Social Determinants of Health and 1-Year Buprenorphine Initiation Among Justice-Involved Veterans - PMC Skip to main content An official website of the United States government Here's how you know Here's how you know Official websites use .gov A .gov website belongs to an official government organization in the United States. Secure .gov websites use HTTPS A lock ( Lock Locked padlock icon ) or https:// means you've safely connected to the .gov website. Share sensitive information only on official, secure websites. Search Log in Dashboard Publications Account settings Log out Search… Search NCBI Primary site navigation Search Logged in as: Dashboard Publications Account settings Log in Search PMC Full-Text Archive Search in PMC Journal List User Guide PERMALINK Copy As a library, NLM provides access to scientific literature. Inclusion in an NLM database does not imply endorsement of, or agreement with, the contents by NLM or the National Institutes of Health. Learn more: PMC Disclaimer | PMC Copyright Notice JAMA Netw Open . 2026 Apr 17;9(4):e268497. doi: 10.1001/jamanetworkopen.2026.8497 Search in PMC Search in PubMed View in NLM Catalog Add to search Social Determinants of Health and 1-Year Buprenorphine Initiation Among Justice-Involved Veterans Sharmin Sultana Sharmin Sultana , BSc 1 Center for Health Optimization & Implementation Research (CHOIR), Veterans Affairs (VA) Bedford Healthcare System, Bedford, Massachusetts 2 Miner School of Computer and Information Sciences, University of Massachusetts Lowell, Lowell Find articles by Sharmin Sultana 1, 2 , Feiyun Ouyang Feiyun Ouyang , PhD 1 Center for Health Optimization & Implementation Research (CHOIR), Veterans Affairs (VA) Bedford Healthcare System, Bedford, Massachusetts 2 Miner School of Computer and Information Sciences, University of Massachusetts Lowell, Lowell Find articles by Feiyun Ouyang 1, 2 , Junhui Qian Junhui Qian , MSc 1 Center for Health Optimization & Implementation Research (CHOIR), Veterans Affairs (VA) Bedford Healthcare System, Bedford, Massachusetts 2 Miner School of Computer and Information Sciences, University of Massachusetts Lowell, Lowell Find articles by Junhui Qian 1, 2 , Yifan Zhang Yifan Zhang , MSc 1 Center for Health Optimization & Implementation Research (CHOIR), Veterans Affairs (VA) Bedford Healthcare System, Bedford, Massachusetts 2 Miner School of Computer and Information Sciences, University of Massachusetts Lowell, Lowell Find articles by Yifan Zhang 1, 2 , Avijit Mitra Avijit Mitra , PhD 1 Center for Health Optimization & Implementation Research (CHOIR), Veterans Affairs (VA) Bedford Healthcare System, Bedford, Massachusetts 3 Manning College of Information and Computer Sciences, University of Massachusetts Amherst, Amherst Find articles by Avijit Mitra 1, 3 , Richeek Pradhan Richeek Pradhan , MD, PhD 4 Centre for Medicine Use and Safety, Monash University, Parkville, Victoria, Australia Find articles by Richeek Pradhan 4 , Yousef Moradi Yousef Moradi , PhD 5 Center for Biomedical and Health Research in Data Sciences (CHORDS), University of Massachusetts Lowell, Lowell 6 Health Metrics and Evaluation Research Center, Research Institute for Health Development, Kurdistan University of Medical Sciences, Sanandaj, Iran Find articles by Yousef Moradi 5, 6 , Terri K Pogoda Terri K Pogoda , PhD 7 CHOIR, VA Boston Healthcare System, Boston, Massachusetts 8 Department of Health Law, Policy & Management, Boston University School of Public Health, Boston, Massachusetts Find articles by Terri K Pogoda 7, 8 , Joel I Reisman Joel I Reisman , AB 1 Center for Health Optimization & Implementation Research (CHOIR), Veterans Affairs (VA) Bedford Healthcare System, Bedford, Massachusetts Find articles by Joel I Reisman 1 , Hong Yu Hong Yu , PhD 1 Center for Health Optimization & Implementation Research (CHOIR), Veterans Affairs (VA) Bedford Healthcare System, Bedford, Massachusetts 2 Miner School of Computer and Information Sciences, University of Massachusetts Lowell, Lowell 3 Manning College of Information and Computer Sciences, University of Massachusetts Amherst, Amherst 5 Center for Biomedical and Health Research in Data Sciences (CHORDS), University of Massachusetts Lowell, Lowell Find articles by Hong Yu 1, 2, 3, 5, ✉ Author information Article notes Copyright and License information 1 Center for Health Optimization & Implementation Research (CHOIR), Veterans Affairs (VA) Bedford Healthcare System, Bedford, Massachusetts 2 Miner School of Computer and Information Sciences, University of Massachusetts Lowell, Lowell 3 Manning College of Information and Computer Sciences, University of Massachusetts Amherst, Amherst 4 Centre for Medicine Use and Safety, Monash University, Parkville, Victoria, Australia 5 Center for Biomedical and Health Research in Data Sciences (CHORDS), University of Massachusetts Lowell, Lowell 6 Health Metrics and Evaluation Research Center, Research Institute for Health Development, Kurdistan University of Medical Sciences, Sanandaj, Iran 7 CHOIR, VA Boston Healthcare System, Boston, Massachusetts 8 Department of Health Law, Policy & Management, Boston University School of Public Health, Boston, Massachusetts Accepted for Publication: February 22, 2026. Published: April 17, 2026. doi: 10.1001/jamanetworkopen.2026.8497 Open Access: This is an open access article distributed under the terms of the CC-BY-NC-ND License , which does not permit alteration or commercial use, including those for text and data mining, AI training, and similar technologies. © 2026 Sultana S et al. JAMA Network Open . ✉ Corresponding Author: Hong Yu, PhD, CHOIR, Veterans Affairs Bedford Healthcare System, 200 Springs Rd (152), Bldg 70, Bedford, MA 01730 ( [email protected] ). Author Contributions: Dr Yu and Ms Sultana had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Ms Sultana and Dr Ouyang contributed equally. Concept and design: Sultana, Ouyang, Qian, Mitra, Pradhan, Moradi, Yu. Acquisition, analysis, or interpretation of data: Sultana, Ouyang, Qian, Zhang, Mitra, Pradhan, Moradi, Pogoda, Reisman. Drafting of the manuscript: Sultana, Ouyang, Mitra, Moradi, Reisman. Critical review of the manuscript for important intellectual content: Ouyang, Qian, Zhang, Pradhan, Moradi, Pogoda, Reisman, Yu. Statistical analysis: Sultana, Ouyang, Qian, Mitra, Pradhan, Moradi, Reisman. Administrative, technical, or material support: Qian, Moradi. Supervision: Yu. Conflict of Interest Disclosures: Dr Pogoda reported receiving grants from Veterans Affairs (VA) Boston Healthcare System during the conduct of the study. No other disclosures were reported. Funding/Support: This work is supported in part by grant I01HX003711 from the US Department of Veterans Affairs (VA), Veterans Health Administration, Office of Research and Development, VA Health Systems Research (Dr Yu) and by grant R01DA056470 from the National Institutes of Health (NIH) (Dr Yu). Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication. Disclaimer: The contents and views in this article are the responsibility of the authors and do not necessarily reflect the position or policy of the VA, NIH, or the US government. Data Sharing Statement: See Supplement 2 . ✉ Corresponding author. Received 2025 Dec 3; Accepted 2026 Feb 22; Collection date 2026 Apr. Copyright 2026 Sultana S et al. JAMA Network Open . This is an open access article distributed under the terms of the CC-BY-NC-ND License, which does not permit alteration or commercial use, including those for text and data mining, AI training, and similar technologies. PMC Copyright notice PMCID: PMC13090846 PMID: 41996110 See commentary " Social Risk and Buprenorphine Initiation Among Justice-Involved Veterans. " on page e268494. This case-control study evaluates the association between social determinants of health (SDOH) and uptake of buprenorphine for opioid use disorder among veterans with incarceration history or other legal involvement. Key Points Question What is the 1-year buprenorphine medication for opioid use disorder (OUD; hereafter buprenorphine) initiation rate among justice-involved veterans (JIVs) with OUD, and how are social determinants of health (SDOH) associated with treatment uptake? Findings In this nested case-control study of 1372 case episodes with buprenorphine initiation matched with 5436 control episodes without initiation drawn from 13 321 OUD episodes involving 12 511 JIVs, only 11.16% initiated buprenorphine within 1 year. Multiple SDOH domains—financial problems, food insecurity, social problems, and transitions of care—were associated with lower odds of buprenorphine initiation. Meaning The findings of this study suggest that buprenorphine initiation among JIVs is low and social challenges substantially hinder treatment uptake, underscoring the need for interventions that address SDOH to improve access to effective OUD treatment. Abstract Importance Justice-involved veterans (JIVs), with incarceration history or other legal involvement, experience high opioid use disorder (OUD) rates but low uptake of medications for OUD (MOUD). Given its common use, assessing the associations of buprenorphine medication for OUD (hereafter buprenorphine) with social determinants of health (SDOH) is critical. Objective To assess the association between SDOH and buprenorphine initiation, and to estimate the 1-year initiation rate across demographic subgroups. Design, Setting, and Participants This nested case-control study obtained data from the Veterans Health Administration (VHA) national electronic health record system. JIVs with an OUD diagnosis receiving VHA care between fiscal years 2015 and 2020 were included. Cases were OUD episodes with buprenorphine initiation, which were matched 1:4 with control episodes without initiation. The unit of analysis was the OUD episode, a period of 1 year following initial OUD diagnosis. Data analyses were conducted in August 2025. Exposures Exposure to defined SDOH domains from the index date (first OUD diagnosis) until buprenorphine initiation date, 365 days after index date, or study end date, whichever occurred first, compared with no exposure. Main Outcomes and Measures The primary outcomes were the 1-year buprenorphine initiation rate and the association of SDOH with initiation. Conditional logistic regression adjusted for clinical and demographic covariates was used to estimate adjusted odds ratios (AORs) and 95% CIs for the association between buprenorphine initiation and 8 domains of SDOH extracted from structured data and unstructured clinical notes. Results Among 13 321 new OUD episodes in 12 511 JIVs, 1372 were case episodes (involving 1279 male JIVs [93.2%]; mean [SD] age, 40.8 [12.8] years) matched with 5436 control episodes (involving 5334 male JIVs [98.1%]; mean [SD] age, 39.2 [12.1] years). The 1-year buprenorphine initiation rate was 11.16% (95% CI, 10.63-11.71). Among SDOH domains identified from combined structured and unstructured data, violence (AOR, 0.31; 95% CI, 0.25-0.38), nonspecific psychosocial problems (AOR, 0.35; 95% CI, 0.30-0.40), food insecurity (AOR, 0.36; 95% CI, 0.30-0.44), and transitions of care (AOR, 0.50; 95% CI, 0.35-0.71) were associated with lower odds of buprenorphine initiation. Additional factors included barriers to care (AOR, 0.54; 95% CI, 0.45-0.64), financial problems (using structured data: AOR, 0.33; 95% CI, 0.28-0.40), and social problems (using structured data: AOR, 0.58; 95% CI, 0.44-0.79). Conclusions and Relevance This case-control study found that buprenorphine initiation remained low and that multiple SDOH domains acted as significant barriers to treatment uptake among JIVs. The findings emphasize the importance of systematic SDOH screening and interventions to address barriers to buprenorphine initiation. Introduction Opioid use disorder (OUD) is a condition characterized by misuse of opioids, including prescription painkillers, heroin, and fentanyl, resulting in tolerance, dependence, and withdrawal. 1 In the US, OUD remains a leading public health crisis, with opioid overdose deaths rising substantially in recent years. 2 Veterans experience nearly twice the overdose mortality of nonveterans, due in part to higher OUD and comorbid conditions. 3 Justice-involved veterans (JIVs)—those with a history of incarceration or legal involvement—experience disproportionately higher OUD rates and compounded challenges, including homelessness, unemployment, mental health conditions, and substance use disorders (SUDs), 4 compared with other veterans. 5 To mitigate these disparities, the US Department of Veterans Affairs (VA) launched the Veterans Justice Outreach (VJO), 6 a program that operates in courts and prisons to connect JIVs to VA health care, housing, employment, and behavioral health services. Similarly, the Veterans Treatment Court program offers rehabilitation over incarceration. 6 , 7 , 8 Despite these efforts, coordination remains fragmented, 9 , 10 leaving many JIVs with OUD vulnerable to overdose and poor long-term outcomes. Medications for OUD (MOUD), specifically methadone, buprenorphine, and naltrexone, are the gold standard of treatment. 11 , 12 Of these drugs, buprenorphine is the most prescribed due to its accessibility and having fewer restrictions than methadone. 13 , 14 Yet, only 45% of veterans with OUD receive MOUD, 15 and nearly one-third discontinue treatment within 1 year. 16 , 17 Among JIVs, only 27% with prior prison involvement and 34% with jail or court histories receive MOUD. 18 This disparity underscores the urgency of improving access to and initiation of buprenorphine MOUD (hereafter buprenorphine) for JIVs. Social determinants of health (SDOH) are substantial factors in health outcomes. 19 Veterans with OUD often face financial hardship, limited social support, and unstable housing, which hinder treatment initiation and retention. 20 , 21 , 22 For JIVs, these barriers are magnified by the stigma and reentry challenges after incarceration. 23 Most studies rely on structured administrative data, which capture only a fraction of SDOH. For example, International Statistical Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) Z codes for SDOH are infrequently documented, whereas unstructured clinical notes contain far richer contextual information. 24 Advances in natural language processing (NLP) make it possible to extract SDOH from these clinical narratives, enabling large-scale study of SDOH’s association with treatment outcomes. 25 The opioid epidemic and criminal-legal reform are both urgent public health priorities; thus, understanding barriers to buprenorphine initiation among JIVs is critical. To achieve this, we conducted a nested case-control study of JIV using data from the Veterans Health Administration (VHA) national electronic health record (EHR) system. Our objectives were to assess the association between SDOH (captured from structured data and unstructured clinical notes) and buprenorphine initiation and to estimate the 1-year initiation rate across demographic subgroups. Identifying which SDOH affect initiation can help address treatment barriers for veterans with OUD. Method Study Design, Setting, and Population This case-control study used data from the VHA EHR system and the VHA Corporate Data Warehouse (CDW), which contains administrative health care information such as demographic characteristics, prescription medications, diagnoses, procedures, clinical notes, and death data. 26 The VA Bedford Healthcare System Institutional Review Board approved the study procedures and waived the informed consent requirement because the study used retrospective, deidentified EHR data and involved minimal risk to participants. We followed the Strengthening the Reporting of Observational Studies in Epidemiology ( STROBE ) reporting guideline. The base population included all JIVs enrolled in the VJO program who had a new OUD diagnosis between fiscal years (FYs) 2015 and 2020. FY 2015 was selected because unstructured clinical notes required for NLP extraction were consistently available beginning in FY 2015, and the study period was limited to FY 2020 (September 30) to minimize potential COVID-19–related disruptions. A new OUD episode required at least 1 inpatient or outpatient encounter coded with an ICD-9 or ICD-10 diagnosis code (eTable 1 in Supplement 1 ) with no OUD diagnosis in the prior 365 days. Eligible veterans were aged 20 to 99 years and had at least 1 VJO encounter in the preceding year. Baseline covariates and SDOH were assessed during the 2 years prior to the index date. The index date was the date of first OUD diagnosis, with the end of follow-up being buprenorphine initiation date, 365 days after index date, or study end date, whichever occurred first. We examined the association between SDOH and buprenorphine initiation using a nested case-control design. The unit of analysis was the OUD episode. Cases included OUD episodes in which veterans initiated buprenorphine within 1 year of diagnosis. Each case was randomly matched with a replacement, with up to 4 controls from episodes in which buprenorphine was not initiated. This design was selected to align with a prior study. 27 The matching was on birth year (within 3 years), FY of cohort entry, sex, and follow-up duration (equal to or longer than the case duration). By design, veterans could serve as controls for multiple cases or prior to becoming cases. The index date for each case was the date of OUD diagnosis; for each matched control, the index date was set to the same number of days from cohort entry as the matched case’s index date, consistent with the follow-up duration matching criterion. SDOH Extraction SDOH were identified from both structured and unstructured VHA EHR data. Unstructured notes across 9 services (eg, emergency department [ED], primary care, mental health, social work, or discharge summaries) were processed using a VA-developed multitask NLP pipeline 27 that was fine-tuned from a pretrained RoBERTa 28 (Robustly Optimized Bidirectional Encoder Representations from Transformers) model to identify 8 SDOH factors, with prior validation demonstrating recall of approximately 79% to 95%. Structured data sources included outpatient and inpatient records, diagnosis codes, and clinic location codes, from which we extracted 6 SDOH factors (eTable 2 in Supplement 1 ). The combined model identified 9 distinct SDOH domains: 5 common to both structured and unstructured data (social problems, financial problems, housing insecurity, legal problems, and violence), 3 unique to NLP-extracted data (barriers to care, transitions of care, and food insecurity), and 1 unique to structured data (nonspecific psychosocial problems). Each factor was dichotomized as present if documented in either data source. 27 Exposure and Covariate Assessment To evaluate the association of SDOH with buprenorphine initiation, we assessed each episode’s exposure to the 9 defined SDOH from the index date (first OUD diagnosis) through the end of the follow-up period. Follow-up ended at the buprenorphine initiation date, 365 days after index date, or study end date, whichever occurred first. All covariates were measured during the 2 years prior to the index date and were modeled as baseline variables. Covariates included sociodemographic characteristics (sex, age, race, and ethnicity), comorbidities, mental health disorders, and baseline SDOH. Baseline SDOH were defined as any occurrence of the 9 SDOH domains during this 2-year period, allowing adjustment for preexisting social risks. Race and ethnicity (American Indian or Alaska Native, Asian, Black or African American, Hispanic or non-Hispanic, Native Hawaiian or Other Pacific Islander, White, other [multiracial, not otherwise specified], or unknown) were self-reported by patients and obtained from the CDW. Race and ethnicity data were collected in this study to assess potential disparities in buprenorphine initiation across demographic groups. Additional covariates included service-connected disability percentage (<50%, ≥50%, not service connected), geographic residence (urban, rural), area deprivation index, mental health disorder (depressive disorder, anxiety, posttraumatic stress disorder, bipolar disorder, schizophrenia), SUD (alcohol, cannabis, tobacco, stimulants), history of opioid overdose, chronic noncancer pain, Charlson Comorbidity Index (range: 0-37, with the highest score indicating greater comorbidity burden and mortality risk), ED visits, and inpatient hospitalizations. Individuals with missing values for covariates were retained using explicit unknown or missing categories. Outcomes The primary outcomes were buprenorphine initiation, defined as the first recorded buprenorphine dispensing event in the CDW following the index OUD diagnosis, and the association of SDOH with initiation. The observation window for initiation extended from the index date through 12 months after diagnosis. Statistical Analysis Population characteristics were summarized using descriptive statistics, with categorical variables presented as frequencies with percentages and continuous variables presented as means with SDs. We estimated unadjusted and adjusted 1-year buprenorphine initiation rate across sociodemographic and baseline SDOH subgroups in the base population. Unadjusted prevalence reflected the crude proportion of initiating buprenorphine, while adjusted estimates were calculated from logistic regression, controlling for age, sex, race, and ethnicity. We applied conditional logistic regression to examine the association between SDOH and buprenorphine initiation. Models were adjusted for key potential confounders, including race, ethnicity, service-connected disability percentage, mental health disorder, SUD, chronic noncancer pain, Charlson Comorbidity Index, ED visits, inpatient hospitalizations, and baseline SDOH. Models were fit for each SDOH domain to examine its independent association with buprenorphine initiation. The analytic unit was episode, and matching strata were defined by each case and its corresponding risk set at the case index date. Controls were sampled from the risk set at each case index date and could serve as controls for multiple cases or could later become cases. Conditional logistic regression stratified by matched case-control sets provided valid estimation under this design. SEs adjusted for clustering at the patient level were estimated to account for correlation across multiple OUD episodes per veteran. All statistical tests were 2-sided, with statistical significance defined as P < .05. Analyses were performed in August 2025 using SAS Enterprise Guide version 8.3 (SAS Institute Inc). Results Cohort Characteristics The base population comprised 13 321 new OUD episodes among 12 511 JIVs ( Figure , A); characteristics of these episodes are provided in eTable 3 in Supplement 1 . The nested case-control cohort included 1372 case episodes (involving 93 females [6.8%], 1279 males [93.2%]; mean [SD] age, 40.8 [12.8] years) and 5436 matched control episodes (involving 102 females [1.9%], 5334 males [98.1%]; mean [SD] age, 39.2 [12.1] years) ( Table 1 ). In all, the largest age group was 30 to 39 years, representing 580 (42.3%) of patients with case episodes and 2486 (45.7%) of patients with control episodes. Among the JIVs, 50 identified as American Indian or Alaska Native (0.7%), 31 as Asian (0.5%), 630 as Black or African American (9.3%; hereafter Black), 376 as Hispanic (5.5%), 52 as Native Hawaiian or Other Pacific Islander (0.8%), and 5755 as White (84.5%); 290 (4.3%) had unknown race and ethnicity. Figure. Flowchart of Cohort Construction and Study Timeline for Justice-Involved Veterans (JIVs) With Opioid Use Disorder (OUD). Open in a new tab A, This flowchart illustrates the identification of JIVs with a new OUD diagnosis between fiscal years (FYs) 2015 and 2020, application of eligibility criteria, and formation of the nested case-control cohort. B, This timeline displays the index date (first OUD diagnosis); the 2-year baseline assessment period for covariates and social determinants of health (SDOH); and follow-up through buprenorphine initiation, 365 days, or study end. VJO indicates Veterans Justice Outreach. Table 1. Baseline Characteristics of Case and Control Episodes. Variable OUD episodes, No. (%) Total (N = 6808) Case (n = 1372) Control (n = 5436) Unique patients a 2963 1332 1631 Sex b Female 463 (6.8) 93 (6.8) 102 (1.9) Male 6345 (93.2) 1279 (93.2) 5334 (98.1) Age, y b Mean (SD) 39.5 (12.3) 40.8 (12.8) 39.2 (12.1) 18-29 1214 (17.8) 220 (16.0) 994 (18.3) 30-39 3066 (45.0) 580 (42.3) 2486 (45.7) 40-49 885 (13.0) 197 (14.4) 688 (12.7) 50-59 1078 (15.8) 218 (15.9) 860 (15.8) ≥60 565 (8.3) 157 (11.4) 408 (7.5) Race b , c American Indian or Alaska Native 50 (0.7) 8 (0.6) 42 (0.8) Asian 31 (0.5) 5 (0.4) 26 (0.5) Black or African American 630 (9.3) 131 (9.6) 499 (9.2) Native Hawaiian or Other Pacific Islander 52 (0.8) 13 (1.0) 39 (0.7) White 5755 (84.5) 1145 (83.5) 4610 (84.8) Unknown 290 (4.3) 70 (5.1) 220 (4.1) Hispanic ethnicity b , c Yes 376 (5.5) 72 (5.3) 304 (5.6) No 6317 (92.8) 1277 (93.1) 5040 (92.7) Unknown 115 (1.7) 23 (1.7) 92 (1.7) Service-connected disability percentage b Not service connected 1284 (18.9) 273 (19.9) 1011(18.6) <50 949 (13.9) 192 (14.0) 757 (13.9) ≥50 4575 (67.2) 907 (66.1) 3668 (67.5) ADI b 0-10 120 (1.8) 39 (2.8) 81 (1.5) 11-20 295 (4.3) 62 (4.5) 233 (4.3) 21-30 572 (8.4) 125 (9.1) 447 (8.2) 31-40 696 (10.2) 145 (10.6) 551 (10.1) 41-50 879 (12.9) 159 (11.6) 720 (13.3) 51-60 806 (11.8) 169 (12.3) 637 (11.7) 61-70 876 (12.9) 169 (12.3) 707 (13.0) 71-80 820 (12.0) 174 (12.7) 646 (11.9) 81-90 847 (12.4) 156 (11.4) 691 (12.7) 91-100 632 (9.3) 149 (10.9) 483 (8.9) Missing data 265 (3.9) 25 (1.8) 240 (4.4) Geographic residence b Rural 1964 (28.9) 377 (27.5) 1587 (29.2) Urban 4798 (70.5) 983 (71.7) 3815 (70.2) Unknown 46 (0.7) 12 (0.9) 34 (0.6) Mental health disorder d Any mental health disorder 5295 (77.8) 1053 (76.8) 4242 (78.0) Depressive disorder 3559 (52.3) 680 (49.6) 2879 (53.0) Bipolar disorder 1075 (15.8) 219 (16.0) 856 (15.8) PTSD 3282 (48.2) 646 (47.1) 2636 (48.5) Schizophrenia 309 (4.5) 62 (4.5) 247 (4.5) Other psychotic disorder 417 (6.1) 88 (6.4) 329 (6.1) SUD d Any SUD 4828 (70.9) 979 (71.4) 3849 (70.8) Cannabis 1490 (21.9) 295 (21.5) 1195 (22.0) Tobacco 2730 (40.1) 558 (40.7) 2172 (40.0) Alcohol 2884 (42.4) 571 (41.6) 2313 (42.6) Stimulant 2362 (34.7) 487 (35.5) 1875 (34.5) Other drug disorder 2658 (39.0) 540 (39.4) 2118 (39.0) None 1980 (29.1) 393 (28.6) 1587 (29.2) History of opioid overdose d 280 (4.1) 68 (5.0) 212 (3.9) Chronic noncancer pain d 4777 (70.2) 963 (70.2) 3814 (70.2) CCI d , e 0 5347 (78.5) 1054 (76.8) 4293 (79.0) 1 834 (12.3) 186 (13.6) 648 (11.9) 2 627 (9.2) 132 (9.6) 495 (9.1) No. of ED visits d 0 2820 (41.4) 561 (40.8) 2259 (41.6) 1 1284 (18.9) 271 (19.7) 1013 (18.6) 2 772 (11.3) 170 (12.4) 602 (11.1) ≥3 1932 (28.4) 370 (27.0) 1562 (28.7) No. of inpatient hospitalizations d 0 4295 (63.1) 890 (64.9) 3405 (62.6) 1 1223 (18.0) 243 (17.7) 980 (18.0) 2 508 (7.5) 112 (8.2) 396 (7.3) ≥3 782 (11.5) 127 (9.3) 655 (12.1) Open in a new tab Abbreviations: ADI, Area Deprivation Index; CCI, Charlson Comorbidity Index; ED, emergency department; OUD, opioid use disorder; PTSD, posttraumatic stress disorder; SUD, substance use disorder. a No OUD episodes and with Veterans Justice Outreach enrollment in prior year. b Assessed at time of OUD diagnosis. c Race and ethnicity were self-reported by veterans and obtained from the Veterans Health Administration Corporate Data Warehouse. d Assessed in the 2 years prior to OUD diagnosis, not including OUD diagnosis date. e CCI range: 0-37, with the highest score indicating greater comorbidity burden and mortality risk. Most episodes were observed in White JIVs (case: 1145 [83.5%]; control: 4610 [84.8%]), and most JIVs were living in urban areas (case: 983 [71.7%]; control: 3815 [70.2%]). Because matching was conducted with replacements, some controls were matched to multiple cases. Descriptive tables report episodes rather than unique individuals because some individuals had multiple episodes. This situation may lead to slight imbalances in exact matching variables (eg, sex) when assessed at the episode level. Detailed characteristics are shown in Table 1 . One-Year Buprenorphine Initiation Rates Across Subgroups One-year buprenorphine initiation rates, demographic characteristics, and baseline SDOH are shown in Table 2 . The 1-year buprenorphine initiation was 11.16% (95% CI, 10.63-11.71). Initiation rate was slightly higher for females than males (12.05% [95% CI, 9.92%-14.19%] vs 11.09% [95% CI, 10.55%-11.64%]). Younger veterans with OUD episodes were more likely to initiate buprenorphine, with adjusted rates highest among JIVs aged 30 to 39 years (15.02%; 95% CI, 13.94%-16.11%) and 18 to 29 years (12.14%; 95% CI, 10.56%-13.71%) compared with 7.88% (95% CI, 6.91%-8.85%) of JIVs aged 50 to 59 years and 8.04% (95% CI, 6.91%-9.19%) of those aged 60 years or older. Adjusted buprenorphine initiation among White JIVs was 12.39% (95% CI, 11.73%-13.04%) compared with a mean (SD) of 9.08% (3.69%) across all other racial and ethnic groups, including American Indian or Alaska Native JIVs (values were masked because of a small number of patients) and Black JIVs (6.26%; 95% CI, 5.27%-7.25%). A lower percentage of Hispanic JIVs initiated buprenorphine than non-Hispanic JIVs (8.21% [95% CI, 6.41%-10.01%] compared with 11.35% [95% CI, 10.79%-11.90%]). Across baseline SDOH, initiation was at or slightly lower than overall rates, ranging from 10.02% (95% CI, 8.84%-11.19%) for food insecurity to 11.18% (95% CI, 10.62%-11.74%) for violence. Table 2. Estimated Unadjusted and Adjusted Buprenorphine Initiation Rates Across Subgroups a , b . Variable Buprenorphine initiation No. Unadjusted rate (95% CI), % Adjusted rate (95% CI), % Sex c Male 1379 11.02 (10.25-11.79) 11.09 (10.55-11.64) Female 108 13.40 (10.29-16.51) 12.05 (9.92-14.19) Age, y c 18-29 200 12.94 (11.05-14.83) 12.14 (10.56-13.71) 30-39 649 16.05 (14.75-17.35) 15.02 (13.94-16.11) 40-49 228 10.86 (8.81-12.91) 10.45 (9.17-11.73) 50-59 233 7.36 (5.80-8.92) 7.88 (6.91-8.85) ≥60 177 7.17 (5.04-9.30) 8.04 (6.91-9.19) Race c , d American Indian or Alaska Native NA e NA e NA e Asian NA e NA e NA e Black or African American 145 5.25 (3.52-6.98) 6.26 (5.27-7.25) Native Hawaiian or Other Pacific Islander 13 14.13 (4.67-23.59) 14.45 (7.24-21.65) White 1240 12.92 (12.06-13.78) 12.39 (11.73-13.04) Other or unknown d 73 11.62 (6.95-16.29) 11.64 (8.96-14.32) Hispanic ethnicity c , d Yes 74 8.96 (5.74-12.18) 8.21 (6.41-10.01) No 1389 11.29 (10.52-12.06) 11.35 (10.79-11.90) Unknown 24 12.83 (6.70-18.96) 12.93 (7.78-18.08) Baseline SDOH f Social problems g 979 9.96 (9.02-10.90) 10.18 (9.58-10.78) Financial problems h 1106 10.53 (9.54-11.52) 10.78 (10.18-11.37) Housing insecurity 1170 10.74 (9.77-11.71) 10.98 (10.40-11.57) Violence 1339 11.20 (10.31-12.09) 11.18 (10.62-11.74) Barriers to care 708 9.66 (8.42-10.90) 10.20 (9.49-10.91) Transitions of care 1189 10.67 (9.79-11.55) 10.87 (10.30-11.45) Food insecurity 251 8.81 (6.30-11.32) 10.02 (8.84-11.19) Nonspecific psychosocial problems 1385 11.09 (10.26-11.92) 11.05 (10.51-11.60) Open in a new tab Abbreviations: NA, not applicable; SDOH, social determinants of health. a Medication for opioid use disorder (MOUD) assessed 30 days prior to 365 days after OUD diagnosis. b Initiation rates represent the proportion of episodes initiating buprenorphine within 12 months. Denominators are based on episode-level subgroup counts from eTable 3 in Supplement 1 . c Assessed at time of opioid use disorder (OUD) diagnosis. d Race and ethnicity were self-reported by veterans and obtained from the Veterans Health Administration Corporate Data Warehouse. Other category included individuals who identified as multiracial or whose race was not otherwise specified in administrative records. e Values were masked due to small numbers (<11 patients). f Assessed in the 2 years prior to OUD diagnosis, not including OUD diagnosis date. g Social problems indicate social or familial problems from structured data, with social isolation from natural language processing (NLP)–extracted data. h Financial problems indicate employment or financial problems from structured data, with job or financial insecurity from NLP-extracted data. Associations Between SDOH and Buprenorphine Initiation Table 3 shows that SDOH identified through structured data were generally less prevalent than those identified using the NLP model. For example, social problems at baseline were identified in only 1006 episodes (14.8%) using structured data compared with 4325 episodes (63.5%) using NLP-extracted data and 4505 episodes (66.2%) from combined data sources. The SDOH from structured data contributed only 180 episodes (2.7%) that were not identified by NLP-extracted data. Similar patterns were observed for the other SDOH factors obtained from both sources. Table 3. Summary Statistics of Social Determinants of Health as Covariate and as Exposure. Baseline SDOH domain OUD episodes, No. (%) SDOH as covariate a SDOH as exposure a Structured data NLP-extracted data Combined data Structured data NLP-extracted data Combined data Social problems b 1006 (14.8) 4325 (63.5) 4505 (66.2) 987 (14.5) 6135 (90.1) 6168 (90.6) Financial problems c 2460 (47.5) 4499 (66.1) 5017 (73.7) 4203 (61.7) 6139 (90.2) 6259 (91.9) Housing insecurity 4205 (61.8) 4573 (67.2) 5319 (78.1) 4642 (68.2) 6018 (88.4) 6136 (90.1) Legal problems 6808 (100) 4737 (69.6) 6808 (100) 4669 (68.6) 6266 (92.0) 6357 (93.4) Violence 4715 (69.3) 3271 (48.1) 5618 (82.5) 4610 (67.7) 5493 (80.7) 6112 (89.8) Barriers to care NA 3105 (45.6) 3105 (45.6) NA 5211 (76.5) 5211 (76.5) Transitions of care NA 5355 (78.7) 5355 (78.7) NA 6575 (96.6) 6575 (96.6) Food insecurity NA 1002 (14.7) 1002 (14.7) NA 2518 (37.0) 2518 (37.0) Nonspecific psychosocial problems 6212 (91.3) NA 6212 (91.3) 5009 (73.6) NA 5009 (73.6) Open in a new tab Abbreviations: NA, not applicable; NLP, natural language processing; OUD, opioid use disorder; SDOH, social determinants of health. a As covariate refers to SDOH identified during the baseline (preindex date) lookback period. As exposure refers to SDOH identified during the follow-up (postindex date) period. Percentages represent the proportion of OUD episodes with the specified SDOH identified from structured data, NLP-extracted data, or either source (combined). b Social problems indicate social or familial problems from structured data, with social isolation from NLP-extracted data. c Financial problems indicate employment or financial problems from structured data, with job or financial insecurity from NLP-extracted data. Conditional logistic regression models assessed the association between 8 SDOH domains and 1-year buprenorphine initiation, excluding legal problems since all JIVs had legal involvement ( Table 4 ). Across the models, several SDOH domains were significantly associated with lower odds of initiation. Violence showed an inverse association with buprenorphine initiation, with adjusted odds ratios (AORs) ranging from 0.28 (95% CI, 0.24-0.33) using structured data to 0.36 (95% CI, 0.31-0.42) using NLP-extracted data and 0.31 (95% CI, 0.25-0.38) using the combined model. Food insecurity (AOR, 0.36; 95% CI, 0.30-0.44) and transitions of care (AOR, 0.50; 95% CI, 0.35-0.71) presented in unstructured notes were responsible for lower odds of buprenorphine initiation. Additionally, nonspecific psychosocial problems (AOR, 0.35; 95% CI, 0.30-0.40) were consistently associated with lower initiation. Barriers to care had an AOR of 0.54 (95% CI, 0.45-0.64). Financial problems had AORs of 0.33 (95% CI, 0.28-0.40) using structured data and 0.73 (95% CI, 0.56-0.94) and 0.64 (95% CI, 0.49-0.84) using NLP-extracted and combined data, respectively. Social problems (AOR, 0.58 [95% CI, 0.44-0.79] for structured data; 0.66 [95% CI, 0.50-0.85] for NLP-extracted data) were consistently associated with lower initiation. Across the data sources, housing insecurity had higher odds of lower initiation, but the odds were significant only with structured data (AOR, 0.78; 95% CI, 0.65-0.93). Effect estimates were consistent when patient-level clustering was applied (eTable 4 in Supplement 1 ). Table 4. Association of Social Determinants of Health With Buprenorphine Initiation. SDOH domain AOR (95% CI) a Structured data NLP-extracted data Combined data b Social problems c 0.58 (0.44-0.79) 0.66 (0.50-0.85) 0.65 (0.52-0.87) Financial problems d 0.33 (0.28-0.40) 0.73 (0.56-0.94) 0.64 (0.49-0.84) Housing insecurity 0.78 (0.65-0.93) 0.87 (0.69-1.11) 0.88 (0.68-1.13) Violence 0.28 (0.24-0.33) 0.36 (0.31-0.42) 0.31 (0.25-0.38) Barriers to care NA 0.54 (0.45-0.64) 0.54 (0.45-0.64) Transitions of care NA 0.50 (0.35-0.71) 0.50 (0.35-0.71) Food insecurity NA 0.36 (0.30-0.44) 0.36 (0.30-0.44) Nonspecific psychosocial problems 0.35 (0.30-0.40) NA 0.35 (0.30-0.40) Open in a new tab Abbreviations: AOR, adjusted odds ratio; NA, not applicable; NLP, natural language processing; SDOH, social determinants of health. a We fit 8 separate conditional logistic regression models, 1 for each SDOH domain. In each model, the SDOH of interest was included as the exposure, and models were adjusted for sociodemographic characteristics, psychiatric symptoms, substance use disorder, chronic pain, clinical comorbidities, and all baseline SDOH domains measured prior to the index date. Other postindex SDOH exposures were not mutually adjusted. Covariates used for matching criteria were excluded. b Combined data represent SDOH coded as present if they were identified in either the source structured data or NLP-extracted unstructured data. c Social problems indicate social or familial problems from structured data, with social isolation from NLP-extracted data. d Financial problems indicate employment or financial problems from structured data, with job or financial insecurity from NLP-extracted data. A sensitivity analysis was performed to investigate the implications of baseline SDOH. Removing baseline SDOH from the model did not greatly affect the association of exposure SDOH (eTable 5 in Supplement 1 ). Discussion We examined the initiation rates of buprenorphine among JIVs receiving care in the VHA health care system between FY 2015 and FY 2020 along with the association of SDOH with buprenorphine initiation. SDOH were assessed over 2 periods (baseline and exposure) and across combined structured and unstructured EHR data. We focused on initiation over a 12-month follow-up window to capture longer-term treatment uptake consistent with OUD’s chronic disease course, given that clinical guidance discourages arbitrary treatment duration limits and prior studies demonstrate clinically meaningful benefits of buprenorphine and methadone over 1 year. 29 , 30 , 31 There were 2 key findings in this study. First, the overall rate of buprenorphine initiation in this population remained low, despite the known benefits of buprenorphine for OUD treatment. None of the subpopulations examined had an initiation prevalence greater than 17%. Second, multiple SDOH domains—violence, food insecurity, nonspecific psychosocial problems, and transitions of care—were associated with lower odds of buprenorphine initiation. These findings suggest that social risk factors, when measured comprehensively using both structured and NLP-derived data, may affect treatment uptake. Buprenorphine initiation is higher among veterans receiving treatment at the VHA than in the general population 32 and varies across sociodemographic factors. Female veterans in VHA settings show higher initiation, engagement, and retention than male veterans. 33 Younger age has been associated with a higher initiation rate for treatment uptake. 16 Studies also have shown that racial and ethnic minority veterans, especially Hispanic and Black individuals, have lower rates of MOUD receipt compared with White veterans. 34 , 35 Consistent with prior research, our study found that buprenorphine initiation was more common among younger, female, White, and non-Hispanic individuals. Racial and ethnic disparities persisted, with American Indian or Alaska Native, Black, and Hispanic veterans having significantly lower odds of initiating buprenorphine despite comparable clinical need. The association between SDOH and diverse health outcomes is well-established, 36 , 37 , 38 but the role of SDOH in buprenorphine initiation among JIVs remain understudied. In this study, several SDOH domains demonstrated inverse associations with buprenorphine initiation, particularly violence, financial problems, food insecurity, and nonspecific psychosocial problems, while housing insecurity and social problems showed associations. These findings likely reflect intersecting mechanisms. Prior legal involvement may heighten stigma, surveillance, and distrust of institutions, discouraging engagement in buprenorphine. Financial instability can limit transportation access, appointment adherence, or managing co-occurring needs, decreasing the likelihood of starting buprenorphine. Poverty and low income have been associated with lower MOUD engagement and higher discontinuation rates. 39 , 40 Disrupted family and community ties among JIVs, may further diminish social support, which is critical for treatment uptake and retention. 41 , 42 Instabilities in housing or food can reduce veterans’ motivation to engage in treatment. 43 , 44 These challenges are often more acute for JIVs, who face employment restrictions due to legal histories, leaving many without stable income, housing, or reliable access to food. 45 Additionally, transitions of care such as release from incarceration or hospital discharge represent vulnerable periods associated with care discontinuity and overdose risk. 46 , 47 , 48 Collectively, these studies suggest that justice involvement is an indicator of lower likelihood of treatment initiation, as justice involvement may magnify SDOH and create logistical and systemic barriers. In contrast to most prior studies, we used NLP to extract SDOH from unstructured EHR notes. With the advancement in NLP and increasing awareness of the clinical relevance of SDOH, many studies have focused on using state-of-the-art NLP models to extract such information from raw text. 49 , 50 , 51 NLP substantially improved SDOH capture compared with structured data, revealing far higher prevalence of risk factors such as social problems and financial hardship 52 and reflecting the richness of clinical notes. 53 Structured data showed associations with initiation, possibly because clinicians code SDOH when risks are severe or clinically salient. 51 Combining structured and NLP-derived data offered broader detection, underscoring the value of integrating both sources. Our results showed that multiple SDOH domains (violence, food insecurity, nonspecific psychosocial problems, transitions of care, barriers to care, financial problems, and social problems) significantly hinder buprenorphine initiation among JIVs. Financial problems and food insecurity can divert resources from treatment, while poor transitions of care, such as postincarceration release, may heighten dropout and overdose risk. Although nonspecific psychosocial problems and violence were prevalent in this cohort, their association overlapped with other risks. These findings highlight the need for interventions addressing SDOH to effectively increase buprenorphine uptake. These findings highlight the need to integrate SDOH assessment into routine VHA workflows, particularly for JIVs. The VHA is addressing this through initiatives such as the ACORN (Assessing Circumstances and Offering Resources for Needs) model, which systematically screens across 9 social risk domains, including food, housing, utilities, employment, and legal needs. 54 Building on these efforts, incorporating NLP-enabled tools into EHR systems could help VJO specialists and clinicians more efficiently identify SDOH domains such as barriers to care, food insecurity, and transitions of care, which cannot be identified from structured data. Earlier recognition would support timely referrals and coordinated services, strengthening treatment engagement and promoting buprenorphine initiation among JIVs. Limitations Although we required a 12-month OUD-free period to define a new episode, we could not distinguish true incident OUD from re-engagement after relapse or care gaps. Some individuals may have had prior OUD or treatment outside the observation window; thus, the findings reflect outcomes following a newly observed OUD care episode rather than first-onset OUD. The VHA population may not be representative of veterans not receiving care from the VHA or the general US population, limiting generalizability. While military service may confer unique stressors, the VHA health care system is often ahead of other health care systems in treatment of social-behavioral disorders such as OUD. 21 , 45 Data were restricted to September 30, 2020, to avoid COVID-19 disruptions and may not reflect current practices. Finally, SDOH may be captured incompletely from administrative data; structured data are limited. Clinical notes may not address all SDOH domains and likely may not document less severe instances of SDOH. Conclusions In this case-control study of JIVs, buprenorphine initiation varied across sociodemographic and baseline SDOH subgroups, with higher initiation among females and younger veterans and lower initiation among racial and ethnic minority groups. Using both structured and NLP-extracted data, we demonstrated that multiple SDOH domains—particularly, financial problems, food insecurity, social problems, and transitions of care barriers—were associated with decreased odds of initiating buprenorphine. NLP enhanced detection of underdocumented risks, while combined data models yielded the most robust assessment. The findings emphasize the importance of systematic SDOH screening and interventions to address barriers to buprenorphine initiation. Supplement 1. eTable 1. ICD Codes to Diagnosis OUD eTable 2. ICD and Stop Codes for Structured SDOH eTable 3. Base Population Statistics eTable 4. Associations of Exposure-Period SDOH in Models Clustered by Patient eTable 5. Associations of Exposure-Period SDOH in Models Without the Corresponding Baseline SDOH eReferences jamanetwopen-e268497-s001.pdf (203.5KB, pdf) Supplement 2. Data Sharing Statement jamanetwopen-e268497-s002.pdf (15KB, pdf) References 1. Dydyk AM, Jain NK, Gupta M. Opioid Use Disorder: Evaluation and Management. StatPearls Publishing; 2026. [ PubMed ] [ Google Scholar ] 2. Dowell D, Nataraj N, Rikard M, Park J, Zhang K, Baldwin G. Why have overdose deaths decreased? Widespread fentanyl saturation and decreased drug use among key drivers. Lancet Reg Health Am. 2025;51:101226. doi: 10.1016/j.lana.2025.101226 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 3. Begley MR, Ravindran C, Peltzman T, et al. Veteran drug overdose mortality, 2010-2019. Drug Alcohol Depend. 2022;233:109296. doi: 10.1016/j.drugalcdep.2022.109296 [ DOI ] [ PubMed ] [ Google Scholar ] 4. Holliday R, Forster JE, Desai A, et al. Association of lifetime homelessness and justice involvement with psychiatric symptoms, suicidal ideation, and suicide attempt among post-9/11 veterans. J Psychiatr Res. 2021;144:455-461. doi: 10.1016/j.jpsychires.2021.11.007 [ DOI ] [ PubMed ] [ Google Scholar ] 5. Taylor E, Gray C, Stimmel M, et al. Influences of the criminal justice system on use of medications for opioid use disorder: a qualitative study. BMC Glob Public Health. 2024;2(1):64. doi: 10.1186/s44263-024-00093-y [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 6. McGuire J, Clark S. Veterans Justice Outreach Initiative (VJO). US Department of Veterans Affairs, Veterans Health Administration; 2009. [ Google Scholar ] 7. Smith JW. The Anchorage, Alaska veterans court and recidivism: July 6, 2004-December 31, 2010. June 2012. Accessed September 23, 2025. https://heinonline.org/HOL/Page?handle=hein.journals/allr29&div=7&g_sent=1&casa_token=&collection=journals 8. Tsai J, Finlay A, Flatley B, Kasprow WJ, Clark S. A national study of veterans treatment court participants: who benefits and who recidivates. Adm Policy Ment Health. 2018;45(2):236-244. doi: 10.1007/s10488-017-0816-z [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 9. Tsai J, Pietrzak RH, Szymkowiak D. The problem of veteran homelessness: an update for the new decade. Am J Prev Med. 2021;60(6):774-780. doi: 10.1016/j.amepre.2020.12.012 [ DOI ] [ PubMed ] [ Google Scholar ] 10. Tsai J, Flatley B, Kasprow WJ, Clark S, Finlay A. Diversion of veterans with criminal justice involvement to treatment courts: participant characteristics and outcomes. Psychiatr Serv. 2017;68(4):375-383. doi: 10.1176/appi.ps.201600233 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 11. Bell J, Strang J. Medication treatment of opioid use disorder. Biol Psychiatry. 2020;87(1):82-88. doi: 10.1016/j.biopsych.2019.06.020 [ DOI ] [ PubMed ] [ Google Scholar ] 12. Gonzalez G, Oliveto A, Kosten TR. Combating opiate dependence: a comparison among the available pharmacological options. Expert Opin Pharmacother. 2004;5(4):713-725. doi: 10.1517/14656566.5.4.713 [ DOI ] [ PubMed ] [ Google Scholar ] 13. Adams KK, Waters K, Sobieraj DM. Initiating buprenorphine to treat opioid use disorder without prerequisite withdrawal: an updated systematic review. Addict Sci Clin Pract. 2025;20(1):19. doi: 10.1186/s13722-025-00548-z [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 14. Oliva EM, Maisel NC, Gordon AJ, Harris AHS. Barriers to use of pharmacotherapy for addiction disorders and how to overcome them. Curr Psychiatry Rep. 2011;13(5):374-381. doi: 10.1007/s11920-011-0222-2 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 15. US Department of Veterans Affairs, Office of Research & Development. Substance use disorders. Accessed August 20, 2025. https://www.research.va.gov/topics/sud.cfm 16. Cleary EN, Rollins AL, McGuire AB, Myers LJ, Quinn PD. Buprenorphine discontinuation and utilization of psychosocial services: a national study in the Veterans Health Administration. Addict Sci Clin Pract. 2025;20(1):35. doi: 10.1186/s13722-025-00562-1 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 17. Hser YI, Evans E, Huang D, et al. Long-term outcomes after randomization to buprenorphine/naloxone versus methadone in a multi-site trial. Addiction. 2016;111(4):695-705. doi: 10.1111/add.13238 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 18. Finlay AK, Morse E, Stimmel M, et al. Barriers to medications for opioid use disorder among veterans involved in the legal system: a qualitative study. J Gen Intern Med. 2020;35(9):2529-2536. doi: 10.1007/s11606-020-05944-6 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 19. Braveman P, Gottlieb L. The social determinants of health: it’s time to consider the causes of the causes. Public Health Rep. 2014;129(Suppl 2):19-31. doi: 10.1177/00333549141291S206 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 20. Madras BK, Ahmad NJ, Wen J, Sharfstein JS. Improving access to evidence-based medical treatment for opioid use disorder: strategies to address key barriers within the treatment system. NAM Perspect. 2020;2020. doi: 10.31478/202004b [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 21. Tsai J, Rosenheck RA, Culhane DP, Artiga S. Medicaid expansion: chronically homeless adults will need targeted enrollment and access to a broad range of services. Health Aff (Millwood). 2013;32(9):1552-1559. doi: 10.1377/hlthaff.2013.0228 [ DOI ] [ PubMed ] [ Google Scholar ] 22. Edidin JP, Ganim Z, Hunter SJ, Karnik NS. The mental and physical health of homeless youth: a literature review. Child Psychiatry Hum Dev. 2012;43(3):354-375. doi: 10.1007/s10578-011-0270-1 [ DOI ] [ PubMed ] [ Google Scholar ] 23. Misra-Hebert AD, Santurri L, DeChant R, et al. Understanding the health needs and barriers to seeking health care of veteran students in the community. South Med J. 2015;108(8):488-493. doi: 10.14423/SMJ.0000000000000326 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 24. Dorr D, Bejan CA, Pizzimenti C, Singh S, Storer M, Quinones A. Identifying patients with significant problems related to social determinants of health with natural language processing. Stud Health Technol Inform. 2019;264:1456-1457. [ DOI ] [ PubMed ] [ Google Scholar ] 25. Gundlapalli AV, Carter ME, Palmer M, et al. Using natural language processing on the free text of clinical documents to screen for evidence of homelessness among US veterans. AMIA Annu Symp Proc. 2013;2013:537-546. [ PMC free article ] [ PubMed ] [ Google Scholar ] 26. Price LE, Shea K, Gephart S. The Veterans Affairs’s Corporate Data Warehouse: uses and implications for nursing research and practice. Nurs Adm Q. 2015;39(4):311-318. doi: 10.1097/NAQ.0000000000000118 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 27. Mitra A, Pradhan R, Melamed RD, et al. Associations between natural language processing-enriched social determinants of health and suicide death among US veterans. JAMA Netw Open. 2023;6(3):e233079. doi: 10.1001/jamanetworkopen.2023.3079 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 28. Liu Y, Ott M, Goyal N, et al. RoBERTa: a Robustly Optimized BERT Pretraining Approach. arXiv. Preprint posted online July 26, 2019. doi: 10.48550/arXiv.1907.11692 [ DOI ] 29. Medications for Opioid Use Disorder for Healthcare and Addiction Professionals, Policymakers, Patients, and Families . Updated 2021. Treatment Improvement Protocol (TIP) Series, No. 63. Substance Abuse and Mental Health Services Administration; 2018. Accessed January 29, 2026. https://www.ncbi.nlm.nih.gov/books/NBK574910/ [ PubMed ] 30. Hadland SE, Kimmel SD, Yan S, et al. Buprenorphine treatment duration and adherence among youth and subsequent health outcomes. Pediatrics. 2025;156(6):e2025071147. doi: 10.1542/peds.2025-071147 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 31. Bozinoff N, Men S, Kurdyak P, Selby P, Gomes T. Prescribing characteristics associated with opioid overdose following buprenorphine taper. JAMA Netw Open. 2022;5(9):e2234168. doi: 10.1001/jamanetworkopen.2022.34168 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 32. Manhapra A, Stefanovics E, Rosenheck R. The association of opioid use disorder and homelessness nationally in the Veterans Health Administration. Drug Alcohol Depend. 2021;223:108714. doi: 10.1016/j.drugalcdep.2021.108714 [ DOI ] [ PubMed ] [ Google Scholar ] 33. Simpson TL, McDowell YE, Livingston NA, et al. Women and men veterans’ initiation, engagement, and retention in treatment for substance use disorders: a cross-sectional investigation using VHA administrative data. J Subst Use Addict Treat. 2025;172:209673. doi: 10.1016/j.josat.2025.209673 [ DOI ] [ PubMed ] [ Google Scholar ] 34. Ezell JM, Simek E, Shetty N, et al. A scoping review of the utilization of opioid use treatment, harm reduction, and culturally tailored interventions among racial/ethnic minorities in the United States. Int J Ment Health Addict. 2025;23(6):4612-4653. doi: 10.1007/s11469-024-01373-2 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 35. Finlay AK, Pivovarova E, Yu M, et al. Receipt of medications for opioid use disorders among veterans by race/ethnicity and legal involvement: an observational study of electronic health records. Health Justice. 2025;13(1):28. doi: 10.1186/s40352-025-00336-6 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 36. Norland K, Schaid DJ, Naderian M, Na J, Kullo IJ. Associations of self-reported race, social determinants of health, and polygenic risk with coronary heart disease. J Am Coll Cardiol. 2024;84(22):2157-2166. doi: 10.1016/j.jacc.2024.06.052 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 37. Bullock JL, Hall YN. Social determinants of health and estimation of kidney function. Clin J Am Soc Nephrol. 2023;18(4):424-426. doi: 10.2215/CJN.0000000000000131 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 38. McCarthy ML, Zheng Z, Wilder ME, et al. Latent class analysis to represent social determinant of health risk groups in the Medicaid cohort of the District of Columbia. Med Care. 2021;59(3):251-258. doi: 10.1097/MLR.0000000000001468 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 39. Cook RR, Jaworski EN, Hoffman KA, et al. Treatment initiation, substance use trajectories, and the social determinants of health in persons living with HIV seeking medication for opioid use disorder. Subst Abus. 2023;44(4):301-312. doi: 10.1177/08897077231200745 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 40. Finlay AK, Harris AHS, Timko C, et al. Disparities in access to medications for opioid use disorder in the Veterans Health Administration. J Addict Med. 2021;15(2):143-149. doi: 10.1097/ADM.0000000000000719 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 41. Benville JR, Compton P, Giordano NA, Cheatle MD. Perceived social support in patients with chronic pain with and without opioid use disorder and role of medication for opioid use disorder. Drug Alcohol Depend. 2021;221:108619. doi: 10.1016/j.drugalcdep.2021.108619 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 42. Banta-Green CJ, Williams JR, Sears JM, Floyd AS, Tsui JI, Hoeft TJ. Impact of a jail-based treatment decision-making intervention on post-release initiation of medications for opioid use disorder. Drug Alcohol Depend. 2020;207:107799. doi: 10.1016/j.drugalcdep.2019.107799 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 43. Wu Y, Cheng J, Thorndike AN. Changes in food insecurity among US adults with low income during the COVID-19 pandemic. JAMA Netw Open. 2025;8(2):e2462277. doi: 10.1001/jamanetworkopen.2024.62277 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 44. Yamamoto A, Needleman J, Gelberg L, Kominski G, Shoptaw S, Tsugawa Y. Association between homelessness and opioid overdose and opioid-related hospital admissions/emergency department visits. Soc Sci Med. 2019;242:112585. doi: 10.1016/j.socscimed.2019.112585 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 45. Hoffman KA, Thompson E, Gaeta Gazzola M, et al. “Just fighting for my life to stay alive”: a qualitative investigation of barriers and facilitators to community re-entry among people with opioid use disorder and incarceration histories. Addict Sci Clin Pract. 2023;18(1):16. doi: 10.1186/s13722-023-00377-y [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 46. Filteau MR, Green B, Jones K. Barriers to community treatment for opioid use disorders among rural veterans. J Veterans Stud. 2021;7(3):83-94. doi: 10.21061/jvs.v7i3.262 [ DOI ] [ Google Scholar ] 47. Brunet N, Moore DT, Lendvai Wischik D, Mattocks KM, Rosen MI. Increasing buprenorphine access for veterans with opioid use disorder in rural clinics using telemedicine. Subst Abus. 2022;43(1):39-46. doi: 10.1080/08897077.2020.1728466 [ DOI ] [ PubMed ] [ Google Scholar ] 48. Staton M, Pike E, Tillson M, Lofwall MR. Facilitating factors and barriers for use of medications to treat opioid use disorder (MOUD) among justice-involved individuals in rural Appalachia. J Community Psychol. 2024;52(8):997-1014. doi: 10.1002/jcop.23029 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 49. Ramachandran GK, Fu Y, Han B, et al. Prompt-based extraction of social determinants of health using few-shot learning. arXiv. Preprint posted online June 12, 2023. doi: 10.18653/v1/2023.clinicalnlp-1.41 [ DOI ] 50. Keloth VK, Selek S, Chen Q, et al. Social determinants of health extraction from clinical notes across institutions using large language models. NPJ Digit Med. 2025;8(1):287. doi: 10.1038/s41746-025-01645-8 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 51. Lybarger K, Dobbins NJ, Long R, et al. Leveraging natural language processing to augment structured social determinants of health data in the electronic health record. J Am Med Inform Assoc. 2023;30(8):1389-1397. doi: 10.1093/jamia/ocad073 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 52. Rouillard CJ, Nasser MA, Hu H, Roblin DW. Evaluation of a natural language processing approach to identify social determinants of health in electronic health records in a diverse community cohort. Med Care. 2022;60(3):248-255. doi: 10.1097/MLR.0000000000001683 [ DOI ] [ PubMed ] [ Google Scholar ] 53. DesRoches CM, Wachenheim D, Garcia A, et al. Clinician and patient perspectives on the exchange of sensitive social determinants of health information. JAMA Netw Open. 2024;7(10):e2444376. doi: 10.1001/jamanetworkopen.2024.44376 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 54. Cohen AJ, Russell LE, Elwy AR, et al. Adaptation of a social risk screening and referral initiative across clinical populations, settings, and contexts in the Department of Veterans Affairs Health System. Front Health Serv. 2023;2:958969. doi: 10.3389/frhs.2022.958969 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Associated Data This section collects any data citations, data availability statements, or supplementary materials included in this article. Supplementary Materials Supplement 1. eTable 1. ICD Codes to Diagnosis OUD eTable 2. ICD and Stop Codes for Structured SDOH eTable 3. Base Population Statistics eTable 4. Associations of Exposure-Period SDOH in Models Clustered by Patient eTable 5. Associations of Exposure-Period SDOH in Models Without the Corresponding Baseline SDOH eReferences jamanetwopen-e268497-s001.pdf (203.5KB, pdf) Supplement 2. Data Sharing Statement jamanetwopen-e268497-s002.pdf (15KB, pdf) Articles from JAMA Network Open are provided here courtesy of American Medical Association ACTIONS View on publisher site Cite Collections Permalink PERMALINK Copy RESOURCES Similar articles Cited by other articles Links to NCBI Databases Cite Copy Download .nbib .nbib Format: AMA APA MLA NLM Add to Collections Create a new collection Add to an existing collection Name your collection * Choose a collection Unable to load your collection due to an error Please try again Add Cancel Follow NCBI NCBI on X (formerly known as Twitter) NCBI on Facebook NCBI on LinkedIn NCBI on GitHub NCBI RSS feed Connect with NLM NLM on X (formerly known as Twitter) NLM on Facebook NLM on YouTube National Library of Medicine 8600 Rockville Pike Bethesda, MD 20894 Web Policies FOIA HHS Vulnerability Disclosure Help Accessibility Careers NLM NIH HHS USA.gov Back to Top