Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2026 Jul 7.
Published before final editing as: J Subst Use. 2025 Jul 7:10.1080/14659891.2025.2529806. doi: 10.1080/14659891.2025.2529806

Predictors of Enacted Stigma Following Disclosure Among People in Recovery From Opioid Use Disorder: A Machine Learning Approach

Mohammad Mousavi 1, Ethel Virginia Sticinski 1, E Carly Hill 1, Natalie M Brousseau 2, Jessica Hulsey 3, Lynn M Morrison 4, John F Kelly 5, Annie B Fox 6, Valerie A Earnshaw 1
PMCID: PMC12311684  NIHMSID: NIHMS2099270  PMID: 40857478

Abstract

Objective:

Individuals who are in recovery from opioid use disorder experience enacted stigma, which can undermine treatment retention and recovery. Stronger understanding of who is at risk of experiencing enacted stigma can inform intervention efforts to reduce experiences of enacted stigma, enhance wellbeing, and promote treatment outcomes among people in recovery from OUD. The current study applies a machine learning framework to examine predictors of enacted stigma among people in recovery from OUD.

Methods:

This study employed a longitudinal approach, with n=112 participants responding to surveys before a possible disclosure and again after three months. We tested three different machine learning models and used a variety of performance metrics to evaluate model performance.

Results:

The random forest model performed the best with an R-squared of 0.85, indicating that our predictors explained 85% of the variance in enacted stigma. Important predictors of enacted stigma were recovery duration, age, disclosure, current issues with drugs, and sobriety commitment.

Conclusions:

Individuals who are in recovery for a shorter time, did not disclose, have greater issues with drugs, and are younger were at higher risk of experiencing enacted stigma. Interventions may be needed to address stigma among people with these characteristics in treatment for OUD.

Keywords: Enacted Stigma, Opioid Use Disorder, Recovery, Machine Learning

Introduction

Although the benefits of treatment for opioid use disorder (OUD) are well established (Hoffman, et al., 2019), people in treatment for OUD face significant stigma. Stigma is a social process supported by power that ultimately results in the devaluation and discrediting of people with certain social statuses (Goffman, 1963; Link & Phelan, 2001), and it is a barrier to sustained recovery among people in treatment from substance use disorders (SUDs) (Kulesza et al., 2013). Enacted stigma, or experiences of prejudice, stereotyping, and discrimination from others, is detrimental to those in recovery given that it is associated with higher levels of distress, lower quality of life, less recovery capital and lower treatment adherence (Vilsaint et al., 2020; Earnshaw, 2020). People in recovery may be particularly at risk of experiencing enacted stigma surrounding disclosures, or when they share information about their substance use history, treatment, or recovery with someone new for the first time (Brousseau et al., 2022). Although past work has explored the negative outcomes of enacted stigma for people in recovery (Earnshaw, 2020), less research has focused on factors that predict enacted stigma among this population. A better understanding of who is at risk of experiencing enacted stigma surrounding disclosure can inform the tailoring of stigma intervention efforts. The current study addresses this gap by using machine learning methods, which enable the exploration of numerous predictor variables without compromising power, to broadly explore predictors of enacted stigma following disclosure among people in treatment for OUD.

Opioid Use Disorder Enacted Stigma

Opioid use disorder is heavily stigmatized (Cheetham et al., 2022), and people in recovery from OUD report continued experiences of enacted stigma during OUD treatment. For example, people in recovery experience enacted stigma from healthcare providers, noting that nurses in particular hold negative attitudes about people with OUD. These views manifest as poor or inequitable care for people seeking treatment for OUD (Kulesza et al., 2013; McCurry et al., 2023). Employers are another potent source of enacted stigma, as people in recovery report being fired or not hired by employers (Earnshaw et al., 2013). Finally, family appears to be a powerful source of enacted stigma. People in treatment for OUD report that their family members continue to view them negatively during recovery (Earnshaw et al., 2013), and enacted stigma from family members is associated with continued opioid use during treatment (Earnshaw et al., 2024). Individuals in treatment for OUD may be at risk of enacted stigma when they are disclosing some aspect of their opioid use history, diagnosis or treatment to others (Earnshaw et al., 2021).

To date, research has largely focused on the negative impacts of enacted stigma on people with OUD and other SUDs. Short-term psychological reactions to experiences of enacted stigma may include feelings of frustration, anger, anxiety, and sadness. Over time, experiences of enacted stigma are associated with higher levels of distress, lower quality of life, and less recovery capital (Vilsaint et al., 2020). Moreover, research suggests that enacted stigma may compromise treatment adherence and recovery efforts, and in some cases may lead to continued substance use (Reichert et al., 2023; Earnshaw, 2020; Earnshaw et al., 2024).

Yet, although somewhat limited in scope, previous research has yielded insight into factors that are associated with enacted stigma. Much of this work has explored associations between socio-demographic factors (e.g., race, gender, age), and enacted stigma. For example, Kulesza et al. (2013) reported that among people who inject drugs, African Americans reported higher levels of enacted stigma than White Americans. This finding was echoed by Mendoza et al. (2019), who suggested that the racialization of OUD has led to inequities in policy and clinical responses which have “differentiated” stigma experiences between racial groups. In contrast to findings on racial differences in enacted stigma, findings for gender and age are more mixed. For example, some studies note women perceive higher levels of enacted stigma than men (Agterberg et al., 2020; Meyers at al., 2021), while other studies have suggested men are more likely to experience stigmatizing attitudes toward SUD than women (Sattler et al., 2021; Goodyear et al., 2018; Miller et al., 2022). Similarly, some studies suggest that younger participants with SUD are more likely to perceive enacted stigma than older participants (Chio et al., 2014; Miller et al., 2022), while other studies suggested that older adults with SUD reported more stigma experiences (Thandi & Browne, 2019; Woodhead et al., 2019).

Previous research has additionally identified treatment- and recovery-related factors that may be associated with enacted stigma. As examples, some work suggests that use of medications like methadone or naltrexone are associated with enacted stigma (Brousseau et al., 2022; McCurry et al., 2023; Witte et al., 2021). Other work suggests that length of time in recovery is associated with enacted stigma, though the nature of the association between time in recovery and enacted stigma remains somewhat unclear. For example, some research suggests a negative association between endorsement of some stigma factors (e.g., perceptions of worthlessness or criminality) and recovery duration (Zurawel et al., 2024), whereas other research suggests a positive association between recovery duration (as measured by number of treatment episodes) and reported stigma among people with OUD (Luoma et al., 2007). Finally, evidence suggests that people with greater commitment to sobriety experience less enacted stigma (Judd et al., 2023; Ashford et al., 2019).

The majority of this past work, however, has studied predictors of enacted stigma in a siloed way. That is, it has largely examined associations between limited sets of predictors with enacted stigma using regression-based models. Methods that allow for the exploration of more predictors of enacted stigma simultaneously, or within the context of one analysis, may improve our understanding of who is at risk of experiencing enacted stigma during OUD treatment.

Current Study

The current study adopts a machine learning approach to assess multiple predictors of enacted stigma within a single study (Dwyer et al., 2018). Machine learning uses data to learn and predict new data based on properties that have been learned (Jordan & Mitchell, 2015). Whereas traditional regression models are often limited by the number of predictors that can be included, machine learning allows us to simultaneously analyze a greater number of predictors without compromising power. Moreover, machine learning achieves better accuracy than traditional regression approaches. We compare three different popular machine learning models used in psychology and health research, which include: support vector machines, random forests, and decision trees (Shatte et al., 2019). After selecting the best predictive model from our data analysis, we use that model to determine predictors of enacted stigma. Ultimately, results of this study can contribute to the identification of predictors of enacted stigma among people in recovery from opioid use disorder, helping to clarify who is at greatest risk of experiencing enacted stigma and therefore may be in need of intervention.

Methods

Procedures

Individuals were eligible to participate in the current study if they were receiving treatment for OUD, intending to disclose one or more people about their SUD history and/or treatment for the first time within the next three months, and were eighteen years of age or older. Participants completed quantitative surveys at baseline (i.e., before their disclosure) and again three months after, at which point they may or may not have disclosed. Surveys were conducted on tablets using Qualtrics. Participants received gift cards for each completed survey. Convenience sampling was used to recruit study participants from two SUD treatment clinics located in suburban and urban areas in Delaware, USA. Data was collected between March 2018 and May 2019. All participants provided informed consent. The protocol received IRB approval from the University of Delaware. A total of n=146 individuals participated at baseline. Participants with a complete follow-up assessment (n=112) were included in the current analyses.

Measures

Predictors include demographic variables, primary substance, current issues with drugs, addiction severity (Raistrick et al.,1994), commitment to sobriety (Kelly & Greene, 2014), and recovery duration. Enacted stigma (Smith et al., 2016) was considered as an outcome in this study.

Participant and Recovery Characteristics

Participants indicated their age, gender, race/ethnicity, education, marital status, employment status, recovery duration, and primary substance or drug. In response to the Risk Behavior Survey, participants indicated if they had used drugs within the previous 30 days (Needle et al., 1995). Responses were summed for participants who indicated they used drugs or alcohol. Those who did not use drugs or alcohol received a score of zero.

Enacted Stigma

Participants were asked about whether they had experienced enacted stigma from the disclosure recipient since disclosing with three questions from the Substance Use-Stigma Mechanisms Scale (Smith et al., 2020). Participants responded on a 5-point Likert-type scale ranging from “never” to “very often”, and a composite score was created by averaging responses (Cronbach’s α = 0.92).

Commitment to Sobriety

The Commitment to Sobriety Scale was used to measure commitment to sobriety. The scale is designed to be a brief assessment tool to gauge a person’s desire and drive to avoid substance use and stay in recovery (Kelly & Greene, 2014). The original response categories were recoded using a scale with 0 representing “strongly disagree” and 5 representing “strongly agree,” where a higher score indicated a stronger commitment to sobriety (Cronbach’s alpha = 0.93).

Data Analysis

We used Python 3 for the data analysis. for the data analysis. Prior to modeling, we searched for outliers and null values, among other data inconsistencies. We cleaned and preprocessed the data, standardized the numerical variables, and fixed any missing values. We used the mean and mode imputation method for missing data. For continuous variables, the mean was utilized, while for categorical variables, the mode was employed to fill in the null values. No outliers were detected in the examined variables. However, we observed a 3% rate of missing data in the recovery duration and primary substance variables, for which imputation was conducted.

Model Selection

Machine learning approaches prioritize description and explanation, and have a flexibility for handling nonlinear patterns (Pargent, et al., 2023). We tested three machine learning models: Support Vector Machine, Decision Tree, and Random Forest. Support Vector Machine (SVM) creates a boundary to separate different classes or predict continuous values by transforming data into a higher-dimensional space. SVM has the benefits of being simple to understand, performing well, and yielding suitable results even with limited data (Vapnik, 1999). Decision trees recursively split data into smaller subsets based on predictor variables, leading to a hierarchical structure. They can handle both regression and classification tasks. (Nasteski, 2017). Random forests combine multiple decision trees, each relying on random subsets of predictors. This ensemble technique is robust and less prone to overfitting (Liaw & Wiener, 2002).

In our analysis, the radial basis function (RBF) kernel was applied by the SVM regression model. The kernel function assesses the similarity among multiple samples (Kim et al., 2023). We used various functions like linear, polynomial, sigmoid, and radial basis functions (RBF) for this purpose. The decision tree model in our analysis had eight nodes and a depth of four. We also used an ensemble of 100 trees to build the Random Forest model. Training (80%) and testing (20%) sets were created from our dataset. Because the dataset was small, we used 5-fold cross-validation. Cross-validation enhances a model’s performance by making more efficient use of the available data. It involves partitioning the dataset into training and validation subsets multiple times to evaluate the model’s performance on unseen data. The process is to divide the dataset into k folds, generalize the behavior, and ultimately increase the efficiency and generalizability of our model (Yadav & Shukla, 2016). 5-fold cross-validation divides the data into five subsets, which makes it possible for the model to train and test on different portions of the data repeatedly.

Evaluation of Prediction Model

Based on recommendations from Chicco et al. (2021), we examined four performance metrics to evaluate model performance, including R-squared (R2), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and Mean Absolute Error (MAE). R-squared (R2): The percentage of the dependent variable’s variance that can be predicted based on the independent variables is known as the coefficient of determination. Mean Squared Error (MSE): MSE measures how far off the predictions of a model are from the actual values. It does this by averaging the squared differences between the predicted and actual values. Squaring the differences makes larger errors stand out more. I’s useful when we want to give more importance to bigger errors. Root Mean Squared Error (RMSE): RMSE is the square root of MSE and is easier to interpret than the MSE because it is on the same scale as the original data. Like the MSE, it highlights bigger mistakes, making it useful when large errors are a concern. Mean Absolute Error (MAE): MAE is the average of the absolute differences between predicted and actual values, without squaring. This means it treats all errors equally, without emphasizing the larger ones. I’s often used when we want a straightforward measure of model performance that is not overly affected by outliers. The model’s performance, however, will be mediocre if the test set contains a large number of outliers (Chicco et sl., 2021). We also computed feature importance scores in our analysis to determine each predictor’s significance. The calculation of feature importance follows standard practices commonly used in the field and was implemented using the Random Forest algorithm from the scikit-learn library. The contribution of each variable to the predictive performance of the model is indicated by these scores. The most significant variable, which is set at 100, is the basis for scoring. A score of 0 indicates that there is very little influence on the model’s performance, but it does not necessarily mean the variable should be eliminated from the model.

Results

Participants ranged in age from 18 to 76, with an average age of 38.43 (SD=11.38). 56.2% of participants identified as women and 43.8% as men. 12.5% of participants were Black or African-American, 78.6% were White, and the remaining participants were other racial or ethnic groups. 15.2% of participants were married, 40.2% were single, and 22.3 were in a relationship. In terms of education, 42.9% of participants were a high school graduate or GED, 22.3% had less than a high school education, and 34.8% had some college education or higher. 37.5% worked full or part-time, 39.3% were unemployed, and 23.2% were retired or disabled.

Table 1 contains the performance metrics (R-squared, MSE, RMSE, and MAE) for the three different machine learning models tested with the training set. The Random Forest showed superior performance in modeling based on all evaluation metrics. Therefore, we assessed the Random Forest model performance using the testing set. With the testing set, the model R-squared was 0.85, which means our predictors in the model explained 85% of the variance in the outcome. The MSE was 0.30, indicating the effectiveness of the Random Forest model in approximating the outcome variable based on our predictors with a relatively small error in the model’s estimation. The RMSE was .56, which indicates the average difference between predicted and actual values. It shows a moderate level of error. Finally, the MAE was 0.43, which means the model was moderately precise in estimating the outcome. Overall, based on R-squared, MSE, RMSE, and MAE scores, the Random Forest model demonstrated a strong predictive performance for enacted stigma.

Table 1.

Evaluation Metrics in Prediction Models

Models R-squared Mean Squared Error Root Mean Squared Error Mean Absolute Error
Random Forest 0.848 0.306 0.553 0.431
Support Vector Machine 0.603 0.659 0.812 0.553
Decision Tree 0.365 0.918 0.958 0.778
Open in a new tab

We extracted the predictors’ importance (Figure 1) from the Random Forest prediction model to explain the effectiveness and importance of each predictor individually on the outcome (enacted stigma). A higher score in predictors indicates greater effectiveness and importance on the enacted stigma. Moreover, based on each predictor’s importance we could identify who is at higher risk of experiencing enacted stigma among individuals in recovery from opioid use disorder.

Figure 1.

Figure 1

Open in a new tab

Predictors Importance on Enacted Stigma

Results suggest that important predictors of enacted stigma included recovery duration, age, disclosure, current issues with drugs, and sobriety commitment, respectively. Other predictors such as education status, relationship status, primary drug, employment status, incarceration, gender, and race had minimal importance on enacted stigma. To further clarify the effects of significant predictors on enacted stigma, we employed Partial Dependence Plots (Figure 2). The plots showed that individuals with shorter recovery duration, who were younger, who had not disclosed, and who had current issues with drug use reported greater enacted stigma than those with longer recovery duration, who were older, who had disclosed, and/or who had fewer issues with current drug use.

Figure 2.

Figure 2

Open in a new tab

Partial Dependence Plots

Discussion

This study aimed to identify risk factors associated with enacted stigma experiences among people in recovery from OUD. Using a machine learning approach to overcome limitations inherent in traditional regression models, several risk factors were identified. Recovery duration was the strongest predictor of enacted stigma, with participants who have been in recovery for shorter periods of time reporting more enacted stigma than those who have been in recovery for a relatively longer duration. This finding is inconsistent with Luoma et al. (2007), which found that individuals with more treatment episodes reported greater enacted stigma. Participants who have been in treatment for a shorter duration may also have greater issues with drugs. Indeed, participants with current issues with drugs also reported greater stigma in the current study. This finding is consistent with previous research (Kelly & Westerhoff, 2010; Hadland et al., 2018; Kulesza et al., 2013; Burgess et al., 2021). Future research should continue to explore the associations between time in recovery, issues with drugs, and enacted stigma to clarify how these variables are related to each other.

The current study also identified important substance use and treatment-related predictors of enacted stigma. Important predictors beyond recovery duration and current issues with drugs included disclosure and commitment to sobriety. Specifically, participants who disclosed reported less enacted stigma. Past research and theory suggests that people who anticipate greater stigma are less likely to disclose (Chaudoir & Fisher, 2010). It’s possible that participants of this study choose not to disclose in relationships wherein they were at greater risk of stigma, thereby reducing their exposure to enacted stigma. Moreover, similar to Judd et al. (2023) and Ashford et al. (2019), we found that people in recovery from OUD with greater commitment to sobriety experience less enacted stigma. Yet, less important predictors included primary substance or drug. The primary substance or drug may not have been an important predictor in this study because it was a relatively homogeneous sample of people in treatment for OUD. Some past work suggests that opioid use is more stigmatized than other types of substance use, perhaps putting people with OUD at greater risk of enacted stigma following disclosure than people with other SUD (Room et al., 2001; Earnshaw et al., 2019). Future research is therefore needed with more diverse samples of people with OUD and other SUDs to determine whether primary substance or drug is associated with experiences of stigma following disclosure.

Finally, regarding socio-demographic characteristics, age emerged as an important predictor of enacted stigma among people in recovery from opioid use disorder. Our research suggests that younger people in recovery experience greater enacted stigma than older people. Recent theory suggests that younger adults may experience more enacted stigma, in part, because they are transitioning between more environments than older adults (Earnshaw et al., 2022). That is, they may be moving between educational, work, social, and familial environments and they may encounter more stigma as they interact with new groups of people. In contrast, less important socio-demographic predictors included education status, relationship status. employment status, incarceration, gender, and race. Much research has been dedicated to understanding the impact of race and gender on stigma experiences. Some findings suggest that race and gender are associated with experiences of stigma among people with OUD or in recovery from SUDs (Earnshaw, 2020; Cheetham et al., 2022; Meyers et al., 2021; Nwanaji-Enwerem et al., 2024); yet in this study, both race and gender were relatively poor predictors of enacted stigma. Because this study assessed the predictive power of multiple factors, it is possible that race and/or gender were confounded by other demographic factors that were not explicitly measured in previous studies. Indeed, this study suggests that education and employment status are slightly stronger demographic predictors than race and gender, suggesting a strong need for quantitative research conducted through an intersectional lens or qualitative research to better understand how these other social positions impact experiences of enacted stigma.

The main strength of this study lies in the development of a prediction model utilizing a diverse set of predictors to identify people in recovery from OUD who may be at risk of experiencing enacted stigma. This addresses a gap in current research, as identifying these risk factors can inform interventions to promote treatment outcomes for this population. Yet, there are several limitations of the study that may be improved upon in future research. The overall performance and reliability of regression models may be impacted by the smaller sample size, which may limit the model’s capacity to generalize the prediction models. Data were collected around a possible disclosure event. Although this is an important time to study enacted stigma given that disclosures can be associated with enacted stigma, future research should explore the extent to which findings generalize to non-disclosure contexts. The sample was relatively homogeneous in terms of geography, some socio-demographic characteristics (e.g., race, ethnicity), and substance use disorder characteristics. Future research should also explore the generalizability of these findings in more diverse samples.

Enacted stigma experienced by people in recovery from OUD acts as a powerful barrier to both treatment access and treatment adherence. This study identified and explored multiple factors associated with enacted stigma experiences among this population. Additional research is required to examine and explain other predictors, such as social support and healthcare accessibility on enacted stigma. Identifying important predictors of enacted stigma among people in recovery from opioid use disorder can offer guidance to protect those at risk, facilitate treatment access and delivery, and promote recovery efforts.

Acknowledgements

The authors are grateful to all participants for their contributions to this study.

Funding

This work was supported by the National Institute on Drug Abuse (NIDA; K01DA042881, VAE) and National Institute on Alcohol Abuse and Alcoholism (NIAAA, K24AA022136, JFK). Funders were not involved in the study design; collection, analysis, or interpretation of the data; write up of the report; or decision to submit the article for publication. The content is solely the responsibility of the authors and does not necessarily represent the official views of its funders.

Footnotes

Conflict of Interest

The authors declared no conflict of interest for this research.

Ethics Approval Statement

All participants provided informed consent. The study was approved by the Institutional Review Board of the University of Delaware.

REFERENCES

  1. Agterberg S, Schubert N, Overington L, & Corace K (2020). Treatment barriers among individuals with co-occurring substance use and mental health problems: Examining gender differences. Journal of substance abuse treatment, 112, 29–35. [DOI] [PubMed] [Google Scholar]
  2. Ashford RD, Brown AM, Canode B, McDaniel J, & Curtis B (2019). A mixed-methods exploration of the role and impact of stigma and advocacy on substance use disorder recovery. Alcoholism Treatment Quarterly, 37(4), 462–480. [Google Scholar]
  3. Brousseau NM, Farmer H, Karpyn A, Laurenceau JP, Kelly JF, Hill EC, & Earnshaw VA (2022). Qualitative characterizations of misinformed disclosure reactions to medications for opioid use disorders and their consequences. Journal of substance abuse treatment, 132, 108593. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Brousseau NM, Karpyn A, Laurenceau JP, Farmer HR, Kelly JF, Hill EC, & Earnshaw VA (2022). The Impacts of Social Support and Relationship Characteristics on Commitment to Sobriety Among People in Opioid Use Disorder Recovery. Journal of Studies on Alcohol and Drugs, 83(5), 646–652. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Burgess A, Bauer E, Gallagher S, Karstens B, Lavoie L, Ahrens K, & O’Connor A (2021). Experiences of stigma among individuals in recovery from opioid use disorder in a rural setting: A qualitative analysis. Journal of Substance Abuse Treatment, 130, 108488. [DOI] [PubMed] [Google Scholar]
  6. Chaudoir SR, & Fisher JD (2010). The disclosure processes model: understanding disclosure decision making and postdisclosure outcomes among people living with a concealable stigmatized identity. Psychological bulletin, 136(2), 236. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Cheetham A, Picco L, Barnett A, Lubman DI, & Nielsen S (2022). The impact of stigma on people with opioid use disorder, opioid treatment, and policy. Substance abuse and rehabilitation, 1–12. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Chicco D, Warrens MJ, & Jurman G (2021). The coefficient of determination R-squared is more informative than SMAPE, MAE, MAPE, MSE and RMSE in regression analysis evaluation. PeerJ Computer Science, 7, e623. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Choi NG, DiNitto DM, & Marti CN (2014). Treatment use, perceived need, and barriers to seeking treatment for substance abuse and mental health problems among older adults compared to younger adults. Drug and alcohol dependence, 145, 113–120. [DOI] [PubMed] [Google Scholar]
  10. Dwyer DB, Falkai P, & Koutsouleris N (2018). Machine learning approaches for clinical psychology and psychiatry. Annual review of clinical psychology, 14, 91–118. [DOI] [PubMed] [Google Scholar]
  11. Earnshaw VA (2020). Stigma and substance use disorders: A clinical, research, and advocacy agenda. American Psychologist, 75(9), 1300–1311. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Earnshaw VA, & Kalichman SC (2013). Stigma experienced by people living with HIV/AIDS. In Stigma, discrimination and living with HIV/AIDS: A cross-cultural perspective (pp. 23–38). Dordrecht: Springer Netherlands. [Google Scholar]
  13. Earnshaw VA, Bogart LM, Menino DD, Kelly JF, Chaudoir SR, Reed NM, & Levy S (2019). Disclosure, stigma, and social support among young people receiving treatment for substance use disorders and their caregivers: A qualitative analysis. International Journal of Mental Health and Addiction, 17, 1535–1549. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Earnshaw VA, Brousseau NM, Hill EC, Laurenceau JP, Sepucha KR, Qiu X, Mousavi M, Morrison LM, & Kelly JF (2024). Enacted stigma from family and drug use during opioid use disorder treatment. Stigma and Health. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Earnshaw VA, Sepucha KR, Laurenceau JP, Subramanian SV, Brousseau NM, Chaudoir SR, … & Kelly JF (2021). Disclosure processes as predictors of relationship outcomes among people in recovery from opioid use disorder: A longitudinal analysis. Drug and Alcohol Dependence, 228, 109093. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Earnshaw VA, Smith LR, & Copenhaver M (2013). Drug addiction in the context of methadone maintenance therapy: An investigation into understudied sources of stigma. International Journal of Mental Health and Addiction, 11(1), 110–122. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Earnshaw VA, Smith LR, Chaudoir SR, Amico KR, & Copenhaver MM (2013). HIV stigma mechanisms and well-being among PLWH: a test of the HIV stigma framework. AIDS and Behavior, 17, 1785–1795. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Earnshaw VA, Watson RJ, Eaton LA, Brousseau NM, Laurenceau JP, & Fox AB (2022). Integrating time into stigma and health research. Nature Reviews Psychology, 1(4), 236–247. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Goffman E Stigma: Notes on the Management of Spoiled Identity. Prentice Hall; 1963. [Google Scholar]
  20. Goodyear K, Haass-Koffler CL, & Chavanne D (2018). Opioid use and stigma: The role of gender, language and precipitating events. Drug and alcohol dependence, 185, 339–346. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Hadland SE, Park TW, & Bagley SM (2018). Stigma associated with medication treatment for young adults with opioid use disorder: a case series. Addiction Science & Clinical Practice, 13, 1–4. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Hoffman KA, Ponce Terashima J, & McCarty D (2019). Opioid use disorder and treatment: challenges and opportunities. BMC health services research, 19, 1–5. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Jordan MI, & Mitchell TM (2015). Machine learning: Trends, perspectives, and prospects. Science, 349(6245), 255–260. [DOI] [PubMed] [Google Scholar]
  24. Judd H, Meier CL, Yaugher AC, Campbell E, & Atismé-Bevins K (2023). Opioid use disorder stigma reduction through story telling narrative and sharing: a qualitative review and lessons learned. International Journal of Mental Health and Addiction, 21(1), 468–483. [Google Scholar]
  25. Judd H, Yaugher AC, O’Shay S, & Meier CL (2023). Understanding Stigma Through the Lived Experiences of People with Opioid Use Disorder. Drug and Alcohol Dependence, 110873. [DOI] [PubMed] [Google Scholar]
  26. Kelly JF, & Greene MC (2014). Beyond motivation: Initial validation of the commitment to sobriety scale. Journal of Substance Abuse Treatment, 46(2), 257–263. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Kelly JF, & Westerhoff CM (2010). Does it matter how we refer to individuals with substance-related conditions? A randomized study of two commonly used terms. International Journal of Drug Policy, 21(3), 202–207. [DOI] [PubMed] [Google Scholar]
  28. Kim JH, Kwon OY, Hwang UJ, Jung SH, & Gwak GT (2023). Prediction model of subacromial pain syndrome in assembly workers using shoulder range of motion and muscle strength based on support vector machine. Ergonomics, 1–10. [DOI] [PubMed] [Google Scholar]
  29. Kulesza M, Larimer ME, & Rao D (2013). Substance use related stigma: what we know and the way forward. Journal of addictive behaviors, therapy & rehabilitation, 2(2). [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Liaw A, & Wiener M (2002). Classification and regression by randomForest. R news, 2(3), 18–22. [Google Scholar]
  31. Link BG, & Phelan JC (2001). Conceptualizing stigma. Annual review of Sociology, 27(1), 363–385. [Google Scholar]
  32. Luoma JB, Twohig MP, Waltz T, Hayes SC, Roget N, Padilla M, & Fisher G (2007). An investigation of stigma in individuals receiving treatment for substance abuse. Addictive behaviors, 32(7), 1331–1346. [DOI] [PubMed] [Google Scholar]
  33. McCurry MK, Avery-Desmarais S, Schuler M, Tyo M, Viveiros J, & Kauranen B (2023). Perceived stigma, barriers, and facilitators experienced by members of the opioid use disorder community when seeking healthcare. Journal of nursing scholarship, 55(3), 701–710. [DOI] [PubMed] [Google Scholar]
  34. Mendoza S, Hatcher AE, & Hansen H (2019). Race, stigma, and addiction. The stigma of addiction: An essential guide, 131–152. [Google Scholar]
  35. Meyers SA, Earnshaw VA, D’Ambrosio B, Courchesne N, Werb D, & Smith LR (2021). The intersection of gender and drug use-related stigma: A mixed methods systematic review and synthesis of the literature. Drug and alcohol dependence, 223, 108706. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Miller PK, Cuthbertson CA, & Loveridge S (2022). Social status influence on stigma towards mental illness and substance use disorder in the United States. Community Mental Health Journal, 58(2), 249–260. [DOI] [PubMed] [Google Scholar]
  37. Nasteski V (2017). An overview of the supervised machine learning methods. Horizons. b, 4, 51–62. [Google Scholar]
  38. Needle R, Fisher DG, Weatherby N, Chitwood D, Brown B, Cesari H, … & Braunstein M (1995). Reliability of self-reported HIV risk behaviors of drug users. Psychology of addictive behaviors, 9(4), 242. [Google Scholar]
  39. Nwanaji-Enwerem U, Redeker NS, O’Connell M, Barry D, Iheanacho T, Knobf TM, … & Sadler LS (2024). Experiences of stigma and discrimination compounded by intersecting identities among individuals receiving medication for opioid use disorder. Journal of Health Care for the Poor and Underserved, 35(1), 94–115. [PMC free article] [PubMed] [Google Scholar]
  40. Pargent F, Schoedel R, & Stachl C (2023). Best practices in supervised machine learning: A tutorial for psychologists. Advances in Methods and Practices in Psychological Science, 6(3), 25152459231162559. [Google Scholar]
  41. Raistrick D, Bradshaw J, Tober G, Weiner J, Allison J, & Healey C (1994). Development of the Leeds Dependence Questionnaire (LDQ): a questionnaire to measure alcohol and opiate dependence in the context of a treatment evaluation package. Addiction, 89(5), 563–572. [DOI] [PubMed] [Google Scholar]
  42. Reichert J, Del Pozo B, & Taylor B (2023). Police stigma toward people with opioid use disorder: A study of Illinois officers. Substance use & Misuse, 58(12), 1493–1504. [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Room R, Rehm J, Trotter RT II, Paglia A, & Üstün TB (2001). Cross-cultural views on stigma valuation, parity, and societal attitudes towards disability. In Üstün TB, Chatterji S, Bickenbach JE, Badley NE, & Trotter EMMN (Eds.), Disability and culture: Universalism and diversity; (pp. 247–291). [Google Scholar]
  44. Shatte AB, Hutchinson DM, & Teague SJ (2019). Machine learning in mental health: a scoping review of methods and applications. Psychological medicine, 49(9), 1426–1448. [DOI] [PubMed] [Google Scholar]
  45. Smith LR, Earnshaw VA, Copenhaver MM, & Cunningham C (2016). Substance use stigma: Reliability and validity of a theory-based scale for substance-using populations. Drug and Alcohol Dependence, 162, 34–43. [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. Smith LR, Mittal ML, Wagner K, Copenhaver MM, Cunningham CO, & Earnshaw VA (2020). Factor structure, internal reliability and construct validity of the Methadone Maintenance Treatment Stigma Mechanisms Scale (MMT-SMS). Addiction, 115(2), 354–367. [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Thandi MKG, & Browne AJ (2019). The social context of substance use among older adults: Implications for nursing practice. Nursing Open, 6(4), 1299–1306. [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Vapnik VN (1999). An overview of statistical learning theory. IEEE transactions on neural networks, 10(5), 988–999. [DOI] [PubMed] [Google Scholar]
  49. Vilsaint CL, Hoffman LA, & Kelly JF (2020). Perceived discrimination in addiction recovery: Assessing the prevalence, nature, and correlates using a novel measure in a US National sample. Drug and alcohol dependence, 206, 107667. [DOI] [PMC free article] [PubMed] [Google Scholar]
  50. Witte TH, Jaiswal J, Mumba MN, & Mugoya GC (2021). Stigma surrounding the use of medically assisted treatment for opioid use disorder. Substance Use & Misuse, 56(10), 1467–1475. [DOI] [PubMed] [Google Scholar]
  51. Woodhead EL, Timko C, Han X, & Cucciare MA (2019). Stigma, treatment, and health among stimulant users: life stage as a moderator. Journal of applied developmental psychology, 60, 96–104. [DOI] [PMC free article] [PubMed] [Google Scholar]
  52. Yadav S, & Shukla S (2016, February). Analysis of k-fold cross-validation over hold-out validation on colossal datasets for quality classification. In 2016 IEEE 6th International conference on advanced computing (IACC) (pp. 78–83). IEEE. [Google Scholar]
  53. Zurawel JA, Lacasse K, Spas JJ, & Merrill JE (2024). Identifying the trajectory of public stigma across the stages of recovery from a substance use disorder. Stigma and Health. [Google Scholar]

RESOURCES