Transitions of care between jail-based medications for opioid use disorder and ongoing treatment in the community: A retrospective cohort study

Abstract

Background:

Offering medications for opioid use disorder (MOUD) in carceral settings significantly reduces overdose. However, it is unknown to what extent individuals in jails continue MOUD once they leave incarceration. We aimed to assess the relationship between in-jail MOUD and MOUD continuty in the month following release.

Methods:

We conducted a retrospective cohort study of linked NYC jail-based electronic health records and community Medicaid OUD treatment claims for individuals with OUD discharged from jail between 2011 and 2017. We compared receipt of MOUD within 30 days of release, among those with and without MOUD at release from jail. We tested for effect modification based on MOUD receipt prior to incarceration and assessed factors associated with treatment discontinuation.

Results:

Of 28,298 eligible incarcerations, 52.8% received MOUD at release. 30% of incarcerations with MOUD at release received community-based MOUD within 30 days, compared to 7% of incarcerations without MOUD (Risk Ratio: 2.62 (2.44–2.82)). Most (69%) with MOUD claims prior to incarceration who received in-jail MOUD continued treatment in the community, compared to 9% of those without prior MOUD. Those who received methadone (vs. buprenorphine), were younger, Non-Hispanic Black and with no history of MOUD were less likely to continue MOUD following release.

Conclusions:

MOUD maintenance in jail is strongly associated with MOUD continuity upon release. Still, findings highlight a gap in treatment continuity upon-reentry, especially among those who initiate MOUD in jail. In the wake of worsening overdose deaths and troubling disparities, improving MOUD continuity among this population remains an urgent priority.

1. INTRODUCTION

With the worsening U.S. fentanyl epidemic (O’Donnell et al., 2021) and deepening disparities in overdose among racial and ethnic minoritized groups (James and Jordan, 2018; Kariisa et al., 2023), improving initiation and continuity of evidence-based treatments for opioid use disorder (OUD) is an urgent priority (Blanco and Volkow, 2019). Methadone, buprenorphine and extended-release naltrexone are highly effective medications for opioid use disorder (MOUD) that significantly improve health outcomes, but are largely under-utilized (Krawczyk et al., 2020; Leshner and Mancher, 2019). Individuals involved in the criminal legal system (CLS) are an important target population that has historically had very low access to MOUD, especially buprenorphine and methadone, for which the evidence on overdose reduction is strongest (Binswanger et al., 2020; N; Krawczyk et al., 2020; Lee and Rich, 2012; Santo et al., 2021). Risk of overdose is especially elevated in the weeks following release from incarceration (Binswanger et al., 2013, 2011) as individuals often lose their opioid tolerance while incarcerated, only to then encounter a dangerous street supply of drugs while facing multiple stressors during re-entry (Fox et al., 2014; Joudrey et al., 2019). Offering MOUD during incarceration can mitigate overdose risk upon release (Cates and Brown, 2023; Degenhardt et al., 2014; Green et al., 2018; Krawczyk et al., 2020). In a recent study of New York City (NYC) adults leaving NYC jails, those who received MOUD prior to release had nearly half the overdose death rate within a year compared to those who did not (Lim et al., 2022).

Despite historical resistance to offering MOUD in the CLS, in recent years, multiple U.S. states have passed laws and regulations mandating that MOUD be made available in carceral settings (Weizman et al., 2021). Still, most states do not offer comprehensive access to MOUD in correctional facilities. For example, a recent report found that only 15 states offer comprehensive MOUD services at either intake or release in a considerable number of its jails or prisons (Homans, Allen, and Mazariegos, 2023). Critical to expanding jail and prison-based MOUD programs and maintaining their protective effects is ensuring individuals can continue treatment upon release from incarceration. In a cohort study in Australia, receiving MOUD in the 4 weeks post-release from incarceration reduced the hazard of death by 75% (Degenhardt et al., 2014). Randomized trials and pilot studies of jail-based MOUD programs have pointed to the value of jail-based MOUD in increasing the likelihood of engaging in treatment upon release (Brinkley-Rubinstein et al., 2018; Gordon et al., 2008; Haas et al., 2021; Magura et al., 1993). Still, there remains a dearth of studies that explore whether and to what extent individuals in real-world jail settings continue MOUD once they leave incarceration. More information about the transition between MOUD maintenance during incarceration and community-based treatment, the role of different medications (e.g. methadone vs. buprenorphine), and what factors influence continuity, is key to broadly improving long-term access to MOUD (Ferguson et al., 2019; Pew Charitable Trusts, 2020; Weizman et al., 2021).

To inform these questions, we analyzed retrospective health data from the NYC jails, which has the oldest and one of the most comprehensive correctional MOUD treatment programs in the country (Bellin et al., 2009) – and linked data to New York State (NYS) Medicaid claims for community-based methadone maintenance treatment (MMT) visits and buprenorphine prescriptions. By doing so, we sought to: 1) assess the relationship between receiving jail-based MOUD and continuing MOUD treatment after release; 2) evaluate whether the association between jail-based MOUD and treatment continuity is moderated by MOUD engagement prior to incarceration; and 3) identify individual sociodemographic, health, substance use, and incarceration-related legal characteristics associated with treatment discontinuation.

2. METHODS

2.1. Setting and data sources

This retrospective cohort study focused on incarceration events in NYC jails with admissions and discharges between May 1, 2011, and December 31, 2017. These years were selected as they were the ones for which the most recent and reliable data were available through a centrally-managed electronic health record, that could be linked to external databases of mortality and Medicaid claims data. NYC has one of the largest municipal jail systems in the country: in 2011, the jail had over 85,000 yearly admissions, reduced to 56,000 by 2017 (NYC Office of the Mayor, 2019). We gathered EHR data of NYC Health + Hospitals Correctional Health Services (CHS), which provides all healthcare, including OUD treatment, social services, and reentry services for individuals who are incarcerated at the jails.

Data from CHS were deterministically matched with NYS Medicaid data and probabilistically matched with NYC death certificate data with a matching quality determined to be acceptable (sensitivity = 97%; specificity = 96%). All data were de-identified using a unique patient ID. Details on datasets, data selection processes, and probabilistic matching are available elsewhere (Lim et al., 2022). The study was reviewed and approved by the Institutional Review Boards of the NYC Department of Health and Mental Hygiene (18–106), NYU Langone (i18–00445 and 1811019740), the Weill Cornell Medicine IRB under protocol #1811019740 and the Biomedical Research Alliance of New York (19-PRS-156–419(HHC)), and received certification from the Office for Human Research Protections of the US Department of Health and Human Services.

2.2. Study population

Since the beginning of our study period, all individuals with OUD at the NYC jail were eligible to receive methadone or buprenorphine. Methadone had been available for withdrawal management, induction, or continued maintenance in the jails since the 1980s, while buprenorphine was available for continued maintenance since 2011 with inductions beginning in 2016. However, individuals who had criminal charges that could result in transfer to prison or other custody (e.g., due to felony charges, parole violations or active warrants, or extradition requests) were eligible to receive only acute withdrawal management with MOUD (rather than maintenance), largely due to historical lack of MOUD access in NYS prisons or other jurisdictions. This policy later changed in late 2017, making anyone interested in MOUD maintenance eligible for the program. For this analysis, we limited records to incarceration events of adults 18–65 with a documented OUD diagnosis based on a comprehensive medical intake evaluation and screening that takes place upon admission to the jail. As our study focused on community-treatment continuity, we limited analyses to incarceration events that had a discharge to the community rather than a transfer to prison or another institution, resulting in 70,275 incarceration events.

We excluded incarceration events with naltrexone (n=1), no associated MOUD receipt for either withdrawal management or for maintenance (n=19,645), or in which there was a taper at time of discharge (n=19,504). No MOUD for withdrawal management or maintenance likely indicated patients were not in acute withdrawal at intake and therefore not necessarily good candidates for MOUD treatment, and a MOUD taper at discharge likely indicated ineligibility for maintenance due to legal charge or a short jail stay that precluded initiating a maintenance dose. Finally, we excluded incarceration events after which individuals were either reincarcerated (n=2,776, 9%) or died within 30 days post-release (n=51, 0.2%), as these events would compete with the ability to enter treatment within the 30-day observation period. Our final analytic dataset included 28,298 incarceration events among 15,609 unique individuals. Of these, we compared 14,948 incarceration events of individuals receiving MOUD maintenance in the 3 last days prior to their release to 13,350 incarceration events of individuals that received MOUD at some point during the jail stay but tapered off MOUD by the 3 last days prior to their release (either due to their own volition or because of ineligibility for maintenance based on charge type).

2.3. Community treatment outcomes

Our main treatment outcome was MOUD receipt (either methadone or buprenorphine) within 30 days of discharge from jail, as indicated by having at least one Medicaid claim for a community-MMT visit or a buprenorphine prescription. Medicaid rate codes and National Drug Codes (NDC) from pharmacy records used to define MMT visits and buprenorphine prescriptions were selected based on standards established in previous protocols utilizing NYS Medicaid data (McNeely et al., 2019), and are detailed in Appendix Table 1. Under federal and state law, Medicaid coverage is suspended for most individuals expected to be incarcerated for at least 30 days but is reactivated upon release. In NYS, low-income adults have been eligible for Medicaid since the beginning of the study period, implying that the majority of individuals leaving incarceration in our cohort (most of whom are low income and unemployed) would be eligible for Medicaid coverage and any claims for services received would appear in the Medicaid administrative data. Indeed, prior studies of the NYC jails have shown high rates of Medicaid enrollment among individuals with substance use disorders who were recently jailed (MacDonald et al., 2015).

2.4. Additional covariates

We included additional covariates, recorded in the EHR during incarceration, to examine correlates of community MOUD and to adjust for confounding when examining associations between jail-based MOUD and community MOUD. Sociodemographic data available from the EHR (collected via a combination of self-report and other documentation available at intake) included age at discharge, sex, race and ethnicity, education level, marital status, veteran status, homeless or shelter stay indicated at intake (e.g. unhoused), and pregnancy status. We also included substance use and other health indicators, including self-reported injection drug history, cocaine and alcohol use disorder diagnoses, mental health needs (as indicated based on active jail-based mental health services enrollment), serious mental illness (SMI) diagnoses based on clinical assessments during incarceration, and HIV and hepatitis C (HCV) status based on self-report and/or diagnostic confirmation. Data on the nature of each incarceration included whether the individual had a prior incarceration (going back to 2009, 2 years prior to the first admission record), whether the incarceration lasted longer than 1 month, and whether the incarceration was for a felony charge or for a prior parole violation at jail admission. We also included CHS covariates on OUD treatment history, including MOUD type received during incarceration (methadone vs. buprenorphine) and self-reported and/or confirmed pre-incarceration OUD treatment. Finally, we used Medicaid claims data from the period prior to incarceration to create two variables based on whether an individual had any claim for MOUD (an MMT visit or buprenorphine prescription) in the 6 months or 2 years prior to jail admission.

2.5. Statistical analysis

First, we compared sociodemographic, clinical, incarceration-related, and prior MOUD treatment characteristics between those with “No MOUD at release” and those with “MOUD at release,” using standardized differences (p-values from chi-squared tests would not be informative due to the large sample size). Second, we conducted multivariable robust log linear Poisson regression to estimate risk ratios (RRs) and 95% confidence intervals (CIs) for the relationship between receiving MOUD maintenance while incarcerated and receiving MOUD within 30 days of discharge, accounting for potential confounders of this association. Robust (or modified) Poisson models that use a robust error variance are used widely to estimate risk ratios in cohort studies (Chen et al., 2018; Zou, 2004). We also tested for effect modification of this relationship based on having received MOUD in the six months prior to incarceration by including an interaction term in our regression model. Third, to test whether community treatment continuity differed by MOUD type, we stratified our sample by individuals who received methadone-only vs. buprenorphine-only (both during incarceration and upon release). Fourth, to understand potential risk factors for discontinuation of MOUD among those who received MOUD maintenance in jail, we limited the sample to those who received MOUD at release and used multivariable robust log linear Poisson regression to assess factors associated with likelihood of receiving MOUD within 30 days of discharge. Given the low frequency of missing data (<3.3% on all covariates), all multivariable regressions were conducted as complete case analyses. Given correlation of multiple incarceration events across unique individuals, we included robust standard errors in all of our regressions. All analyses were conducted in Stata 17 (StataCorp, 2021).

2.6. Sensitivity analyses

We conducted multiple sensitivity analyses to ensure findings were robust to analytic decisions. First, we tested for effect modification based on having MOUD claims in the two years rather than six months prior to incarceration. Second, to ensure there wasn’t significant selection bias given our assessment of the outcome—community MOUD—was based on Medicaid claims (e.g. a small proportion of individuals may have had other insurance types), we conducted a sensitivity analysis only including individuals who had any active claim in Medicaid (for any health service/prescription) during the month following release. Finally, we assessed whether findings differed when we extended our main outcome to having a claim for MOUD within three months of incarceration release, rather than one month.

3. RESULTS

3.1. Characteristics of the study sample

Characteristics of the sample are displayed in Table 1. Among our sample of incarceration events for individuals with OUD who had received MOUD earlier in their jail stay, 13,350 (47.2%) did not receive MOUD at release, while 14,948 (52.8%) did receive MOUD at release (94% MMT and 6% buprenorphine). Table 1 includes sociodemographic, clinical, incarceration-related, and prior treatment characteristics of the sample. Overall, the sample was primarily male (82%), 35–54 years old (62%), Hispanic (44%) or Non-Hispanic Black (30%). More than one-third self-reported having injected drugs or having mental health needs, 44% of incarcerations were related to a felony, 49% of incarcerations had a duration of a month or longer, and 84% had at least one prior incarceration in NYC in the 2 years prior to the first incarceration event in the study period. Of the sample, 37% self-reported pre-incarceration OUD treatment. MOUD at release was more common among individuals who were female; who had history of injection, co-occurring cocaine use disorder, or HCV; who were incarcerated <1 month, not facing a felony charge or a parole violation; self-reported prior treatment for OUD, and had a claim for MOUD in the 6 months or 2 years prior to incarceration.

Table 1.

Characteristics of individuals with incarceration events, by receipt of medications for opioid use disorder (MOUD, i.e. methadone or buprenorphine) within 3 days prior to release, NYC Jail 2011–2017 (n=28,298)

	n (%)
	Total (N=28,298)	Did not receive MOUD at release (N=13,350)	Received MOUD at release (N=14,948)	Standardized difference
Age (years)
18–24	1,424 (5.0)	978 (7.3)	446 (3.0)	0.22
25–34	6,400 (22.6)	3,207 (24.0)	3,193 (21.4)
35–44	7,903 (27.9)	3,562 (26.7)	4,341 (29.0)
45–54	9,512 (33.6)	4,204 (31.5)	5,308 (35.5)
55–64	3,059 (10.8)	1,399 (10.5)	1,660 (11.1)
Male (vs. female)	23,325 (82.5)	11,885 (89.1)	11,440 (76.6)	0.33
Education Level				0.02
Less than High School	12,178 (44.5)	5,699 (44.1)	6,479 (44.9)
High School Degree or Equivalent	10,351 (37.8)	4,864 (37.6)	5,487 (38.0)
Some College/Trade School	4,834 (17.7)	2,359 (18.3)	2,475 (17.1)
Race/Ethnicity
Non-Hispanic, Black	8,495 (30.1)	4,111 (30.9)	4,384 (29.4)	0.10
Non-Hispanic, White	6,629 (23.5)	2,858 (21.5)	3,771 (25.3)
Hispanic, Any Race	12,478 (44.2)	6,009 (45.1)	6,469 (43.4)
Non-Hispanic of another Race	646 (2.3)	347 (2.6)	299 (2.0)
Marital Status
Single/Never married	22,162 (78.7)	10,439 (78.5)	11,723 (78.8)	0.04
Married/Partnered	4,571 (16.2)	2,206 (16.6)	2,365 (15.9)
Divorced/Separated/Widowed	1,445 (5.1)	648 (4.9)	797 (5.4)
Unhoused (vs. stable housing) 1	4,766 (16.84%)	2,009 (15.1)	2,757 (18.4)	0.11
Health-Related
History of injection drug use	12,712 (44.9)	5,173 (38.8)	7,539 (50.4)	0.23
Cocaine Use Disorder	9,583 (33.9)	4,032 (30.2)	5,551 (37.1)	0.16
Alcohol Use Disorder	7,007 (24.8)	3,305 (24.8)	3,702 (24.8)	0.00
HIV Diagnosis	2,240 (7.9)	947 (7.1)	1,293 (8.7)	0.06
HCV Diagnosis	9,952 (35.2)	3,719 (27.9)	6,233 (41.7)	0.29
Serious Mental Illness	1,503 (5.3)	742 (5.6)	761 (5.1)	0.02
Any Mental Health Needs	9,994 (35.3)	4,829 (36.2)	5,165 (34.6)	0.04
Incarceration-Related
Prior Incarceration (since 2009)	23,848 (84.3)	10,796 (80.9)	13,052 (87.3)	0.19
Incarceration lasted >1 month	13,771 (48.7)	8,120 (60.8)	5,651 (37.8)	0.47
Incarceration related to felony charge	12,408 (43.9)	8,272 (62.0)	4,136 (27.7)	0.76
Incarceration related to parole violation	2,087 (7.4)	1631 (12.2)	456 (3.1)	0.35
Prior Treatment-Related
Self-reported opioid use treatment prior to incarceration	10,528 (37.2)	2,926 (21.9)	7,602 (50.9)	0.63
MOUD claim in two years prior to incarceration	10,238 (36.2)	3,393 (25.4)	6,845 (45.8)	0.43
MOUD claim in six months prior to incarceration	7,251 (25.6)	2,204 (16.5)	5,047 (33.8)	0.40

¹Unhoused: homeless or transient shelter stay. Reference group is housed, staying with family/friends

3.2. Association between jail-based MOUD and community MOUD engagement

Of those with incarceration events in our sample, 5,373 (19%) received methadone or buprenorphine treatment within one month of discharge (Table 2). Of these, 4,289 (80%) had a claim for a community MMT visit and 1,142 (21%) for a buprenorphine prescription (N=58 had both an MMT visit and buprenorphine prescription).

Table 2:

Association between receipt of jail-based MOUD at release and community MOUD engagement within one month post-release, stratified by whether individual had an MOUD claim within 6 months prior to incarceration

Population	No MOUD at release		Any MOUD at release		Comparison
	Total	Received community MOUD within one month post-release	Total	Received community MOUD within one month post-release
	N	N (%)	N	N (%)	aRR (95% CI)1
All incarcerations	13,350	965 (7.2%)	14,948	4,408 (29.5%)	2.62 (2.44–2.82)2
Incarcerations without MOUD claim in six months prior	11,154	272 (2.4%)	9,907	935 (9.4%)	5.47 (4.66–6.42)3
Incarcerations with MOUD claim in six months prior	2,196	693 (31.6%)	5,041	3,473 (68.9%)	2.01 (1.87–2.15)3

Notes:

Adjusted risk ratios (aRR) were calculated using robust log linear Poisson regression. All models account for clustering by person and complete-case analysis and exclude individuals who died within 30 days (n=51) or were re-incarcerated within 30 days (n=2,776)

¹All multivariable models account for age, sex, education level, race/ethnicity, marital status, history of injection drug use, serious mental illness, cocaine use disorder, alcohol use disorder, HIV diagnosis, HCV diagnosis, unhoused/unstably housed, mental health needs, prior incarceration (since 2009), incarceration lasted >1 month, incarceration related to felony charge, incarceration related to parole violation, self-reported opioid use disorder treatment prior to incarceration, unhoused/unstably housed, mental health needs.

²Multivariable all incarcerations model additionally adjusts for prior MOUD treatment within 6 months of incarceration.

³Interaction coefficient between past-6-months MOUD and MOUD was significant at p<0.05 level (0.56 p=<0.001)

As shown in Table 2, among those with MOUD at release, 30% had a subsequent community-based MOUD claim within 30 days, while only 7% of those who no longer had MOUD at release had a community-based MOUD claim within 30 days. After adjusting for potential confounders as well as prior treatment history, those with MOUD were 2.62 times as likely to enter MOUD within one month (95% CI: 2.44–2.82).

A majority of those who had in-jail MOUD and a prior MOUD claim continued MOUD in the community (69%), while only 9% of those with new initiation of MOUD in-jail continued MOUD in the community (Table 2). In both groups, having MOUD at release was associated with a significantly increased probability of entering MOUD within one month of discharge, with an even larger association for those without prior MOUD (Adjusted Risk Ratio (aRR): 5.47 (95% CI: 4.66–6.42)) than for those with prior MOUD (aRR: 2.01 (95% CI: 1.87–2.15)). Prior MOUD use therefore significantly modified the effect of MOUD at release (interaction term p<0.001). When we stratified records by receipt of methadone vs. buprenorphine, a strong association between in-jail MOUD and MOUD continuity within one month of discharge persisted. Methadone at release had a particularly strong association with continuing community-based MMT relative to those not receiving MMT at release (aRR: 3.98 (95% CI: 3.60–4.40), though buprenorphine at release was still strongly associated with continuing community-based buprenorphine (aRR: 2.36 (95% CI: 2.05–2.72)), and had a higher prevalence of treatment continuity in the community (40.86%) relative to methadone (26.71%) (Table 3).

Table 3:

Association between receipt of jail-based MOUD and community MOUD engagement one month post-release; analyzed separately for methadone and buprenorphine

3A. Methadone maintenance treatment (MMT)
Population	No MOUD1 at release		Any MMT at release		Comparison
	Total	Received community MMT w\hin one month post-release	Total	Received community MMT within one month post-release
	N	N (%)	N	N (%)	aRR (95% CI)1
All incarcerations 2	13,350	499 (3.74%)	13,974	3,733 (26.71%)	3.98 (3.60–4.40)
Incarcerations without MOUD claim in six months prior	11,154	170 (1.42%)	9,947	913 (9.18%)	9.65 (7.96–11.69)
Incarcerations with MOUD claim in six months prior	1,364	329 (24.12%)	4,027	2,820 (70.03%)	2.56 (2.32–2.83)
3B. Buprenorphine treatment
Population	No MOUD1 at release		Any Buprenorphine at release		Comparison
	Total	Received community bup. within one month post-release	Total	Received community bup. within one month post-release
	N	N (%)	N	N (%)	aRR (95% CI)1
All incarcerations 2	13,350	478 (3.58%)	974	398 (40.86%)	2.36 (2.05–2.72)
Incarcerations without MOUD claim in six months prior	12,404	173 (1.39%)	484	49 (10.12%)	6.56 (4.64–9.27)
Incarcerations with MOUD claim in six months prior	946	305 (32.24%)	490	349 (71.22%)	1.84 (1.64–2.07)

Notes:

Adjusted risk ratios (aRR) were calculated using robust log linear Poisson regression. All models account for clustering by person and complete-case analysis and exclude individuals who died within 30 days (n=51) or were re-incarcerated within 30 days (n=2,776)

¹All multivariable models account for age, sex, education level, race/ethnicity, marital status, history of injection drug use, serious mental illness, cocaine use disorder, alcohol use disorder, HIV diagnosis, HCV diagnosis, unhoused/unstably housed, mental health needs, prior incarceration (since 2009), incarceration lasted >1 month, incarceration related to felony charge, incarceration related to parole violation, self-reported opioid use disorder treatment prior to incarceration, unhoused/unstably housed, mental health needs.

²Multivariable all incarcerations model additionally adjusts for prior MOUD treatment within 6 months of incarceration.

3.3. Correlates of community MOUD continuity

When we limited the sample to those who received MOUD at release, multiple characteristics were independently associated with subsequent community MOUD (Figure 1, with detailed estimates provided in Appendix Table 2). These included: having an MOUD treatment claim in the six months prior to the index incarceration (aRR: 6.76 (95% CI: 6.30–7.27), older age (e.g., 55–64 vs. 18–24, aRR: 1.37 (95% CI: 1.13–1.64)), and being male (aRR: 1.10 (95% CI: 1.04–1.17)). Non-Hispanic Black individuals were less likely to continue MOUD compared to Non-Hispanic White individuals (aRR: 0.89 (95% CI: 0.83–0.95)), while Hispanic individuals and those of another race/ethnicity had a similar likelihood to continue MOUD compared to Non-Hispanic White individuals.

Figure 1:

Factors associated with MOUD engagement at one-month post-release among people who received MOUD at release based on multivariable analysis (N=14,948)

Note: Adjusted and unadjusted estimates and confidence intervals are available in Appendix Table 2

Multiple characteristics of the incarceration itself were associated with MOUD continuity, including having had a prior incarceration (aRR: 1.19 (95% CI: 1.10–1.27)), felony charge at the index incarceration (aRR: 1.16 (95% CI: 1.10–1.22)) or parole violation at the index incarceration (aRR: 1.30 (95% CI: 1.17–1.43)). Those who received buprenorphine in jail had a slightly greater likelihood of continuing MOUD compared to those who received methadone (aRR: 1.09 (95% CI: 1.02–1.18)), as did those who reported being in treatment prior to incarceration compared to those who did not (aRR: 1.16 (95% CI: 1.10–1.24)). None of the health-related factors we assessed were associated with treatment continuation, nor was housing stability or marital status.

3.4. Sensitivity analyses

Results of sensitivity analyses can be found in Appendix Tables 3–5. First, when we tested whether the association between in-jail MOUD and community MOUD was moderated by prior two-year MOUD treatment (vs. six-months only), findings did not qualitatively change, and the interaction between prior MOUD and MOUD at release was still highly significant (Appendix Table 3). Results also did not qualitatively change when we limited the dataset to individuals with active Medicaid claims upon release (Appendix Table 4), indicating findings were not biased by differential access to Medicaid among the MOUD vs. non-MOUD groups. Finally, the impact of in-jail MOUD on community-based MOUD decreased slightly but was still strong (aRR: 2.11 (2.00–2.23)) when we extended the main outcome to having a claim for MOUD within three months of incarceration discharge, rather than one month (Appendix Table 5).

4. DISCUSSION

The current study provides a unique opportunity to observe real-world patterns and correlates of MOUD continuity after release from incarceration that is rarely available in US population-based samples. Particularly, our study sought to better understand 1) the relationship between receiving jail-based MOUD and continuing MOUD treatment after release, 2) whether this association was moderated by MOUD engagement prior to incarceration, and 3) what sociodemographic, health, substance use, and incarceration-related legal characteristics were associated with treatment discontinuation.

In regards to the first aim, findings corroborate prior clinical trial and pilot study results (Brinkley-Rubinstein et al., 2018; Haas et al., 2021; Magura et al., 1993; McKenzie et al., 2012) indicating that jail-based MOUD is strongly associated with MOUD continuity upon release. In regards to the second aim, we found that this effect was even larger for those initiating MOUD without prior treatment, pointing to the importance of offering jail-based MOUD initiation in addition to maintenance. Continuity of MOUD in the post-incarceration period is critical to sustaining improved health outcomes offered by these treatments, including reductions in overdose risk, and decreasing healthcare costs (Jalali et al., 2022; Krawczyk et al., 2020). Another recent analysis from our team using these data found that community utilization of MOUD significantly mediated the effect of in-jail MOUD on mortality outcomes within a year of jail release.(Sungwoo Lim, n.d.)

Despite improved MOUD linkage outcomes among people receiving MOUD in jail, overall rates of transition to community treatment among this group were low, below 30%. The proportion of individuals entering community MOUD was not substantially larger by three months post-release (35%), indicating MOUD engagement rates are unlikely to rise over time. Community re-entry is known to be a high-risk time for individuals leaving jails, and these low rates of community linkage to care are in line with those reported for other serious conditions such as HIV (Loeliger et al., 2018). Still, this low rate is concerning given NYC jails’ MOUD program is one of the most comprehensive in the country, offering both initiation and maintenance for both methadone and buprenorphine. Most U.S. jails do not routinely offer access to MOUD or do so on a limited basis, such as for those already in MOUD treatment or special populations such as pregnant women, and often prioritize naltrexone over agonist medications (Nunn et al., 2009; Sufrin et al., 2022; Weizman et al., 2021). In addition, few other cities and counties have access to as many community MOUD providers as are available in NYC (Abraham et al., 2018; Haffajee et al., 2019). Lack of community MOUD availability is often cited by jail leaders in other jurisdictions as a challenge for offering and linking incarcerated individuals to MOUD (Krawczyk et al., 2022).

The low rate of MOUD continuity observed in our data was overwhelmingly driven by low enrollment after release by people initiated on MOUD in jail (9%), whilst a majority of those who had treatment prior to incarceration and received MOUD in jail did go on to continue treatment in the community (69%). This was despite the fact that receiving MOUD in jail at release, regardless of prior treatment, was still a strong predictor of MOUD following discharge. Therefore, better incentives and opportunities for engagement in care post-release are needed for individuals who initiate MOUD during incarceration.

In regards to our third aim of understanding individual characteristics associated with greater treatment discontinuation following release, we found that Non-Hispanic Black individuals were less likely to continue MOUD. Prior studies have had mixed findings on racial differences in MOUD retention; a meta-analysis of retention in MOUD found no differences across race/ethnicity groups, while a recent study found lower MOUD retention among Non-Hispanic Black individuals receiving buprenorphine (Hochheimer and Unick, 2022). Only one other study to our knowledge specifically looked at differences in methadone continuity among those leaving a jail setting in Connecticut, and found no significant differences across race groups (Haas et al., 2021). Our study highlights a need for greater efforts to understand and address structural barriers to care and treatment continuity, especially among minoritized groups, for whom racist drug war and policing practices (Netherland and H., 2017) and a long history of discrimination in substance treatment and other healthcare settings, have resulted in disproportionate criminalization, marginalization, and, increasingly, worse overdose outcomes (James and Jordan, 2018; Khatri et al., 2023; Kunins, 2020; Netherland and H., 2017). The same was true for younger adults who, regardless of pre-incarceration history of MOUD, had significantly worse MOUD continuity post-release than older persons, and for whom MOUD engagement is known to be particularly challenging (Buchholz et al., 2023; Hogue et al., 2023; Proctor et al., 2015; Samples et al., 2018). Interestingly, while women were more likely to have MOUD at release than men (possibly due to lower presence of severe charge types that may have led to ineligibility for maintenance treatment), men were more likely to continue treatment in the community. More work is needed to address particular challenges women face at re-entry in engaging in services and continuing treatment (Olphen et al., 2009).

Challenges for continuing MOUD after incarceration, particularly among those initiating care in jail, younger individuals, and individuals from minoritized groups, may reflect the chaotic nature of jail-reentry (e.g., unplanned discharges, little opportunity to discuss discharge plans) and its synergistic effects with larger barriers to MOUD utilization in the community. A recent national analysis estimated that only 13% of those with OUD receive MOUD (Krawczyk et al., 2022). Those with criminal-legal involvement additionally experience a range of social, medical, and structural barriers to treatment (Aronowitz and Laurent, 2016; Fiscella et al., 2018; Krawczyk et al., 2017). Expanding the availability of MOUD in low-threshold and flexible settings that can address needs of this population during the chaotic period of re-entry might be particularly important. Examples include mobile van settings (Krawczyk et al., 2019), telehealth buprenorphine clinics (Duncan et al., 2021; Tofighi et al., 2022), which might be particularly attractive in more rural areas (Weintraub et al., 2021), and the Transitions Clinic Network (Howell et al., 2021). Recent trials have also shown the potential promise of injectable forms of buprenorphine in facilitating treatment continuity following discharge (Berk et al., 2022; Cheng et al., 2022; Lee et al., 2021), but uptake of these treatments in carceral settings remains slow.

Those in our sample who received buprenorphine (vs. methadone), though a smaller sample due to the historical limits on buprenorphine induction in the jail, were more likely to continue MOUD, which is consistent with findings from a prior clinical trial in NYC (Magura et al., 2009). Improved continuity of buprenorphine may relate in part to greater facility in accessing buprenorphine via office-based prescribing, compared to methadone, which can only be dispensed at highly-regulated and often burdensome opioid treatment programs (OTPs). Continued efforts to maintain COVID-19-era flexibilities for methadone take-homes and initiation of buprenorphine via telehealth, which have facilitated initiation and retention in MOUD, is key (Krawczyk et al., 2023b, 2023c). Expanding methadone to other treatment settings, including mobile vans (which have recently launched in NY) (Gibbons et al., 2022) and office-based practices, as is the norm in non-US settings and which has been proposed by recent legislation (Krawczyk et al., 2023a), may also improve MMT access and continuity. Importantly, ensuring connections post-release to harm reduction programs is critical for ensuring reduced overdose among this population. In addition to expanding the number of settings where evidence-based services can be obtained, ensuring training of providers in person-centered, non-stigmatizing care for people with a history of incarceration is critical to improving health outcomes among this group. Finally, expanding alternatives to incarceration remains a vital goal as incarceration itself is known to highly destabilize multiple factors that promote health and well-being such as employment, housing, and family connections (Wildeman and Wang, 2017).

Lastly, our study found that even in the NYC jail, which has a robust MOUD program, only half of OUD patients received MOUD at the time of release. This is likely explained by historical limitations on eligibility for MOUD based on charges throughout our study period (reflecting the historical unavailability of MOUD in NYS prisons), a policy that was later changed. Indeed, those with a felony charge and parole violation or with longer incarceration periods in our sample were least likely to have MOUD at release. This points to the importance of making MOUD available in all prison systems to ensure care continuity and access to evidence-based treatment regardless of charge or disposition. Unfortunately, most prison systems across the U.S. still do not offer MOUD (“Data - Jail and Prison Opioid Project,” 2023). Those not receiving MOUD maintenance were also more likely to be younger, male, not having injection drug use history, and not having co-occurring disorders such as cocaine use, HIV or HCV, potentially indicating lower severity of OUD for which incarcerated individuals or health care providers in jails may have decided against MOUD. Greater efforts to understand decision making around in-jail MOUD, including the role of stigma, mistrust, and preferences among patients, and how barriers encountered in the post release era (e.g. resistance from parole) (Vail et al., 2021) may influence treatment decisions, are important for informing continuous efforts to offer evidence-based treatment in these settings and upon release, particularly as efforts grow across the U.S. to increase access to MOUD in both jail and prison settings (Weizman et al., 2021).

This study is subject to multiple limitations. First, all our analyses were based on administrative and health records not collected for research, and may therefore omit important variables and covariates that limit our ability to draw some conclusions such as on the severity of OUD or interest in treatment. Our outcomes data on community MOUD was only limited to observable Medicaid claims, and individuals who received MOUD covered by other insurance types or private pay may have been missed by this analysis. However, given that the great majority of individuals leaving NYC jail are Medicaid-eligible, and that our findings did not differ on sensitivity analyses restricting to those actively enrolled in Medicaid, this limitation is not likely to have introduced significant bias. Second, our outcome focused on whether individuals received any claim for MOUD, and we did not measure retention or differentiate between individuals who continued or discontinued MOUD after the first claim. Third, our analytic sample excluded individuals who were receiving MOUD for withdrawal management only at the time of discharge (as opposed to MOUD maintenance). Thus, we are not able to draw conclusions about the impact of MOUD used for withdrawal management (e.g., detox) on subsequent treatment outcomes. We also were not able to assess the impact of distinct length of MOUD treatment in jail on subsequent continuity outcomes. Lastly, generalizability may be limited given our sample was uniquely situated in the NYC jails, which have a robust MOUD infrastructure and a wide network of community MOUD providers, and during the years 2011–2017, prior to the overwhelming shift to a fentanyl street drug supply. More work is needed to understand more recent impact of fentanyl on MOUD initiation and continuation, including the role of MOUD type, and to study other U.S. settings which have lower access to MOUD in correctional facilities and where community-based MOUD providers are not as readily present.

CONCLUSIONS

MOUD maintenance while in jail is strongly associated with going on to receive MOUD in the community upon release, regardless of prior history of MOUD. MOUD access in jail mitigated discontinuity for those with previous treatment experience and resulted in better treatment continuity among those initiated on MOUD during incarceration. Still, findings highlight a continued gap in MOUD engagement upon re-entry even among those who do receive MOUD in-jail, especially among groups already experiencing barriers to community treatment: persons who are younger, Non-Hispanic Black, and newly initiated on MOUD while incarcerated. Findings support the dire need to invest in re-entry programs and low-barrier treatment models to encourage MOUD continuity among the highest-risk groups, while emphasizing the need for alternative pathways to engage individuals in treatment and harm reduction services without the criminal justice system as an intermediary. In the wake of worsening overdose deaths and troubling disparities, improving access to evidence-based care among this population must be an urgent public health and policy priority.

Highlights.

• Receiving MOUD in jail increases likelihood of MOUD continuity following release

• Less than one third of those receiving in jail MOUD continue care following release

• Receiving MOUD prior to incarceration is highly predictive of MOUD post-release

• Certain groups remain at high risk for MOUD discontinuity upon jail release

• Efforts are needed to improve MOUD care continuity upon re-entry from jail

Appendix

Appendix Table 1:

Rate codes and definitions used to define opioid treatment program visits and buprenorphine prescriptions in Medicaid data

	NY Medicaid NDC/Rate Codes
Buprenorphine prescription claim	00054017613; 00054017713; 00054018813; 00054018913; 00093537856; 00093537956; 00093572056; 00093572156; 00228315303; 00228315403; 00228315473; 00228315503; 00228315573; 00228315603; 00378092393; 00378092493; 00406192303; 00406192403; 12496120201; 12496120203; 12496120401; 12496120403; 12496120801; 12496120803; 12496121201; 12496121203; 12496127802; 12496128302; 12496130602; 12496131002; 16590066630; 35356000407; 35356000430; 35356055530; 35356055630; 42291017430; 42291017530; 43063018407; 43063018430; 43063066706; 49999039507; 49999039515; 49999039530; 49999063830; 49999063930; 50383028793; 50383029493; 50383092493; 50383093093; 51927354800; 52959030430; 52959074930; 54123011430; 54123091430; 54123092930; 54123095730; 54123098630; 54569549600; 54569573900; 54569573901; 54569573902; 54569639900; 54569640800; 54868570700; 54868570701; 54868570702; 54868570703; 54868570704; 54868575000; 55045378403; 55700014730; 55700018430; 55700030230; 55700030330; 58284010014; 59385001201; 59385001230; 59385001401; 59385001430; 59385001601; 59385001630; 60429058630; 60429058730; 63481016101; 63481016160; 63481020701; 63481020760; 63481034801; 63481034860; 63481051901; 63481051960; 63481068501; 63481068560; 63481082001; 63481082060; 63481095201; 63481095260; 63629402801; 63629403401; 63629403402; 63629403403; 63874108403; 63874108503; 63874117303; 65162041503; 65162041603; 66336001630; 68071138003; 68071151003; 68258299903; 68308020230; 68308020830
Methadone maintenance treatment rate code	1116; 1120; 1130; 1131; 1134; 1135; 1531; 1543; 1564; 1567; 1618; 1671; 2531; 2532; 2533; 2534; 2872; 2973; 1471; 1555

Appendix Table 2:

Factors associated with MOUD engagement at one-month post-release among people who received MOUD at release (N=14,948)

	Unadjusted Risk Ratio (95% CI)	Adjusted Risk Ratio (95% CI)
Age
18–24	1.00	1.00
25–34	1.49 (1.18, 1.89)	1.30 (1.08, 1.55)
35–44	1.66 (1.31, 2.10)	1.30 (1.09, 1.55)
45–54	1.62 (1.28, 2.05)	1.31 (1.09, 1.57)
55–64	1.79 (1.39, 2.29)	1.37 (1.13, 1.64)
Sex
Female	1.00	1.00
Male	1.26 (1.15, 1.37)	1.10 (1.04, 1.17)
Educational Level
Less than High School	1.00	1.00
High School Degree or Equivalent	0.98 (0.92, 1.05)	1.00 (0.95, 1.04)
Some College/Trade School	0.91 (0.83, 1.00)	0.98 (0.92, 1.04)
Race/Ethnicity
Non-Hispanic, White	1.00	1.00
Non-Hispanic, Black	0.82 (0.75, 0.90)	0.89 (0.83, 0.95)
Hispanic, Any Race	1.09 (1.01, 1.18)	0.98 (0.93, 1.04)
Non-Hispanic, All other races than Black or White	1.05 (0.87, 1.28)	1.08 (0.94, 1.23)
Marital Status
Single, Never Married	1.00	1.00
Married/Partnered	1.04 (0.97, 1.13)	0.99 (0.94, 1.05)
Divorced/Separated/Widowed	1.08 (0.95, 1.22)	0.98 (0.90, 1.07)
Unhoused or Unstably Housed	1.11 (1.04, 1.19)	1.04 (0.99, 1.09)
Health-Related
Injection Drug Use	1.05 (0.99, 1.12)	0.97 (0.92, 1.01)
Cocaine Use Disorder	0.90 (0.85, 0.95)	1.00 (0.96, 1.05)
Alcohol Use Disorder	0.82 (0.76, 0.88)	0.95 (0.91, 1.00)
HIV	1.10 (0.99, 1.22)	1.01 (0.94, 1.08)
HCV	1.13 (1.06, 1.21)	0.98 (0.94, 1.03)
Serious Mental Illness	1.12 (1.00, 1.26)	1.02 (0.93, 1.11)
Mental Health Needs	1.16 (1.10, 1.23)	1.04 (0.99, 1.09)
Incarceration-Related
Prior Incarceration (since 2009)	1.46 (1.33, 1.61)	1.19 (1.10, 1.27)
Incarceration lasted >1 month	1.15 (1.09, 1.21)	0.97 (0.93, 1.02)
Incarceration related to felony charge	1.31 (1.24, 1.39)	1.16 (1.10, 1.22)
Incarceration related to parole violation	1.81 (1.65, 2.00)	1.30 (1.17, 1.43)
Received buprenorphine at release (vs. methadone)	1.62 (1.49, 1.76)	1.09 (1.02, 1.18)
Self-reported opioid use treatment prior to incarceration	1.99 (1.87, 2.11)	1.16 (1.10, 1.24)
MOUD claim if six months prior to incarceration	7.32 (6.83, 7.83)	6.76 (6.29, 7.26)

Appendix Table 3:

Association between receipt of jail-based MOUD and community MOUD engagement within one month post-release, stratified by whether individual had MOUD claim within two years prior to incarceration

Population	No MOUD at release		Any MOUD at release		Comparison
	Total	Received community MOUD within one month post-release	Total	Received community MOUD within one month post-release
	N	N (%)	N	N (%)	aRR (95% CI)1
Incarcerations without MOUD claim in two years prior	8,113	166 (1.67%)	9,965	585 (7.21%)	6.97 (5.68–8.55) 2
Incarcerations with MOUD claim in two years prior	3,385	799 (23.60%)	6,835	3,823 (55.93%)	2.20 (2.05–2.36) 2

Notes:

Adjusted risk ratios (aRR) were calculated using robust log linear Poisson regression. All models account for clustering by person and complete-case analysis and exclude individuals who died within 30 days (n=51) or were re-incarcerated within 30 days (n=2,776)

¹All multivariable models account for age, sex, education level, race/ethnicity, marital status, history of injection drug use, serious mental illness, cocaine use disorder, alcohol use disorder, HIV diagnosis, HCV diagnosis, unhoused/unstably housed, mental health needs, prior incarceration (since 2009), incarceration lasted >1 month, incarceration related to felony charge, incarceration related to parole violation, self-reported opioid use disorder treatment prior to incarceration, unhoused/unstably housed, mental health needs.

²Interaction coefficient between past-6-months MOUD and MOUD at release was significant at the p<0.05 level (0.52 p=<0.001)

Appendix Table 4:

Association between receipt of jail-based MOUD and community MOUD engagement one month post-release, limited to individuals with active Medicaid claim within one month of release

Population	No MOUD1 at release		Any MOUD at release		Comparison
	Total	Received community MOUD within one month post-release	Total	Received community MOUD within one month post-release
	N	N (%)	N	N (%)	aRR (95% CI)1
All incarcerations 2	8,744	961 (10.99%)	10,848	4,402 (34.45%)	2.59 (2.42–2.78)
Incarcerations without MOUD claim in six months prior	6,571	272 (4.14%)	5,834	934 (16.01%)	5.13 (4.39–5.99)
Incarcerations with MOUD claim in six months prior	2,173	689 (31.71%)	5,014	3,468 (69.17%)	2.01 (1.87–2.15)

Notes:

Adjusted risk ratios (aRR) were calculated using robust log linear Poisson regression. All models account for clustering by person and complete-case analysis and exclude individuals who died within 30 days (n=51) or were re-incarcerated within 30 days (n=2,776)

¹All multivariable models account for age, sex, education level, race/ethnicity, marital status, history of injection drug use, serious mental illness, cocaine use disorder, alcohol use disorder, HIV diagnosis, HCV diagnosis, unhoused/unstably housed, mental health needs, prior incarceration (since 2009), incarceration lasted >1 month, incarceration related to felony charge, incarceration related to parole violation, self-reported opioid use disorder treatment prior to incarceration, unhoused/unstably housed, mental health needs.

²Multivariable all incarcerations model additionally adjusts for prior MOUD treatment within 6 months of incarceration.

Appendix Table 5:

Association between receipt of jail-based MOUD and community MOUD engagement within 3 months post-release, stratified by whether individual had an MOUD claim within 6 months prior to incarceration

Population	No MOUD1 at release		Any MOUD at release		Comparison
	Total	Received community MOUD within three months post-release	Total	Received community MOUD within three months post-release
	N	N (%)	N	N (%)	aRR (95% CI)1
All incarcerations 2	13,350	1,513 (11.33%)	14,948	5,150 (34.45%)	2.11 (2.00–2.23)
Incarcerations without MOUD claim in six months prior	11,154	561 (5.03%)	9,907	1,316 (13.28%)	3.37 (3.00–3.78)
Incarcerations with MOUD claim in six months prior	2,196	952 (43.35%)	5,041	3,834 (76.06%)	1.68 (1.59–1.78)

Notes:

Adjusted risk ratios (aRR) were calculated using robust log linear Poisson regression. All models account for clustering by person and complete-case analysis and exclude individuals who died within 30 days (n=51) or were re-incarcerated within 30 days (n=2,776)

¹All multivariable models account for age, sex, education level, race/ethnicity, marital status, history of injection drug use, serious mental illness, cocaine use disorder, alcohol use disorder, HIV diagnosis, HCV diagnosis, unhoused/unstably housed, mental health needs, prior incarceration (since 2009), incarceration lasted >1 month, incarceration related to felony charge, incarceration related to parole violation, self-reported opioid use disorder treatment prior to incarceration, unhoused/unstably housed, mental health needs.

²Multivariable all incarcerations model additionally adjusts for prior MOUD treatment within 6 months of incarceration.