Medications for Opioid Use Disorder in County Jails: Outcomes After Release

Abstract

Background:

In 2019, seven county correctional facilities (jails) in Massachusetts initiated pilot programs to provide all FDA-approved medications to treat opioid use disorder (MOUD).

Methods:

This observational study used linked state data to examine post-release MOUD receipt, overdose, mortality, and reincarceration among 6,400 persons in carceral settings with probable OUD who did (n=2,711; 42%) or did not (n=3,689; 58%) receive MOUD from these programs from 09/01/2019 through 12/31/2020. Log-binomial and proportional hazard models adjusted for propensity score (PS) weights and baseline covariates that remained imbalanced after PS weighting.

Results:

The study cohort included 6400 persons with probable OUD: 2711 (42.4%) received MOUD in jail and 3689 (57.6%) did not. Among persons treated with MOUD in jail, 67.9% received buprenorphine, 25.7% received methadone, and 6.5% received naltrexone. Treated persons were more likely than those not treated to be White (75.4% vs. 58.1%), to be sentenced (31.6% vs. 13.2%), to be receiving MOUD at jail entry (73.7% vs. 17.1%), and to receive MOUD during the first 30 days after community release (60.2% vs. 17.6%; adjusted relative risk, 1.44; 95% confidence interval [CI], 1.38 to 1.50). Only 50.4% of MOUD recipients engaged in MOUD treatment for 75% of the first 90 days after release, and 57.5% were receiving MOUD at 180 days. Receipt of MOUD in jail, as compared with no such receipt, was associated with lower postrelease risks of fatal overdose (adjusted hazard ratio, 0.48; 95% CI, 0.36 to 0.64), nonfatal overdose (adjusted hazard ratio, 0.76; 95% CI, 0.68 to 0.85), death from any cause (adjusted hazard ratio, 0.44; 95% CI, 0.35 to 0.56), and reincarceration (adjusted hazard ratio, 0.88; 95% CI, 0.81 to 0.94). The incidence of hospitalizations did not differ substantially between the two groups.

Conclusion:

Receipt of MOUD in jail was associated with an increased likelihood of postrelease MOUD initiation and decreased risks of overdose, death from any cause, and reincarceration. (Funded by the National Institutes of Health and others.)

The opioid epidemic in the US resulted in 81,083 known fatalities in 2023.¹ In Massachusetts, fatal opioid-related overdoses quadrupled over the past two decades.² People who use opioids are at increased risk of incarceration, and those with incarceration histories a have higher risk of fatal overdose than those without that history.^3–5 In states with the highest overdose rates, 22% of the incarcerated population has opioid use disorder (OUD) compared with 4% of the general population.^6,7 Modeled estimates indicate that overdose fatalities among formerly incarcerated persons may account for 47% of community overdose deaths.⁸ Carceral facilities thus have opportunities for public health and clinical interventions to treat OUD and prevent opioid overdose fatalities.⁵

Medications for opioid use disorder (MOUD), including a full agonist (methadone), partial agonist (buprenorphine), and antagonist (naltrexone), are effective in treating OUD and reducing opioid overdose mortality.^9,10 Prior MOUD treatment studies among sentenced persons in prisons have reported positive outcomes after release to the community.^11,12 However, MOUD treatment options, especially the agonist medications, have disseminated slowly into jails, which serve the majority of incarcerated persons^6,13; only 13% of jails nationwide offer MOUD to all incarcerated persons with OUD.¹³

In Massachusetts, 7 of 13 county correctional facilities (i.e., jails) initiated provision of all FDA-approved forms of MOUD as part of a state mandated pilot program.^14,15 Other states and jurisdictions have also begun to offer MOUD in carceral settings.^16–19 As MOUD access expands in jails, evidence is needed of its impact on treatment continuity, fatal and non-fatal opioid overdose, mortality, recidivism, and related community-based outcomes.²⁰

In collaboration with Massachusetts jails and its Department of Public Health (MDPH), the Massachusetts Justice Community Opioid Innovation Network (MassJCOIN) research hub assessed the implementation and outcomes of the pilot program. We hypothesized that receipt of MOUD in jail, the goal of the legislative mandate, would be associated with improved treatment continuity, opioid overdose, reincarceration, and mortality among persons with OUD returning to the community after release from jail.

METHODS

Chapter 208 of the Acts of 2018 in Massachusetts mandated that five county jails pilot-test the provision of all FDA-approved forms of MOUD; two other county jails also opted in.^14,15 MassJCOIN conducted an effectiveness-implementation study of these pilot programs.²¹ All the authors contributed to the design of the study, critical revisions of the manuscript, and the decision to submit the manuscript for publication (Table S1 in the Supplementary Appendix, available with the full text of this article at NEJM.org). The first three authors and the last author wrote the first draft of the manuscript. The authors vouch for the accuracy and completeness of the data and for the fidelity of the study to the protocol (available at NEJM.org).

An elected county sheriff leads each jail, which holds both pre-trial defendants and convicted persons sentenced to 30 months or less (Table S2). All seven jails offered buprenorphine and methadone as of 09/01/2019; most already had injectable extended-release naltrexone available.

Data Collection and Linkage

Research assistants and jail staff collected data directly from individual detainees, and from jail administrative and clinical records, and entered it into the Bureau of Substance Addiction Services’ (BSAS) treatment database. MDPH’s Public Health Data Warehouse (the PHD) linked BSAS data with over 35 state administrative databases from 01/01/2014 through 12/31/2021.^22,23 The PHD captures information on service use, incarceration and mortality in the state, as well as national mortality data for all Massachusetts residents.^22,23 State statute mandates strict confidentiality of the PHD.^22,23

Cohort and Study Group Definitions

Eligible participants were: incarcerated at one of the seven jails that piloted this program between 09/01/2019 and 12/31/2020; identified as having probable OUD based on criteria specific to each jail (Table 1); had jail data that linked in the PHD (Fig. 1); and exited jail on or before 06/30/2021 to ensure at least 180 days of follow-up through 12/31/2021. Study group classification came from BSAS data that indicated whether or not a participant received MOUD during their first incarceration during the study period.

Table 1.

Site-Specific Source for Determining Probable Opioid Use Disorder (OUD) in Massachusetts, 2019–2021

Jail	Source
A	• Self-reported recent opioid use or substance use disorder during medical/mental health intake, or • Screened positive for OUD via Texas Christian University Drug Screen (TCUDS) tool
B	• Self-reported recent opioid use during medical/mental health intakes, or • Drug Abuse Screening Test 10 (DAST-10) screening (score of 2 or greater out of 10)
C	• Rapid urine drug screen (iCup) positive for an opioid, or • OUD diagnosis assessed during mental health assessment
D	• Diagnosed with any OUD during behavioral health evaluation and self-reported recent opioid use
E	• Self-reported OUD diagnosis during mental health assessment
F	• Self-reported history of substance use disorder or medication for opioid use disorder use during medical/mental health intake, or • Screened positive for OUD via Texas Christian University Drug Screen (TCUDS) tool, or • Self-reported recent opioid use
G	• Public Health Data (PHD) Warehouse algorithm due to inconsistent screening onsite*

^*Probable OUD indicator based on pre-incarceration historic community events in state administrative databases, including opioid overdose, diagnosis of OUD or prior MOUD treatment.²⁷

Figure 1.

Consort Diagram, Massachusetts Justice Community Opioid Innovation Network, 2019–2021

Post-Release Outcomes

The primary outcomes were postrelease MOUD treatment, nonfatal and fatal opioid overdose, reincarceration, and death from any cause. Postrelease MOUD treatment was detected from the All-Payer Claims Database, Prescription Monitoring Program, and BSAS records within the PHD5,^23–25; outcomes were defined as MOUD initiation within 30 days (i.e., any community MOUD treatment or prescription started within 30 days after the date of jail exit), treatment engagement (i.e., MOUD treatment or prescriptions covering ≥75% of the first 90 days after jail release), and treatment retention at 180 days (i.e., any MOUD treatment or prescription on day 180 after release).

A determination of postrelease opioid-related overdose was derived from ambulance-encounter, hospital-discharge, and death-certificate data, dichotomized according to fatal and nonfatal outcomes within the first 30, 90, and 180 days after release as well as across all follow-up time censored after December 31, 2021. Postrelease reincarceration included incarceration after the index release in 10 of the 13 county jails for which data were available and the state prison system; it did not include out-of-state or federal incarceration. A determination of death from any cause during a participant’s follow-up time was derived from death-certificate data, which included out-of-state deaths only for Massachusetts residents.

Statistical Methods

Descriptive analyses generated standard measures of central tendency for continuous variables, and frequencies and percentages for categorical measures. Log-binomial models for time-restricted outcomes and Cox proportional hazards models for time-to-event outcomes assessed post-release differences between those receiving MOUD (all types) vs. those not receiving MOUD.²⁶ Since jail receipt of MOUD was non-randomized, adjusted models use propensity score weights to account for potential selection bias. Propensity score weights were generated based on existing literature and imbalances found in baseline measures; these weights accounted for age, race, ethnicity, sex, veteran status, education level, homeless history, history of opioid-related overdose, MOUD at jail entry (the strongest correlate of MOUD in jail consistent with the legislative mandate), adjudication status, days incarcerated (i.e., the number of days from jail intake to release), and jail county. Adjusted models included covariates that remained imbalanced after propensity score weighting (Fig. S1). Confidence interval widths were not adjusted for multiplicity and should not be used in place of hypothesis testing.

A sensitivity analysis used an algorithm developed for the PHD to identify persons with OUD on the basis of historical community indicators,²⁷ including opioid overdose and diagnosis of OUD or MOUD treatment before the index incarceration, to address differences among jails in procedures used to identify OUD. Another sensitivity analysis accounted for competing events (e.g., deaths in models of incarceration) that were censored in survival models and not factored into time-restricted models.²⁶ An additional sensitivity analysis assessed unmeasured confounding.²⁸

RESULTS

Intake data identified 7,196 persons with probable OUD (Table 1) who were incarcerated in the seven participating jails. After exclusions, the cohort included 6,400 people with OUD who did (n=2,711; 42%) and did not (n=3,689; 58%) receive MOUD in jail (Fig. 1).

Participants’ median age was 36 years and 4785 (74.7%) were male (Table 2). The 2711 jailed persons treated with MOUD were more likely than the 3689 non-treated persons to be White non-Hispanic (2043 [75.4%] vs 2144 [58.1%]), sentenced (858 (31.6%) vs. 487 [13.2%]), and to have had previous overdoses (1236 [45.6%] vs. 1321 [35.8%]). In keeping with the mandate to maintain community medication prescriptions, 1999 (73.7%) of those who received MOUD in jail had been on community MOUD at jail entry vs. 631 (17.1%) of those who did not receive MOUD. Of the total sample, 4521 (70%) had experienced homelessness.

Table 2.

Participant Demographics and Baseline Measures among Persons Incarcerated in Massachusetts Jails, 2019–2021.

	MOUD* N=2711	No MOUD* N=3689
Age, median [IQR]	36 [31, 42]	35 [29, 43]
Sex, N (%)
Male	2013 (74.3)	2772 (75.1)
Female	698 (25.7)	917 (24.9)
Race/Ethnicity, N (%)
White, non-Hispanic	2043 (75.4)	2144 (58.1)
Black, non-Hispanic	156 (5.8)	607 (16.5)
Asian/Pacific Islander, non-Hispanic	…†	16 (0.4)
Hispanic	492 (18.2)	889 (24.1)
American Indian or Other	13 (0.5)	33 (0.9)
Education, N (%)
High school or less	1867 (68.9)	2334 (63.3)
More than high school	639 (23.6)	854 (23.2)
Missing/Not collected	205 (7.6)	501 (13.6)
Veteran, N (%)
No	2418 (89.2)	3313 (89.8)
Yes	293 (10.8)	376 (10.2)
Any homelessness 2019–2021, N (%)
No	543 (20.0)	765 (20.7)
Yes	1922 (70.9)	2599 (70.5)
Unstable housing	245 (9.0)	230 (6.2)
Unknown	…†	95 (2.6)
Adjudication status, N (%)
Pre-trial	1821 (67.2)	3020 (81.9)
Sentenced	858 (31.6)	487 (13.2)
Safekeep	…†	162 (4.4)
Unknown/Missing	…†	20 (0.5)
Days incarcerated, median [interquartile range]	47 [14, 130]	17 [2, 63]
Pre-incarceration§ opioid overdose, N (%)
No	1475 (54.4)	2368 (64.2)
Yes	1236 (45.6)	1321 (35.8)
Pre-incarceration4 opioid overdose count, mean (std)	1.4 (2.6)	1.0 (2.3)
Pre-existing community MOUD* at jail entry‡, N (%)
No	712 (26.3)	3058 (82.9)
Yes	1999 (73.7)	631 (17.1)
MOUD* type in jail, N (%)
Methadone	696 (25.7)
Buprenorphine (Suboxone)	1840 (67.9)
Naltrexone (Vivitrol)	175 (6.5)
County House of Corrections / Jail, N (%)
A	390 (14.4)	431 (11.7)
B	171 (6.3)	158 (4.3)
C	773 (28.5)	1009 (27.4)
D	64 (2.4)	211 (5.7)
E	299 (11.0)	315 (8.5)
F	322 (11.9)	210 (5.7)
G	692 (25.5)	1355 (36.7)

^*Medication for opioid use disorder (MOUD).

^†Suppressed for small cell size or complementary cell suppression.

^‡MOUD within 14 days of jail entry or positive report at intake.

^§The Public Health Data Warehouse (PHD) includes pre-incarceration data back to 01/01/2014.

Among participants treated with MOUD in jail, the most commonly used medications were buprenorphine (in 1840 participants [67.9%]) and methadone (in 696 [25.7%]). A total of 1465 participants (54.0%) in the MOUD group were from two jails (C or G), and the likelihood of MOUD receipt ranged from 64 of 275 (23.3%) in jail D to 322 of 532 (60.5%) in jail F.

Postrelease MOUD Initiation, Engagement, and Retention

Treatment with MOUD while jailed was associated with a higher incidence of initiation, engagement, and retention in community MOUD after release (Table 3). Of participants in the MOUD group, 1633 (60.2%) initiated community MOUD in the first 30 days after release, as compared with 648 (17.6%) of the participants in the non-MOUD group (adjusted relative risk, 1.44; 95% confidence interval [CI], 1.38 to 1.50). Of participants in the MOUD group, 1365 (50.4%) engaged in MOUD treatment for at least 75% of the first 90 days after release, as compared with 454 (12.3%) of the participants in the non-MOUD group (adjusted relative risk, 1.77; 95% CI, 1.64 to 1.92). At 180 days, 1560 participants (57.5%) in the MOUD group were retained in treatment, as compared with 840 participants (22.8%) in the non-MOUD group (adjusted relative risk, 1.40; 95% CI, 1.31 to 1.49).

Table 3.

Post-Release Outcomes among People Who Received and Did Not Receive Medications for Opioid Use Disorder (MOUD) While Incarcerated in Massachusetts’ County Correctional Facilities, 2019–2021.

	MOUD N=2711	No MOUD N=3689	Unadjusted RR‡ or HR§ (95% CI)¶	Adjusted# RR‡ or HR§ (95% CI)¶
Days of follow-up, median [IQR]	620 [470, 710]	650 [500, 745]
Post-release MOUD Treatment
MOUD initiation, N (%)*	2243 (82.7)	1921 (52.1)	2.92 (2.75, 3.12)§	1.94 (1.82, 2.06)§#
MOUD initiation within 30 days, N (%)	1633 (60.2)	648 (17.6)	3.43 (3.18, 3.70)‡	1.44 (1.38, 1.50)‡#
MOUD engagement 75% of 90 days, N (%)	1365 (50.4)	454 (12.3)	4.09 (3.72, 4.49)‡	1.77 (1.64, 1.92)‡#
MOUD retention at 180 days, N (%)	1560 (57.5)	840 (22.8)	2.53 (2.36, 2.70)‡	1.40 (1.31, 1.49)‡#
MOUD type post-release, N (%)
None	252 (9.3)	1684 (45.6)
Buprenorphine	1656 (61.1)	1209 (32.8)
Methadone	687 (25.3)	489 (13.3)
Injectable extended-release naltrexone	116 (4.3)	307 (8.3)
Post-release Opioid Overdose
Opioid overdose post-release, N (%)	596 (22.0)	841 (22.8)	0.97 (0.87, 1.08)§	0.73 (0.66, 0.81)§#
Fatal opioid overdose, N (%)	71 (2.6)	130 (3.5)	0.76 (0.57, 1.02)§	0.48 (0.36, 0.64)§#
Non-fatal opioid overdose, N (%)	545 (20.1)	746 (20.2)	1.00 (0.89, 1.11)§	0.76 (0.68, 0.85)§#
Opioid overdose in 30 days
Any	91 (3.4)	198 (5.4)	0.63 (0.49, 0.80)‡	0.55 (0.43, 0.70)§#
Fatal	…†	17 (0.5)	0.80 (0.37, 1.75)‡	1.10 (0.52, 2.32)§#
Non-fatal	…†	182 (4.9)	0.61 (0.47, 0.78)‡	0.50 (0.39, 0.65)§#
Opioid overdose in 90 days
Any	191 (7.1)	336 (9.1)	0.77 (0.65, 0.92)‡	0.65 (0.53, 0.77)§#
Fatal	22 (0.8)	30 (0.8)	1.00 (0.58, 1.73)‡	1.03 (0.60, 1.75)§#
Non-fatal	171 (6.3)	309 (8.4)	0.75 (0.63, 0.90)‡	0.62 (0.52, 0.74)§#
Opioid overdose in 180 days
Any	300 (11.1)	478 (13.0)	0.85 (0.75, 0.98)‡	0.69 (0.60, 0.79)§#
Fatal	33 (1.2)	51 (1.4)	0.88 (0.57, 1.36)‡	0.77 (0.50, 1.19)§#
Non-fatal	273 (10.1)	435 (11.8)	0.85 (0.74, 0.99)‡	0.68 (0.59, 0.79)§#
Death from any cause
Deaths, all-cause, N (%)	99 (3.7)	198 (5.4)	0.70 (0.55, 0.89)§	0.44 (0.35, 0.56)§#
Reincarceration
Reincarceration, N (%)	1203 (44.4)	1834 (49.7)	0.85 (0.79, 0.91)§	0.88 (0.81, 0.94)§#
Reincarceration within 90 days	377 (13.9)	718 (19.5)	0.71 (0.64, 0.80)‡	0.78 (0.70, 0.88)§#
Reincarceration within 180 days	646 (23.8)	1053 (28.5)	0.83 (0.77, 0.91)‡	0.89 (0.81, 0.96)§#

^*Initiation required MOUD record with start date on or after jail release date.

^†Suppressed for small cell size or complementary cell suppression.

^‡Log-binomial regression for time-restrained outcomes, RR: relative risk.

^§Cox proportional hazards for survival analysis, HR: hazard ratio.

^¶Confidence interval widths have not been adjusted for multiplicity and should not be used in place of hypothesis testing.

^#Adjusted for propensity score weights generated by regressing study group outcome on age, race, ethnicity, sex, veteran status, educational level, homeless history, pre-incarceration overdose, MOUD at jail entry, adjudication status, days incarcerated and county of the jail; as well as covariates that remained imbalanced after propensity score weighting (homeless history and days incarcerated).

Postrelease Overdose, Death, and Reincarceration

MOUD in jail was associated with a lower risk of fatal overdose (adjusted hazard ratio, 0.48; 95% CI, 0.36 to 0.64), nonfatal overdose (adjusted hazard ratio, 0.76; 95% CI, 0.68 to 0.85), death from any cause (adjusted hazard ratio, 0.44; 95% CI, 0.35 to 0.56), and reincarceration (adjusted hazard ratio; 0.88; 95% CI, 0.81 to 0.94). The incidence of hospitalizations did not differ substantially between the two groups (Table S3).

In this study involving persons with OUD in Massachusetts county jails, those who received MOUD in jail had significantly greater postrelease MOUD continuity in the community than those who did not receive MOUD. Competing events also did not substantially change the adjusted results (Table S5).²⁶ E-value analyses (Table S5) suggest that a hypothetical unmeasured confounder of modest magnitude (i.e., relative risk < 2) could explain away some overdose or incarceration findings, while the post-release treatment and death from any cause results appear more robust to unmeasured confounding.²⁸

DISCUSSION

In the present study receipt of MOUD treatment among persons with OUD in Massachusetts county jails was associated with significant improvements in post-release MOUD continuity in the community compared to jailed persons with OUD who did not receive MOUD. This receipt of MOUD, made possible through a state mandate that these jails deliver all FDA-approved forms of MOUD, was similarly associated with lower non-fatal and fatal opioid overdose, all-cause mortality, and reincarceration. These findings extend to jail settings the decades of accumulated evidence for the benefits of MOUD treatment in community settings.^11,29

The association of the implementation of these jail pilot programs with post-release treatment in the community is consistent with prior studies.^30,31 The literature supports the importance of treatment post-release. A study of agonist treatment in Australian prisons found that the effect of in-prison treatment alone did not last long after community reentry, but that ongoing care in the four weeks after release was associated with a 75% mortality reduction.¹¹ Older literature on the effect of correctional treatment similarly emphasizes the importance of care during the transition back to the community.³² In a similar manner, the association of in-jail MOUD with ongoing care remained consequential out to 180 days post-release.

The magnitude of the associated difference in overdose-related mortality aligns with prior studies of carceral MOUD.¹¹ After Rhode Island’s centralized correctional facility implemented all three FDA-approved MOUDs, a retrospective evaluation of community overdose fatalities found that 26 of 179 (14.5%) had a history of incarceration prior to implementation versus 5 of 157 (5.7%) afterwards – suggesting a 60% decrease in in the strength of association between overdose fatality and previous incarceration.¹⁷ The likely effects of MOUD on mortality after jail release are large and represent a highly effective approach to reduce community opioid overdose deaths.

The association of jail-based MOUD with lower reincarceration is consistent with prior work suggesting that buprenorphine receipt in jail is related to lower recidivism.³³ Our results differ from prior research that did not find lower reincarceration rate,^31,34 possibly because the propensity score weighting here mitigated selection effects. We believe that the reincarceration finding is important for policymakers and advocates seeking to reduce the carceral population.

Several jails in Massachusetts offered injectable long-acting naltrexone prior to this study, while institutional and other barriers slowed the uptake of the agonist medications, buprenorphine and methadone, which have stronger evidence for reducing mortality.²⁹ Across the US, jails that offer agonist medications often restrict their use to pregnant people.¹³ Consequently, many persons with OUD who are incarcerated in the US, even those on appropriate agonist pharmacotherapy prior to entry, suffer through forced withdrawal. Forced withdrawal places the person at high-risk for recurrent use and overdose during community reentry because of the person’s reasonable aversion to initiating MOUD out of fear of repeated forced withdrawal during even minor legal detention.^35,36

The lower delivery of MOUD to Black non-Hispanic and Hispanic people in the jails in the present study is of concern, especially given their overrepresentation in carceral populations. The minimum requirement of the legislative mandate in Massachusetts was to maintain MOUD treatment among persons with a confirmed community prescription (i.e., buprenorphine or naltrexone) or opioid treatment program (i.e., methadone) dose at intake, and otherwise to initiate pharmacotherapy within 30 days of jail release for those with OUD.²¹ One possible explanation for the disparities is that well-described community inequities in MOUD access were amplified because maintenance of prior community treatment was the predominant indication for MOUD in these jails.^37,38 Some jails did go beyond the legislative mandate to offer MOUD induction to all persons with OUD soon after intake, including those on street opioids with short stays (e.g. non-sentenced, pre-trial detention) — jails with a greater Black non-Hispanic and Hispanic population might have been less proactive in offering early induction to persons who were not on MOUD at jail entry. Of note, 300 female residents transferred to one of the study sites in October 2019 were predominantly White, non-Hispanic and most were receiving MOUD, which could partially explain the observed disparities. Nonetheless, these data cannot rule out bias and discrimination as to which people were offered MOUD, prior negative experiences with treatment, including under- or mistreatment, distrust of the medical or carceral community, self- or other stigma, or personal preferences that might have lessened receipt of MOUD among populations of color in these jails.

A strength of our study is its real-world assessment of outcomes in county jails under the supervision of elected Sheriffs, which is more generalizable across U.S. county jails than prior evaluations from centralized state facilities and prisons.^34,39 However, the study has certain limitations: these findings should only be extrapolated with the caveat that the sample came from 7 of 13 county jails that volunteered to participate in a state program in a single northeastern U.S. state with a higher prevalence of OUD and a lower rate of incarceration than most of the U.S., and that White, non-Hispanic persons and women were overrepresented (Table S6). The use of the PHD allowed detection of post-release treatment, opioid overdose and most incarcerations statewide, as well as mortality for Massachusetts residents nationwide; nonetheless, its inability to capture out-of-state or federal treatment, overdose and incarceration outcomes for people who left Massachusetts is a limitation. Furthermore, the process for identifying persons with OUD reflected real-world practices in these jails and thus was not standardized; a sensitivity analysis that applied the PHD OUD identification algorithm to all participants was reassuring for the subset with OUD-related histories in the community prior to incarceration (Table S4). In addition, most persons received MOUD because they had been receiving it in the community, and thus patient selection likely reflected community prescribing practices. Propensity score methods cannot establish causality because of their sensitivity to outliers, risk of model misspecification, and inability to account for unmeasured confounding, which is most salient for the overdose and incarceration outcomes (Table S5).⁴⁰ Also, the PHD maintains a high level of data privacy, so the inability of our analysts to see the data directly made it difficult to assess data quality and the adequacy of data linkage. Finally, the study period overlapped with the pandemic, which may have introduced distortions, such as greater all-cause and opioid overdose mortality, as well as early releases and fewer incarcerated persons than usual.²⁴

Provision of MOUD in county jails demonstrated beneficial associations with post-release treatment continuity, overdose-specific and all-cause mortality, and reincarceration; the agonist medications buprenorphine and methadone, which comprised 94% of MOUD received, likely explained most of the benefit. Observed racial and ethnic disparities in MOUD receipt merit further investigation. Given the overrepresentation of Black non-Hispanic and Hispanic people in the carceral population, jails may have the opportunity to improve community equity in access to MOUD by offering MOUD induction to all new intakes who meet criteria for OUD, regardless of prior treatment history, and linking them to post-release community treatment.^15,38 These findings provide the impetus for the majority of jails in the U.S. and internationally that do not provide agonist pharmacotherapy for OUD to implement these effective medications to reduce opioid overdose, mortality and reincarceration when jailed persons reenter the community. In the U.S., this research also has important implications for the effective investment of opioid settlement funds and other policy initiatives to augment the delivery of MOUD to quell the ongoing crisis of opioid use and its related morbidity and mortality.

We found that receipt of MOUD in jail was associated with improved postrelease MOUD initiation and lower risks of overdose, death from any cause, and reincarceration.