Access to and predictors of substance use treatment and support among people experiencing incarceration in the United States: Analysis of a national cross-sectional study

Abstract

Substance use-related overdose continues to be one of the leading causes of preventable death in the U.S. People returning from prisons and jails are at heightened risk. Certain substance use treatment methods in carceral facilities – especially medication for opioid use disorder (MOUD) – have shown promise in reducing overdose rates. Treatment availability has been under-studied, with past research often measuring whether facilities offer treatment, but not whether individuals actually receive it. This study used individual-level data to characterize who qualifies for prison-based treatment, who receives it, and what factors affect one’s likelihood of being treated, drawing from the most recent nationally-representative U.S. Survey of Prison Inmates. Descriptive statistics indicate that people with substance use disorder (SUD) who entered American prisons from 2014 to 2016 had lower levels of educational attainment, employment and housing, and higher levels of physical and mental illness. Just 13 % of individuals with SUD received any form of substance use treatment in prison; pharmacotherapies like methadone were almost nonexistent (<1 %). In controlled analysis, individuals who self-identified as Non-Hispanic Black or Hispanic had lower odds of receiving any treatment or support compared to Non-Hispanic White individuals. People convicted of a violent offense had lower odds of treatment than those convicted of other classes of crimes. These novel findings indicate that, in 2016, America’s prison-based substance use treatment had both poor accessibility and inequitable distribution. This raises concerns that, even as facility-level availability of modalities like MOUD continues to expand, certain groups may be left out without attention to individual-level availability.

Highlights

• •Just 13 % of those with substance use disorder received treatment in US prisons.
• •Medication assisted treatment was almost non-existent in US prisons in 2016.
• •Versus White individuals, Black and Hispanic individuals had lower odds of treatment.
• •By offense class, those with violent convictions had the lowest odds of treatment.
• •Survey of Prison Inmates variables make treatment quality difficult to assess.

1. Introduction

Preventable drug overdose deaths – especially those related to opioids – represent a global cause of mortality in which the United States stands apart. In 2022, overdose death rates controlling for population were almost one and a half times higher in the US than the next closest nation (Gumas, 2025). One theory for this gap is the relationship of overdose deaths to another issue in which the U.S. is singular: mass incarceration. It is estimated that, among those recently released from carceral facilities in the US, overdose risk ranges from 40 to 123 times higher than the general public (Hartung et al., 2023, Joudrey et al., 2019, Ranapurwala et al., 2018). With an estimated 600,000 people released each year back into American communities – many of whom with a history of substance use – the number of people at risk of overdose is massive (Hamilton and Belenko, 2019, Bronson et al., 2017).

In response, advocates have called for jails and prisons to begin offering more treatment for substance use disorder (SUD). Though efficacy varies by modality (Mitchell et al., 2007), one treatment approach with strong empiric backing is medication for opioid use disorder (MOUD), in which methadone, buprenorphine or naltrexone are administered to individuals during incarceration to pharmacologically treat opioid use disorder. Studies have shown that prison-based treatment with methadone and buprenorphine reduces individuals’ risk of fatal overdose upon reentry by 81 % (Santo et al., 2021), lowers rates of recidivism by 32 % (Evans et al., 2022), and reduces rates of drug use and infectious disease (Moore et al., 2019). Naltrexone has not demonstrated an effect on post-release overdose deaths, but may lower drug use on reentry (Lee et al., 2015). Given this evidence, buprenorphine and methadone are considered the gold standard for substance use treatment on both sides of the prison wall (Homans et al., 2023).

Prisons and jails have been slow to adopt these medications, though precise estimates of how many people receive this and other treatments in prisons are hard to come by. One recent national survey of U.S. jails found that, though 44 % of facilities offered MOUD to at least some persons, only 13 % offered them to anyone with an opioid use disorder (Balawajder et al., 2024). These results are indicative of larger challenges in both prison-based treatment availability, and existing studies measuring it. Many facilities only offer MOUD to select groups, like pregnant women or those in active withdrawal (Fortino et al., 2024, Friedmann et al., 2012), or exclusively carry naltrexone, the only MOUD to not demonstrate meaningful effects on post-release overdose rates (Sufrin et al., 2023). As a result, much of the extant literature – which relies on facility-level reporting of availability (Nunn et al., 2009, Taxman et al., 2007) – may consider MOUD “available,” even if the real number of people receiving meaningful treatment is low. Thus, there is a gap in the literature on how much treatment is actually being received by people with SUD in prison, as well as where gaps or inequities exist in that provision.

The most recent Survey of Prison Inmates (SPI) offers an opportunity to comprehensively investigate individual-level treatment rates. The 2016 SPI is the most current nationally-representative dataset regarding people experiencing incarceration in the country’s state and federal prisons, though a successor survey is now under development (Pollack and Dosani, 2024). Implemented by Research Triangle Institute (RTI), the 2016 SPI successfully surveyed more than 24,000 incarcerated individuals from 306 state prisons and 58 federal prisons across the United States, capturing person-level data between 2014 and 2016 (Glaze, 2019). Though the data comes from before some states began mandating treatment availability (Weizman et al., 2021), the breadth and depth of the SPI makes it one of the few tools capable of characterizing how widely available treatment was for individuals in prison, a question whose answer holds important implications for practice today.

Using the 2016 SPI, we sought to characterize the need for and availability of substance use treatment within America’s prisons. Focusing on individuals with diagnosable substance use disorder (SUD), we sought to answer three questions. Firstly, how many people in America’s prisons required substance use treatment, and what was their sociodemographic profile? Comparing those with SUD to peers without that level of documented use allows us to characterize the unique needs and traits of the population with a demonstrated need for treatment. Second, how many people with SUD reported that they were currently receiving substance use treatment in America’s prisons? Understanding this rate at the level of individuals will more accurately characterize availability at a key moment in the opioid overdose epidemic, when fentanyl overdose rates were rapidly rising (Seth et al., 2018). Finally, what characteristics made one more or less likely to receive treatment? This is a novel analysis, investigating which individuals with SUD received treatment in 2016 to establish a baseline understanding of the (in)equitable distribution of treatment resources that may continue today. Ultimately, we sought to investigate the hypothesis that, at a critical time in the opioid epidemic, substance use treatment in US prisons was both inadequate and inequitably distributed, playing a role in the dramatic rise in overdose deaths among this vulnerable population.

2. Methods

2.1. Data

To address our study aims, we analyzed data from the most recent iteration (2016) of the Bureau of Justice Statistics’ Survey of Prison Inmates (SPI). The study’s methodology has been described in-depth elsewhere (Glaze, 2019). Summarized briefly, the SPI was designed to be nationally-representative of the 2016 prison population (2001 facilities, 1502,671 individuals), with a two-step random selection process that generated a sample of 24,848 individuals from facilities across the United States. Surveys administered to each selected individual contained several hundred questions, ranging from demographics to health and substance use information. Each individual’s responses were assigned a weight to correct for sampling biases and make the final dataset more nationally representative. Reflecting a statistical pattern known as length-biased sampling, this cross-section more heavily represents persons serving longer sentences. To study a sample more representative of the cohort of persons who entered prison at the time of survey administration, we examined the subgroup of SPI respondents who entered prison within three years of the beginning of sample collection (2014, 2015 and 2016). All data used for this study is public-use and therefore IRB exempt.

2.2. Variables

Demographic variables included age (18–24; 25–34; 35–44; 45–54; 55–64; 65 +), sex, and combined race and ethnicity categories (Non-Hispanic [NH] White, NH Black, Hispanic; American Indian / Alaskan native [AIAN]; Asian / Native Hawaiian / Pacific Islander [ANHPI]; 2 + races). A variable on gender was available, but so few participants identified as non-cis that we did not use the variable in analyses. Geography data was most granularly available in the dataset at the level of state of residence. Education level was an ordinal variable (less than high school; graduated high school; some college; graduated college or above). Additional demographics included: employment in the 30 days prior to arrest; insurance status in the 30 days prior to arrest; housing status in the 12 months prior to arrest.

We also created composite variables to capture physical and mental health status. Physical health status was an ordinal variable with three levels: no physician-diagnosed physical health conditions; one condition; or two or more conditions. Conditions included: cancer; hypertension; stroke; diabetes; heart disease; kidney problems; arthritis / gout / lupus / fibromyalgia; asthma; cirrhosis; tuberculosis; hepatitis B; hepatitis C; or HIV. Mental health status had three levels: no physician-diagnosed mental health conditions; one condition; or two or more conditions. These conditions included: manic depression / bipolar disorder / mania; depressive disorder; schizophrenia / other psychotic disorder; post-traumatic stress disorder; anxiety disorder; or any personality disorder.

The 2016 SPI captured data on use of the following substance classes: alcohol; cannabis; heroin; prescription drugs not prescribed by a doctor; methamphetamines; cocaine; crack; phencyclidine (PCP); party drugs (LSD, MDMA); and other hallucinogens. All substances had variables indicating their use at time of arrest and in the 30 days before arrest, except alcohol, which only captured the former. We created a substance use disorder variable based on whether, in the past 12 months, individuals endorsed two or more DSM-V criteria captured by the survey (American, 2013). Offense information existed in the survey at the level of broad categories (violent, property, drug, public order, or other) as well as more specific sub-classes (e.g. homicide as a sub-class of violent offense). Across variables, the small number that answered “NA / Unknown” were classified in the negative (e.g. no use of the particular substance).

Finally, the SPI included variables for different forms of treatment or support both before and after prison entry. The following options were captured: detoxification; an alcohol or drug treatment program in which participants lived in a special facility / unit; counseling while not in a special living facility; a self-help or peer group; an education or awareness program; pharmacotherapy for substance use disorder (e.g. methadone, buprenorphine, disulfiram); or other. We term the pharmacotherapy category “medication for addiction treatment (MAT)” rather than MOUD given that the survey included non-opioid related treatments (e.g. disulfiram). Among these variables, we differentiated between treatment (detox, special living facility, counseling, MAT) and support (self-help or peer group, education or awareness campaign), given that only the former would be considered a treatment modality from a clinical perspective.

2.3. Analyses

To understand who needed substance use disorder treatment in prison, we compared characteristics of the sample population with a diagnosable substance use disorder in the 12 months prior to arrest (SUD sample) to those of respondents who did not report such use (non-SUD sample). We performed chi-square tests to determine whether the two samples differed significantly across several variables. To address the stratified and weighted nature of the SPI, all descriptive statistics were calculated using the SPI’s analytic sample weights.

To understand the rates of SUD treatment and support before and during incarceration, we calculated the percentage of the SUD sample that received each treatment modality, before and after entering prison. We also looked specifically at MAT rates among people with documented use of heroin in the 30 days prior to their arrest.

The final analysis was a logistic regression looking at which variables affected one’s odds of receiving substance use treatment or support in prison. The outcome variable was receipt of any treatment or support since time of prison entry. Potential confounding variables were identified using directed acyclic graphing and added to the model, and included sociodemographics (age, sex, race / ethnicity, education status, insurance status, employment status), recent drug use, mental health status, offense class and state or territory (including Washington D.C. and Puerto Rico) (Shrier and Platt, 2008). Reference categories for each variable are specified below. All analyses were conducted in RStudio (RStudio version 2022.07.1 +554).

3. Results

3.1. Sample characteristics

In total, the population that had entered prison within three years of survey administration was 12,586 (weighted: 667,498). Of this total, 5327 (weighted: 272,346) met the criteria for a substance use disorder (42.3 % of sample). Comparison of the samples with and without SUD yielded significant differences across nearly all queried variables, with p values less than 0.001 for all of the following results, except where noted (Table 1). Across self-identified race, the SUD sample was more Non-Hispanic White (44 % vs. 28 %) and slightly more AIAN (2 % vs. 1 %) and 2 +  races (11 % vs. 9 %), but less Non-Hispanic Black (21 % vs. 32 %) and Hispanic (19 % vs. 27 %), with ANHPI percentages roughly equivalent (both 1 %). The sample with SUD also had lower employment rates in the 30 days pre-arrest (55 % vs. 62 %), and fewer people housed in the 12 months pre-arrest (85 % vs. 92 %). There was no significant difference between pre-arrest insurance status (p = 0.93).

Table 1.

Sociodemographics of incarcerated individuals with and without substance use disorder, 2014–2016 sample.

	No SUD (n = 395,151)	SUD (n = 272,346)	Percent of Total (n = 667,498)	Chi Squared
Age Group* **				< 0.001
18–24	14 %	15 %	15 %
25–34	35 %	39 %	37 %
35–44	26 %	26 %	26 %
45–54	16 %	13 %	15 %
55–64	7 %	5 %	6 %
65 +	2 %	1 %	1 %
Sex (Male)* **	92 %	88 %	90 %	< 0.001
Race* **				< 0.001
NH White	28 %	44 %	35 %
NH Black	32 %	21 %	27 %
Hispanic	27 %	19 %	24 %
AIAN	1 %	2 %	2 %
ANHPI	1 %	1 %	1 %
Employed Pre-Arrest* **	62 %	55 %	59 %	< 0.001
Insured Pre-Arrest	51 %	51 %	51 %	0.93
Housed Pre-Arrest* **	92 %	85 %	89 %	< 0.001
Education Level* **				< 0.001
<  High School	57 %	59 %	58 %
High School	23 %	24 %	24 %
Some College	13 %	13 %	13 %
College +	5 %	3 %	4 %

*significance < 0.05;

* * significance < 0.01;

* **  significance < 0.001

Health, offense class and substance use data showed some of the starkest differences between the SUD and non-SUD samples (Table 2). All results were statistically significant (p < 0.001). Compared with those without SUD, people with SUD were more likely to endorse a chronic medical condition (26 % vs. 22 %) as well as two or more mental health conditions (40 % vs. 24 %). In regard to offenses, the sample with SUD were less likely to be serving a violent offense (27 % vs. 36), with differences in violent crime rates due mostly to lower rates of homicide (2 % vs. 5 %) and sexual assault (4 % vs. 8 %) rates. Substance use mirrored each other across samples in terms of most frequently used substances, although those with SUD endorsed having a higher overall rate of use, with prescription drugs and heroin being the 4th and 6th most-commonly used. The vast majority of people with SUD used two or more substances the month prior to arrest (70 % vs. 26 %).

Table 2.

Health, offense class and pre-arrest substance use among incarcerated individuals with and without substance use disorder, 2014–2016 sample.

	No SUD (n = 395151)	SUD (n = 272346)	Percent of Total (n = 667498)	Chi Squared
Physical Health* **				< 0.001
2 +  Conditions	14 %	15 %	14 %
1 Condition	22 %	26 %	24 %
None	64 %	59 %	62 %
Mental Health* **				< 0.001
2 +  Conditions	24 %	40 %	31 %
1 Condition	11 %	14 %	13 %
None	65 %	46 %	57 %
Offense Class* **				< 0.001
Homicide	5 %	2 %	4 %
Rape / Sexual Assault	8 %	4 %	7 %
Robbery	8 %	7 %	8 %
Assault	12 %	12 %	12 %
Other Violent	2 %	2 %	2 %
Total Violent	36 %	27 %	32 %
Burglary	6 %	10 %	8 %
Other Property	11 %	15 %	13 %
Total Property	17 %	25 %	20 %
Drug Trafficking	17 %	17 %	17 %
Drug Possession	5 %	8 %	6 %
Other Drug	1 %	1 %	< 1 %
Total Drug	22 %	26 %	24 %
Weapons	7 %	5 %	6 %
Other Public Order	16 %	16 %	16 %
Total Public Order	23 %	21 %	22 %
Other Unspecified	< 1 %	< 1 %	< 1 %
Substance Use
Cannabis* **	40 %	61 %	48 %	< 0.001
Alcohol (at arrest)* **	20 %	40 %	28 %	< 0.001
Methamphetamines* **	11 %	36 %	21 %	< 0.001
Prescription Drugs* **	10 %	33 %	19 %	< 0.001
Cocaine* **	9 %	25 %	15 %	< 0.001
Heroin* **	4 %	20 %	11 %	< 0.001
Party Drugs (LSD, MDMA)* **	5 %	11 %	7 %	< 0.001
Crack* **	3 %	11 %	6 %	< 0.001
Phencyclidine* **	1 %	3 %	2 %	< 0.001
2 or more substances* **	26 %	70 %	44 %	< 0.001

*significance < 0.05;

* * significance < 0.01;

* **  significance < 0.001

3.2. Treatment and support availability

In the SUD sample, most common treatment or support modalities mirrored each other before and after prison entry, though rates differed noticeably (Table 3). The most common modalities were those considered “support” rather than true “treatment,” and included self-help or peer groups (37 % pre-incarceration vs. 17 % during incarceration) and education or awareness campaigns (31 % vs. 14 %). The most common treatment modality was special living facility (27 % vs. 8 %); maintenance medication was least common (9 % vs. <1 %). Of the 272,346 weighted respondents, only 1686 reported receiving MAT (0.6 %). Looking specifically at individuals who used heroin prior to arrest and met criteria for a SUD diagnosis – i.e., individuals who would best-qualify for MAT – only 2.2 % of this population received MAT after prison entry. In total, 24 % of people with SUD had received some form of treatment or other substance use support since prison entry. When considering treatments alone (special living facility, counseling, detox and MAT), this number drops to 13 %.

Table 3.

Treatment & support modalities pre- and post-prison entry in substance use disorder group, 2014–2016 sample.

	SUD, pre-Arrest (n = 272346)	SUD, while incarcerated (n = 272346)
Self-help / peer group	37 %	17 %
Education campaign	31 %	14 %
Special living facility	27 %	8 %
Counseling w/o facility	26 %	7 %
Detoxification	13 %	1 %
Medication (MAT)	9 %	< 1 %
Other	3 %	< 1 %
Total	47 %	24 %
Treatment Only (facility; counseling; detox; MAT)	38 %	13 %

3.3. Predictors of treatment provision

In logistic regression analysis, we found that several variables were significant predictors of SUD treatment or other supports in prison (Table 4). Demographically, individuals had significantly lower odds of receiving treatment or support if they identified as Black (adjusted odds ratio [aOR]: 0.71; 95 % confidence interval [CI]: 0.63–0.81) or Hispanic (aOR: 0.71; CI: 0.54–0.71) compared to White individuals. In terms of conviction class, compared to those convicted of public order crimes, individuals convicted of violent crimes had lower odds of treatment or support (aOR: 0.63; CI: 0.55–0.72). Geographically, more than half of the states had a significant relationship to treatment or support provision. We chose Texas as the reference variable as it is the state with the largest number of participants in the SPI sample. Compared to Texas, only Virginia had a lower odds of treatment or support (aOR: 0.55; CI: 0.38–0.78). States with the highest odds of receiving treatment were New York (aOR: 3.73; CI: 2.9–4.79) and Pennsylvania (aOR: 3.49; CI: 2.78–4.39). Regionally, all nine US census regions had states with statistically significant odds of receiving treatment. In several regions, more than half of the states were associated with higher or lower odds of treatment, including: Pacific (3/5); Mountain (5/8); West North Central (4/7); East South Central (3/4); Mid-Atlantic (2/3); and South Atlantic (5/9). New England had the fewest states rise to the level of significance, with only Massachusetts and Vermont showing higher odds than Texas (2/6).

Table 4.

Results from logistic regression with primary outcome of receipt of any form of substance use treatment or support, 2014–2016 sample.

		ß	Adjusted Odds Ratio (aOR)	aOR 95 % CI Lower	aOR 95 % CI Upper
Age(ref: 18–24)
	25–34 * **	0.34	1.41	1.22	1.63
	35–44 * **	0.46	1.59	1.36	1.85
	45–54 * **	0.47	1.61	1.35	1.91
	55–64 * **	0.42	1.52	1.21	1.92
	65 +	−0.18	0.83	0.49	1.36
Female Sex* ** (ref: male)		0.21	1.23	1.11	1.37
Race / Ethnicity (ref: Non-Hispanic White)
	Non-Hispanic Black* **	−0.34	0.71	0.63	0.81
	Hispanic* **	−0.48	0.62	0.54	0.71
	American Indian / Alaska Native	−0.06	0.94	0.66	1.32
	Asian / Native Hawaiian	−0.35	0.71	0.44	1.1
	Two or More Races	−0.11	0.89	0.78	1.03
Educational Attainment(ref: < high school)
	High School	−0.1	0.91	0.81	1.01
	Some College	0.05	1.06	0.93	1.2
	College +	−0.14	0.87	0.71	1.06
Mental Health Status* ** (linear)		0.3	1.35	1.25	1.45
Drugs used pre-arrest
	Heroin* **	0.49	1.63	1.42	1.86
	Cocaine* **	0.22	1.24	1.08	1.43
	Crack * **	0.32	1.38	1.14	1.66
	Prescription Drugs * **	0.33	1.39	1.25	1.56
	Phencyclidine	−0.23	0.79	0.54	1.14
	Methamphetamines* **	0.46	1.58	1.41	1.77
	Other Hallucinogens	−0.06	0.94	0.69	1.27
	Cannabis* *	0.16	1.17	1.06	1.29
	Party Drugs (LSD, MDMA)	−0.05	0.95	0.79	1.14
	Alcohol (at time of arrest only)* **	0.62	1.86	1.7	2.04
Pre-Arrest Employment		−0.04	0.96	0.88	1.05
Pre-arrest Insurance		0.03	1.03	0.94	1.13
Offense Class (ref: Public Order)
	Violent* **	−0.47	0.63	0.55	0.72
	Property	0	1	0.88	1.14
	Drug* **	0.23	1.26	1.11	1.44
	Other	−0.31	0.74	0.39	1.31
State (ref: TX)
	AK* *	0.68	1.97	1.17	3.27
	AL* *	0.63	1.88	1.27	2.74
	AR*	0.37	1.45	1.01	2.06
	AZ*	0.45	1.57	1.1	2.22
	CA* *	0.32	1.38	1.08	1.75
	CO* **	1.16	3.18	2.2	4.58
	CT	0.19	1.2	0.82	1.74
	DC* *	1.22	3.39	1.37	7.66
	DE* *	0.8	2.23	1.22	4.02
	FL	−0.06	0.94	0.72	1.23
	GA	−0.13	0.88	0.66	1.16
	HI* **	0.8	2.23	1.52	3.25
	IA* **	0.89	2.43	1.58	3.7
	ID*	0.72	2.05	1.15	3.61
	IL	0.27	1.32	1	1.73
	IN*	0.31	1.36	1	1.84
	KS	0.48	1.61	0.98	2.59
	KY	0.52	1.68	0.99	2.82
	LA	0.19	1.21	0.81	1.8
	MA* **	1.06	2.89	2.01	4.15
	MD	−0.37	0.69	0.43	1.07
	ME	1.06	2.89	0.89	8.87
	MI* **	0.92	2.51	1.93	3.25
	MN	0.46	1.59	0.93	2.67
	MO * **	1	2.71	2.12	3.46
	MS*	0.43	1.53	1.08	2.14
	MT	0.79	2.2	0.62	6.97
	NC*	0.32	1.38	1.03	1.85
	ND	−0.44	0.65	0.03	4.69
	NE*	0.61	1.84	1.04	3.2
	NH	0.19	1.21	0.06	7.72
	NJ	0.35	1.42	0.84	2.3
	NM* **	0.72	2.06	1.34	3.13
	NV	−0.03	0.97	0.64	1.45
	NY* **	1.32	3.73	2.9	4.79
	OH* **	0.76	2.14	1.68	2.73
	OK* *	0.49	1.63	1.14	2.32
	OR	0.07	1.08	0.72	1.58
	PA* **	1.25	3.49	2.78	4.39
	PR	−0.18	0.83	0.24	2.27
	RI	1.2	3.33	0.57	16.58
	SC	0.1	1.1	0.75	1.6
	SD* **	1.04	2.82	1.76	4.5
	TN	0.25	1.28	0.89	1.82
	UT* **	1.13	3.11	1.73	5.65
	VA* **	−0.6	0.55	0.38	0.78
	VT* *	0.91	2.48	1.26	4.73
	WA	0.21	1.23	0.84	1.79
	WI* **	0.95	2.59	1.92	3.49
	WV*	0.5	1.65	1.01	2.65
	WY	0.71	2.03	0.39	8.97

*significance < 0.05;

* * significance < 0.01;

* **  significance < 0.001

4. Discussion

This study sought to characterize, at an individual level, the availability and distribution of treatments and other supports for substance use disorders among people experiencing incarceration in U.S. prisons. Our results indicate that a significant portion of America’s prison population met the criteria for SUD prior to arrest. Compared to peers who did not meet these criteria, this group was more vulnerable across several measures, from housing to mental health. Less than a quarter received treatment or support of any kind for their substance use since their entry to prison. Even fewer received medication-based treatments like MOUD. What treatment was available was not equitably distributed, with disparities across racial, geographic and sentence-based lines, among others.

This study’s sociodemographic findings contextualize the heightened risk for post-release overdose experienced by incarcerated populations. By demonstrating that people with SUD in prison have higher rates of unemployment, homelessness, physical and mental health conditions and polysubstance use, this study confirms at the national level what smaller-scale studies have already shown (Gates et al., 2017, McNiel et al., 2005, Nolte-Troha et al., 2023, Bunting et al., 2020). Many of these factors are associated with an overall increase in overdose risk: housing instability is linked to a six-fold greater risk of opioid overdose (Yamamoto et al., 2019); unemployment, disability or history of prior incarceration all carry more than twice the risk of fatal opioid overdose (Altekruse et al., 2020); and mental health and SUD comorbidities are associated with much higher risk of nonfatal and fatal OD, with hazard ratios of 8.7 and 4.1, respectively (Keen et al., 2022). Taken together, these findings help explain why those returning from prison are at higher risk, while also highlighting other needs among the SUD group at a national level.

Our study also shows that, as of 2016, treatment for SUD was almost nonexistent in prisons, even as opioid overdose deaths were accelerating. Though past research has looked at treatment availability at the facility level (Neill-Gubitz et al., 2022, Nunn et al., 2009), our results paint a more granular picture. Just 13 % of people with SUD received treatment since admission, and less than 1 % received MOUD, despite strong evidence of its efficacy (Moore et al., 2019, Wakeman et al., 2020). Recent research suggests that this is largely due to lack of availability, with most facilities not offering MOUD, or only providing it for specialized populations (Nunn et al., 2009). Further, mechanisms for screening individuals for treatment need are also lacking. One study estimates that less than half of prisons screen for MOUD need, likely exacerbating the gap between availability as reported at the facility vs. individual level (Scott et al., 2021). Taken together, these findings indicate that treatments like MOUD were almost nonexistent at the individual level in 2016, a finding that should be followed up on with more current research.

Finally, this study found that, among the few people receiving prison-based treatment or other support, large disparities existed. Though many studies have looked at predictors of treatment outcomes in US prisons and jails (Messina et al., 2006, Welsh and McGrain, 2008, Casares-López et al., 2013), few have studied overall treatment distribution. One study documented lower treatment rates among self-identifying Hispanic individuals (Nowotny, 2015), but our study may be the first to show a similar disparity among non-Hispanic Black individuals. Past research indicates that racial / ethnic differences are unlikely to be due to differences in willingness to engage in treatment (Acevedo et al., 2012). The impact of prison staff is likely to play a role, with ground-level workers exerting a significant influence on who gets screened and treated, opening space for bias (Pfaff et al., 2023). These disparities are made more concerning by recent changes in overdose trends: mortality rates have stabilized for non-Hispanic Whites, but continue to grow for Hispanic, Native Americans, and non-Hispanic Black individuals (Spencer et al., 2024). Given these groups are overrepresented in carceral facilities (Nellis, 2021), the fact that they may also be less likely to receive treatment suggests this mortality gap may only widen.

Another disparity surfaced was the lower odds of treatment among people convicted of violent crimes. People serving time for violent offenses already have a higher risk of mortality post-release than peers (Kariminia et al., 2007). Though engagement in drug treatment during incarceration can have a protective effect (Lize et al., 2015), our research shows that these individuals are less likely to receive it. Several possible explanations exist for this disparity. At the system level, many facilities base service eligibility on security concerns, thereby outright disqualifying those with violent convictions (Lee, 2019). In addition, the higher-security facilities that house these individuals may offer less programming broadly (Hill et al., 2016), or could be slower to adopt new treatment methods. Ultimately, disparities by offense class mean that those at higher risk of post-release mortality had lower odds of receiving potentially life-saving treatments as of 2016.

The final notable finding from the controlled analysis related to state-level disparities in treatment. Compared to Texas, most states had higher odds of treatment. Why these regions rose to the level of significance is difficult to explain given the relative paucity of literature on prison-based SUD treatment availability at the state level (Homans et al., 2023, Scott et al., 2021). In the last decade, several states have begun offering MOUD across their prison systems, though the first to do so (RI) wasn’t until 2016 (Longley et al., 2023, Weizman et al., 2021). Though the survey represents data from several years ago, overall availability still seems to vary widely by state, with a 2023 survey showing that the majority of jails and prisons still do not offer these medications (Longley et al., 2024). Another study looking at community treatment rates among individuals referred from criminal-legal settings found significant disparities by states, indicating geographic variety in treatment outside of prisons (Donahoe et al., 2024). Ultimately, future research may contextualize these geographic disparities by examining how variations in facility-based treatment can be tied to other indicators, such as state policies, funding initiatives, rates of OUD, etc.

Our findings should be read in the context of several limitations. The data is from 2016, making it dated in a field where the treatment landscape changes rapidly (Longley, 2024, Dadiomov et al., 2022). In addition, the lack of variables on specific drugs – like fentanyl, which is frequently implicated in overdose deaths – reduced our ability to more accurately produce use-related statistics. There were also issues with treatment and support variables. For one, there was no indication of intensity, making it difficult to understand whether the SUD treatment services that individuals received were delivered properly. There were also no variables to clarify whether low treatment levels were due to uptake or availability issues. Though recent research shows high rates of treatment use when it is available (Treitler et al., 2022), variables on availability vs. uptake would have allowed us to more precisely locate the root of low treatment rates.

5. Conclusion

The findings of this study add a new level of texture to calls for expanded substance use treatment in carceral facilities. Though MOUD access has expanded in recent years (Weizman et al., 2021), this study points to disparities in treatment that may not be fully resolved by changes in facility-level access. More research is needed to understand and address disparities, especially since much of the treatment screening and referral process is mediated by individuals, so top-down pushes to expand treatment availability (e.g. lawsuits, policy changes) may not resolve the individual-level factors that introduce bias by race or offense class. Thoughtfully-crafted treatment expansion that considers existing disparities will be essential to stemming the growing disparities we see in overdose mortality among these vulnerable groups.

Ensuring that access is equitable will require careful monitoring and data collection to track and avoid possible disparities. Nationally-representative data like that captured in the SPI is an invaluable part of shaping programs and monitoring whether services are reaching those in need. Unfortunately, the Bureau of Justice Statistics, which once released the SPI every five to seven years, has slowed in the pace of their publications (Sawyer, 2019), and the new administration will likely only worsen the visibility barriers that already exist across government agencies (Fathi, 2010, Dolovich, 2022). The contributions of efforts such as the NIDA-funded Justice Community Opioid Innovation Network (JCOIN) underscore the severe under-funding of the National Institute of Justice and allied agencies that support criminal-legal system research. In addition to pushing for expanded access to MOUD, interested parties should also lobby for better research on the needs they’re trying to meet. So long as overdose deaths remain one of our nation’s leading causes of death (Drug, 2024), data on their potential causes – and where people fall through the cracks – are essential.

CRediT authorship contribution statement

Chen Samuel: Writing – review & editing, Writing – original draft, Visualization, Project administration, Methodology, Formal analysis, Conceptualization. Pollack Harold A.: Writing – review & editing, Writing – original draft, Project administration, Methodology, Conceptualization. Salisbury-Afshar Elizabeth M.: Writing – review & editing, Conceptualization. Pho Mai T.: Writing – review & editing, Writing – original draft, Supervision, Project administration, Methodology, Conceptualization.

Funding

Research was funded by University of Chicago Pritzker School of Medicine Summer Research Program (SRP) funding available to first-year medical students. The specific grant was made possible by alumnus Ken Ouriel, MD’81, and his wife Joy Bracker, MST’81. Article processing charges were covered by the following NIDA grants: NIDA 3UG1DA050066 NIDA 1R01DA057665.

Author agreement statement

We the undersigned declare that this manuscript is original, has not been published before and is not currently being considered for publication elsewhere.

We confirm that the manuscript has been read and approved by all named authors and that there are no other persons who satisfied the criteria for authorship but are not listed. We further confirm that the order of authors listed in the manuscript has been approved by all of us.

We understand that the Corresponding Author is the sole contact for the Editorial process.

He/she is responsible for communicating with the other authors about progress, submissions of revisions and final approval of proofs

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.