of the Color-Blind Racial Attitudes Scale (CoBRAS), which has been shown to be associated with higher levels of racial prejudice (Neville et al., 2006). The CoBRAS includes 8 survey items rated on a 5-point scale from “Strongly disagree” to “Strongly agree” and included items such as “White people in the U.S. have certain advantages because of the color of their skin”, and “Racial and ethnic minorities do not have the same opportunities as white people in the U. S.” Items were summed to create the scale, with higher scores representing lower perceived awareness of white racial privilege. Cronbach’s alpha was 0.88.

Political party affiliation

Current political affiliation was collected as a categorical variable: Democrat, leaning Democratic, no declared affiliation/independent, leaning Republican, or Republican.

History of opioid misuse

To measure the respondent’s personal experience with opioid misuse, we asked respondents “Have you ever misused opioids of any kind – such as heroin, fentanyl, or prescription pain medications other than exactly as prescribed for you?”

Similar questions were asked if they had family members or close friends who ever misused opioids in their lifetime. Opioid misuse was defined for the respondent as use of opioids or prescription pain medication illicitly obtained or used in a way not prescribed by a doctor.

Experience with criminal justice system

We asked respondents whether they themselves and whether a family member or close friend ever had a conviction for a misdemeanor or felony crime or been incarcerated in jail or prison.

Background factors

Data were collected on the sociodemographic characteristics of the respondents from the AmeriSpeak panel which updates these items annually, including age, sex, race/ethnicity, education, income and place of residence based on the US Census region (Northeast, South, Midwest, and West).

Analysis

All descriptive statistics are weighted to national census benchmarks, taking into account selection probabilities (balanced by sex, age, education, race/ethnicity, and region) (Saloner et al., 2018) and nonresponse (using a response propensity approach calculating the conditional probability that a particular respondent completed the survey given observed covariates) (Wiley, 2021).

Logistic regression was used to examine associations between racial attitudes and OUD stigma and the government policy and opioid treatment policy support outcomes controlling for potential confounders (age, region, gender, education, personal CJI, family CJI, personal OUD, family OUD; see Table 2&3), chosen a priori. Each of the 4 outcomes were analyzed using separate models, including CoBRAS score, political affiliation, and OUD stigma as predictors. Predicted support and 95% CI for each model were calculated, using the mean for all other predictors, and are presented in Fig. 1. All data were analyzed using IMB SPSS 24.1 and R software 4.1.0.

Table 2

Logistic regression beta estimates (95% CI) of government policy support, among AmeriSpeak participants, October 2020.

	ExpandMedicaidfor low-incomefamilies tocoveraddictiontreatment	Increasegovernmentspending toimprovetreatment ofOUD	Naloxoneavailable &affordablewithoutprescription	Mandatorytreatment isan effectiveway to helppeople withOUD
COBRS	−0.56 (−0.73, −0.39)	−0.51 (−0.68, −0.38)	−0.39 (−0.56, −0.23)	−0.01 (−0.16, 0.17)
OUD stigma score	−0.55 (−0.77, −0.378)	−0.50 (−0.71, −0.30)	−0.27 (−0.47, −0.07)	−0.02 (−0.20, 0.17)
Affiliation
Democrat	1.30 (0.87, 1.73)	0.93 (0.53, 1.34)	0.88 (0.47, 1.29)	0.57 (0.19, 0.96)
Lean democrat	0.78 (0.24, 1.33)	0.88 (0.36, 1.42)	0.04 (−0.47, 0.55)	0.27 (−0.21, 0.74)
Independent/no declared affiliation	0.19 (−0.25, 0.62)	−0.06 (−0.49, 0.37)	0.24 (−0.18, 0.67)	−0.08 (−0.49, 0.32)
Lean republican	0.31 (−0.16, 0.78)	−0.09 (−0.57, 0.38)	0.11 (−0.35, 0.57)	0.01 (−0.46, 0.43)
Republican	Ref	Ref	Ref	Ref
Race/ethnicity
Black	−0.75 (−1.24, −0.27)	−0.65 (−1.11, −0.21)	−0.94 (−1.38, −0.49)	−0.20 (−0.61, 0.21)
White	Ref	Ref	Ref	Ref
Asian	−1.05 (−2.08, −0.01)	−0.12 (−1.13, 0.92)	−0.79 (−1.79, 0.23)	0.43 (−0.52, 1.46)
Hispanic	−0.55 (−0.96, −0.14)	−0.42 (−0.82, −0.03)	−0.28 (−0.67, 0.10)	0.18 (−0.17, 0.54)
Other	0.06 (−0.65, 0.80)	−0.06 (−0.72, 0.62)	0.38 (−0.28, 1.10)	0.07 (−0.51, 0.66)
Sample size for analysis	1139	1140	1137	1137

Fig. 1.

A–C. Predicted probabilities of support for government policy, by political affiliation, racial attitudes, and OUD stigma. In each graph, we used the R margins package to compute and plot predicted probabilities and 95% CI for each dependent variable, holding constant all other independent variables at the sample mean. A. Shows the predicted support for each of the Four OUD support outcomes, by Political Affiliation; B. Shows the predicted support for each OUD support outcome, by CoBRAS score; and C. Shows the predicted support for each OUD support outcome, by OUD Stigma score.

Results

Table 1 shows descriptive statistics. Our sample was 52% female, 63% White; 54% were at least 45 years of age. Forty-four percent leaned or identified as Democrats, and 36% leaned or identified as Republican. Overall, the median CoBRAS score was 3.0 and the average OUD stigma score was 3.33, each were on a scale of one to five. One-third of respondents reported knowing a family member with a history of opioid use. Nine percent reported histories of personal use. These proportions were slightly higher than the proportions reported in a 2017 national survey, where 12% of respondents reported having a family member currently addicted to opioids, and 5% reported that they themselves were addicted to some form of opioid (News, 2017).

Table 1

Sample characteristics (weighted).

	Total %
Age
18–29	20.7%
30–44	25.0%
45–59	24.3%
60+	30.0%
Race/ethnicity
Black	12.0%
Multiracial/other	4.6
Hispanic	16.7%
Asian	4.1%
White non-Hispanic	62.8%
Gender
Female	51.6%
Male	48.2%
Missing/other	0.2%
Education
Less than high school	9.8%
HS diploma/GED	27.8%
Vocational/some college	27.6%
Bachelor’s degree	19.9%
Post-grad/professional degree	14.8%
Income
<$25,000	20.0%
$25,000–49,000	25.3%
$50,000–84,000	22.5%
$85,000–150,000	24.6%
Over $150,000	7.6%
Geography
Northeast	17.3%
Midwest	20.7%
West	23.9%
South	38%
Personal conviction ever	11.1%
Family conviction ever	36.3%
Personal use ever	8.9%
Family use ever	33.2%
Mean COBRAS (SD)	3.0 (1.04)
Mean OUD stigma (SD)	3.3 (0.68)
Political affiliation
Democrat	34.3%
Lean democrat	10.8%
Don’t lean/ independent	15.5%
Lean republican	10.8%
Republican	28.1%
Favor expanding Medicaid insurance benefits for low-income families to provide coverage for prescription opioid disorders.
Strongly disagree	7.7%
Somewhat disagree	10.5%
Neither agree nor disagree	21.0%
Somewhat agree	32.9%
Strongly agree	28.0%
Favor making naloxone (also known as “Narcan”) available and affordable without a prescription
Strongly disagree	6.5%
Somewhat disagree	10.2%
Neither agree nor disagree	25.0%
Somewhat agree	29.8%
Strongly agree	28.4%
Mandatory addiction treatment can be an effective intervention.
Strongly disagree	5.8%
Somewhat disagree	10.3%
Neither agree nor disagree	29.0%
Somewhat agree	34.4%
Strongly agree	20.5%
Favor increasing government spending to improve treatment of opioid use disorder/addiction
Strongly disagree	9.0%
Somewhat disagree	11.0%
Neither agree nor disagree	25.4%
Somewhat agree	32.3%
Strongly agree	22.4%

In terms of specific public health responses, 61% agreed or strongly agreed with expanding Medicaid to finance addiction services for low-income patients. Fifty-eight percent of respondents agreed with making naloxone available without a prescription. Support was lower for mandatory treatment (55% agree/strongly agree) and increased government spending for opioid-related services (54%). While 42% of participants identified as Democrat or leaning Democrat, this group made up 71% of those supporting increased government spending. Conversely, Republican/lean Republican made up 36% of the full sample, but 71% of those who disagreed with increasing government spending.

Table 2 show the results of logistic regression analysis that explore respondents’ support for specific public health responses. Item nonresponse slightly reduced our sample available for multivariate analysis. As indicated in the last row of Table 2, our analysis sample was between 1137 and 1141 (out of a possible 1161) individuals for all reported regression analyses.

Examining political affiliations in the full model, that includes both CoBRAS and stigma scores, Democratic affiliation was associated with stronger support (relative to Republican affiliation) for all four dependent variables, with estimated logistic regression coefficients (β) of 1.30 (95% CI 0.87, 1.73) for Medicaid expansion; 0.93 (95% CI 0.53, 1.34) for increased government spending; 0.88 (95% CI 0.470, 1.29) for naloxone availability; and 0.57 (95% CI 0.19, 0.96) for mandatory treatment.

This association was stronger for Medicaid supports and was weakest for mandatory treatment. We found a clear trend by political affiliation, with smaller but significant associations for Lean Democrat and, in some models, Independent/no declared affiliation, compared to Republican, but not significant differences between Lean Republican and Republican. (Six individuals declined to answer this question. We included them within the category of independent/no declared affiliation. Excluding these individuals from the analysis or identifying them with a separate dummy variable had a negligible impact on our estimated coefficients or predicted probabilities.)

As seen in Fig. 1A. these results can be interpreted as predicted support (assuming mean values for all other variables) of 82% (95% CI 77%, 87%) for Medicaid expansion among strong Democrats compared with only 55% (95% CI 50%, 60%) among strong Republicans. For increased government spending, predicted support was 71% (95% CI 65%, 77%) for strong Democrats and only 49% (95% CI 44%, 54%) for strong Republicans. Similarly, predicted support for naloxone treatment was 76% (95% CI 70%, 81%) for strong Democrats and 56% (95% CI 51%, 61%) for strong Republicans. Predicted support for mandatory treatment was less politically polarized—favored by 64% (95% CI 58%, 70%) among strong Democrats compared to 50% (95% CI 45%, 54%) for strong Republicans.

In the fully adjusted models higher CoBRAS scores (indicating more negative attitudes towards Black Americans) were associated with lower support for all outcomes except mandatory treatment: adjusted β estimates of −0.55 (95% CI −0.77, −0.33) for Medicaid expansion; −0.51 (95% CI −0.68, −0.35) for increased government spending; and – 0.39 (95% CI −0.56, −0.23) for naloxone availability.

As seen in Fig. 1B, this translates to 85% predicted support for Medicaid expansion (95% CI 80%, 90%) at the lowest CoBRAS score, compared with 38% support (95% CI 29%, 47%) at the highest CoBRAS scores. For increased government spending, predicted public support was 78% (95% CI 72%, 84%) at the lowest CoBRAS scores and only 32% (95% CI 24%, 40%) at the highest CoBRAS scores. Finally, for naloxone availability, predicted support ranged from 79% (95% CI 73%, 85%) at the lowest CoBRAS score to 44% (95% CI 35%, 53%) at the highest CoBRAS score.

Higher OUD stigma was also associated with lower support for the same three outcomes: adjusted β estimates −0.55 (95% CI −0.77, −0.33) for Medicaid expansion; −0.50 (95% CI −0.71, −0.30) for increased government spending; and – 0.27 (95% CI −0.47, −0.07) for naloxone availability.

Again as seen in Fig. 1C, this can be interpreted as predicted support for Medicaid expansion of 88% (95% CI 82%, 94%) at the lowest stigma level compared with 45% (95% CI 35%, 54%) at the highest level. For increased government spending, predicted support was 82% (95% CI 74%, 89%) as the lowest stigma level and 37% (95% CI 29%, 46%) at the highest stigma level. Predicted support for increased naloxone availability was 77% (95% CI 68%, 86%) at the lowest level of stigma and 53% (95% CI 44%, 62%) at the highest level.”

Finally, we also examined respondents’ race/ethnicity as a predictor of public health responses to the opioid epidemic. We found that Black participants displayed lower support for Medicaid expansion, increased government spending on opioids, and naloxone availability when compared to non-Hispanic Whites. In like fashion, Hispanic and Asian-American participants displayed lower support than did non-Hispanic for Medicaid expansion, increased government spending on opioid-related efforts, and naloxone distribution.

Discussion

We found that race-conservative attitudes, as well as Republican affiliation were associated with lower support for Medicaid expansion, for increased government spending to address with OUD epidemic, or for expanding naloxone availability. As expected (Kennedy-Hendricks et al., 2017; Ezell et al., 2021) OUD stigma was also associated with decreased support for Medicaid expansion, naloxone availability, and government spending. While media depictions of the opioid epidemic may be less polarized on racial grounds than were highly-racialized prior drug epidemics (Netherland and Hansen, 2016; Mendoza et al., 2016; Mendoza et al., 2019), attitudes around race still play an important role in public support for public health responses to the opioid epidemic.

We also found racial and ethnic differences in support for naloxone distribution, in contrast to prior work, such as Kennedy-Hendricks et al., who found no difference in earlier survey data by race or ethnicity in support for similar treatment measures (Kennedy-Hendricks et al., 2017). This is one fruitful area for future studies to explore whether specific culturally-competent dialogue within Black communities is required regarding these evidence-based interventions.

While destigmatizing messages can increase support for public health responses to the opioid epidemic (Wakeman and Rich, 2018; McGinty and Barry, 2020; McGinty et al., 2018; Corrigan et al., 2017; McGinty et al., 2015), it is unclear if these would address political differences in support for such policies. Messaging that focuses on structural determinants may likewise be differentially effective by political group (Kennedy-Hendricks et al., 2017; McGinty and Barry, 2020). Novel culturally-competent messaging approaches by trusted messengers within different political and race/ethnic communities may be important to secure support for evidence-based treatment and harm reduction policies.

Our study cannot address whether such racial attitudes have a unique or distinctive association in the area of opioids, or whether these patterns reflect broader racialization of policy attitudes and beliefs in the Obama and Trump eras (Tesler, 2021). Our results may also reflect changing attitudes that track the changing nature of the opioid epidemic itself. Much published research reflects surveys conducted in 2014 or before. In subsequent years, opioid overdose mortality rates have rapidly increased among Black and Hispanic Americans, and now approach those observed among non-Hispanic whites (Opioid, 2021). Both areas are potentially fruitful for future research.

Despite partisan conflict in many domains, bipartisan legislation enacted under the George W. Bush, Obama, and Trump administrations supported mental health and substance use disorder parity in insurance coverage. These efforts provided significant resources for OUD treatment, and sought to emphasize prevention and treatment rather than criminal sanctions directed at drug users. Indeed, the behavioral health components of the Affordable Care Act won unanimous support within the Senate Finance Committee, including every Republican on the committee who voted against the final Senate bill (Friedmann et al., 2017). The opioid overdose epidemic has been cited as a key factor in bipartisan legislative support for the ACA’s Medicaid expansion in New Hampshire and other states (Grogan et al., 2020). This history suggests the potential for broad, bipartisan support for OUD treatment. Our findings suggest that this potential remains unrealized.

Conclusion

We find that race-conservative attitudes and political affiliation are associated with support (or lack thereof) for public-health responses to address OUD, even after controlling for OUD stigma. Although the opioid epidemic has been portrayed as less racially and politically divisive than were previous drug epidemics, underlying racial and political attitudes remain important. The design and implementation of politically and culturally-competent public messaging remains a key challenge in crafting and implementing evidence-based responses to the opioid epidemic.

	Expand Medicaid for low-income families to cover addiction treatment				Increase government spending to improve treatment of OUD
	Baseline(affiliation only)(model 1)	Baseline +CoBRAS(model 2)	Baseline +stigma(model 3)	Baseline +stigma +CoBRAS (model 4)	Baseline(affiliation only)(model 1)	Baseline +CoBRAS (model2)	Baseline +stigma(model 3)	Baseline +stigma +CoBRAS (model4)
CoBRAS	–	−0.59 (−0.77, −0.42)	–	−0.56 (−0.73, −0.39)	–	−0.55 (−0.75, −0.35)	–	−0.51 (−0.68, −0.35)
OUD stigma score	–	–	−0.60 (−0.81, −0.39)	−0.55 (−0.77, −0.33)	–	–	−0.54 (−0.71, −0.38)	−0.50 (−0.71, −0.30)
Affiliation
Democrat	2.06 (1.69, 2.44)	1.93 (1.56, 2.32)	1.93 (0.56, 2.32)	1.30 (0.87, 1.73)	1.71 (1.36, 2.06)	1.01 (0.61, 1.41)	1.55 (1.20, 1.91)	0.93 (0.53, 1.34)
Lean democrat	1.52 (1.04, 2.03)	0.90 (0.37, 1.45)	1.40 (0.91, 1.92)	0.78 (0.24, 1.33)	1.62 (1.14, 2.12)	0.96 (0.44, 1.49)	1.46 (0.97, 1.96)	0.88 (0.36, 1.42)
Independent/non-identified	0.59 (0.18, 1.00)	0.38 (−0.04, 0.81)	0.49 (0.08, 0.91)	0.19 (−0.25, 0.62)	0.43 (0.02, 0.84)	0.01 (−0.42, 0.43)	0.24 (−0.18, 0.65)	−0.06 (−0.49, 0.37)
Lean republican	0.39 (−0.06, 0.84)	0.40 (−0.06, 0.87)	0.33 (−0.13, 0.79)	0.31 (−0.16, 0.78)	0.04 (−0.42, 0.49)	−0.03 (−0.51, 0.43)	−0.05 (−0.52, 0.41)	−0.09 (−0.57, 0.38)
Republican	Ref	Ref	Ref	Ref	Ref	Ref	Ref	Ref
Race/ethnicity
Black	−0.42 (−0.88, 0.05)	−0.73 (−1.21, −0.25)	−0.46 (−0.93, 0.01)	−0.75 (−1.24, −0.27)	−0.40 (−0.83, 0.04)	−0.64 (−1.09, −0.20)	−0.42 (−0.86, 0.02)	−0.65 (−1.11, −0.21)
White	Ref	Ref	Ref	Ref	Ref	Ref	Ref	Ref
Asian	−1.19 (−2.22, −0.16)	−1.13 (−2.17, −0.10)	−1.08 (−2.12, −0.04)	−1.07 (−2.11, −0.03)	−0.30 (−1.31, 0.75)	−0.22 (−1.23, 0.83)	−0.19 (−1.20, 0.86)	−0.12 (−1.13, 0.92)
Hispanic	−0.46 (−0.86, −0.07)	−0.55 (−0.96, −0.15)	−0.45 (−0.85, −0.05)	−0.55 (−0.96, −0.14)	−0.38 (−0.76, −0.00)	−0.44 (−0.83–0.06)	−0.36 (−0.74, 0.03)	−0.42 (−0.82, −0.03)
Other	0.08 (−0.59, 0.78)	0.06 (−0.63, 0.79)	0.08 (−0.61, 0.80)	0.06 (−0.66, 0.80)	−0.02 (−0.65, 0.623)	−0.03 (−0.68, 0.64)	−0.03 (−0.67, 0.63)	−0.06 (−0.72, 0.62)
Sample size for analysis	1140	1140	1139	1139	1141	1141	1140	1140
	Naloxone available & affordable without prescription				Mandatory treatment is an effective way to help people with OUD
	Baseline(affiliation only)(model 1)	Baseline +CoBRAS(model 2)	Baseline +stigma(model 3)	Baseline +stigma +CoBRAS (model4)	Baseline(affiliation only)(model 1)	Baseline +CoBRAS (model2)	Baseline +stigma (model3)	Baseline +stigma +CoBRAS (model4)
CoBRAS	–	−0.42 (−0.58, −0.26)	–	−0.39 (−0.56, −0.23)	–	–	−0.01 (−0.19, 0.17)	−0.01 (−0.16, 0.14)
OUD stigma score	–	–	−0.32 (−0.52, −0.12)	−0.27 (−0.47, −0.07)	–	−0.01 (−0.20, 0.17)	–	−0.02 (−0.20, 0.17)
Affiliation
Democrat	1.44 (1.09, 1.80)	0.92 (0.51, 1.33)	1.36 (1.00, 1.72)	0.88 (0.47, 1.29)	0.60 (0.28, 0.93)	0.58 (0.20, 0.96)	0.60 (0.27, 0.93)	0.57 (0.19, 0.96)
Lean democrat	0.59 (0.14, 1.06)	0.07 (−0.43, 0.58)	0.52 (0.05, 0.99)	0.04 (−0.47, 0.55)	0.30 (−0.14, 0.73)	0.27 (−0.20, 0.75)	0.29 (−0.15, 0.73)	0.27 (−0.21, 0.74)
Independent/non-identified	0.52 (0.12, 0.93–0.50 (−0.71, −0.30))	0.27 (−0.14, 0.70)	0.47 (0.06, 0.89)	0.24 (−0.18, 0.67)	−0.07 (−0.46, 0.33)	−0.08 (−0.49, 0.32)	−0.07 (−0.47, 0.33)	−0.08 (−0.49, 0.32)
Lean republican	0.17 (−0.28, 0.62)	0.14 (−0.31, 0.60)	0.15 (−0.32, 0.59)	0.11 (−0.35, 0.57)	0.02 (−0.42, 0.46)	0.02 (−0.42, 0.46)	0.01 (−0.43, 0.45)	−0.01 (−0.46, 0.43)
Republican	Ref	Ref	Ref	Ref	Ref	Ref	Ref	Ref
Race/ethnicity
Black	−0.73 (−1.17, −0.30)	−0.93 (−1.38, −0.49)	−0.75 (−119, −0.32)	−0.94 (−1.38, −0.49)	−0.20 (−0.61, 0.20)	−0.20 (−0.61, 0.21)	−0.20 (−0.61, 0.20)	−0.20 (−0.61, 0.21)
White	Ref	Ref	Ref	Ref	Ref	Ref	Ref	Ref
Asian	−0.88 (−1.88, 0.13)	−0.84 (−1.83, 0.18)	−0.83 (−1.83, 0.18)	−0.79 (−1.79, 0.23)	0.42 (−0.53, 0.65)	0.42 (−0.53, 1.45)	0.42 (−0.54, 1.45)	0.42 (−0.53, 1.45)
Hispanic	−0.24 (−0.62, 0.14)	−0.30 (0.068, 0.09)	−0.23 (−0.61, 0.16)	−0.28 (−0.67, 0.10)	0.18 (−0.17, 0.54)	0.18 (−0.17, 0.54)	0.18 (−0.17, 0.54)	0.18 (−0.17, 0.54)
Other	0.39 (−0.27, 1.09)	0.39 (−0.28, 1.10)	0.39 (−0.27, 1.10)	0.38 (−0.28, 1.10)	0.07 (−0.51, 0.66)	0.07 (−0.51, 0.66)	0.07 (−0.51, 0.66)	0.07 (−0.51, 0.66)
Sample size for analysis	1138	1138	1137	1137	1138	1138	1137	1137

Adjusting for: Age, region, gender, education, personal CJI, family CJI, personal OUD, family OUD and all variables for which estimates are presented.

Bold signifies p < 0.05.