The Impact of Policy Changes on Heroin and Nonmedical Prescription Opioid Use among an Incarcerated Population in Kentucky, 2008-2016

Abstract

In response to the opioid epidemic, there have been several national and state-level policies enacted. Consideration of how criminal justice-involved individuals are affected by such policies has received limited attention, despite disproportionately higher use among this population. Bivariate statistics examined yearly trends and logistic regressiosns examined demographic correlates of nonmedical prescription opioid and heroin use among Kentucky inmates over an eight-year time-span of important national and local policy changes (n=34,542). Results indicate that among incarcerated individuals, prior use of heroin increased 204% from 2008 to 2016, with increases possibly linked to key policy changes associated with OxyContin reformulation and state implementation of a prescription drug monitoring program. The current incarcerated population had more severe use patterns when considering general population research. Consideration of criminal justice-involved populations is crucial to understanding and treating the opioid epidemic.

Introduction

The use of nonmedical prescription opioids (NMPO) and heroin have reached prevalence rates emblematic of a national crisis, as demonstrated by increases in incidence rates, emergency room visits, treatment admissions, and overdose fatalities (Cai, Crane, Poneleit, & Paulozzi, 2010; Martins et al., 2017; Paulozzi, 2012; Slavova et al., 2017) Drug overdose is now the leading cause of accidental death for individuals under the age of 50 (Ahmad, Rossen, Spencer, Warner, & Sutton, 2017). In 2016, opioids were involved in 66.4% of all overdose mortalities for a total of 42,249 lives lost (CDC, 2017). Societal costs of opioids are estimated between $55 billion to $1 trillion, owing to health care costs, lost earnings from premature deaths, and criminal justice costs (Altarum, 2018; Birnbaum et al., 2011).

The increase in use and adverse outcomes have been primarily described as a progression of initiation to abuse – from NMPO to heroin – and this progression has been explained by a host of factors. The factors of most interest to the current study include the unintended consequences of federal- and state-level policies (Cicero, Ellis, Surratt, & Kurtz, 2013; Muhuri, Gfroerer, & Davies, 2013; Haegerich, Paulozzi, Manns, & Jones, 2014;). To that end, the current research seeks to explore trends in NMPO and heroin use in the context of national and state-level policies among a population of justice-involved persons – a vulnerable population that has been underrepresented from national conversation and data related to the opioid crisis.

The Opioid Landscape in Kentucky

Appalachian states have suffered extensively from negative outcomes related to opioid misuse. West Virginia, Ohio, and Kentucky ranked among the top ten states for overdose mortalities (Ahmad et al., 2017). For Kentucky, the rate of 33.5 deaths per 100,000 persons is nearly double the national rate (CDC, 2018a). Further, nonfatal opioid overdoses significantly increased in Kentucky from 2016 to 2017 (CDC, 2018b). A national vulnerability assessment found nearly 25% of the 220 identified counties to be at-risk for rapid dissemination of an HIV outbreak due to injection drug use (IDU) practices are located in Kentucky (Van Handel et al., 2016). In response to the opioid crisis, Kentucky has enacted several state-level policies to expand treatment efforts, implement a prescription drug monitoring program (PDMP), and increase diversion-related criminal penalties (Slavova et al., 2017; Victor, Walker, Cole, & Logan, 2017). These policies build on national level changes, such as reformulation of OxyContin.

In 2010, the US Food and Drug Administration (FDA) designated OxyContin as an “abuse-deterrent” drug owing to the reformulation of the drug which made it more difficult to crush, snort, or inject. This policy, along with the state-level interventions that included implementation of a PDMP, were initiated to make NMPO more difficult to misuse or divert and to increase the regulation of the distribution of NMPO through a supply-side drug policy approach. Supply-side policy interventions are concerned with the reduction of availability of substances to deter use and abuse. However, a recent study by Alpert and colleagues (2018) demonstrated that the disruption to NMPO supply caused by the OxyContin reformulation caused large and significant increases in heroin overdose deaths.

Other supply-side policy interventions include PDMPs. PDMPs collect data on prescribed controlled substances so that physicians (Beletsky, 2018) and in some cases other health and criminal justice officials, may review a patient’s prescribing history in order to identify potential diversion or misuse of NMPO. Research has indicated that required PDMPs, as opposed to merely enacted programs, may reduce misuse of NMPO (Buchmueller & Carey, 2017). Kentucky House Bill 1 (H.B. 1 2012; amendment H.B. 217), bolstered the state’s PDMP, which mandated close regulation of pain management clinics and the implementation of new practice standards for those prescribing and distributing prescription opioids (PDMP Center of Excellence, 2016). However, recent research demonstrates that while H.B. 1 may have reduced the misuse of NMPO, it is also possibly associated with increasing rates of heroin use (Slavova et al., 2017; Victor et al., 2017).

Kentucky Senate Bill 192 (S.B. 192) was passed in 2015, which designates sources of funding for substance use treatment and authorizes expanded access to naloxone. The impact of S.B. 192 on opioid trends in Kentucky is not yet clear. Kentucky appears to mirror national-level trends that suggest many individuals who initiate use of heroin, do so at points indicative of policy reform aimed at reducing the NMPO epidemic (Compton, Jones, & Baldwin, 2016; Muhuri et al., 2013; Slavova et al., 2017; Victor et al., 2017). However, increased access to treatment, including the provision for pre-release use of injectable naltrexone as designated in S.B. 192 could possibly decrease heroin use compared to previous policies which were focused on restricting access to NMPO through supply-side interventions.

Opioid Use among Incarcerated Populations

Despite substance use rates that are disproportionately higher than general populations (Belenko, Hiller, & Hamilton, 2013), incarcerated populations have received limited attention in the literature regarding how recent drug policy efforts may have affected their drug use patterns. Upon reentry, criminal justice-involved individuals are approximately 129 times more likely to die of a drug overdose compared to the general public (Binswanger et al., 2007; Dumont, Brockmann, Dickman, Alexander, & Rich, 2012). The heightened risk of mortality reflects the volatility of circumstances in the days following release, the period of abstinence from substances while incarcerated, and the initiation of high-risk behaviors – such as using opioids with reduced tolerance.

In a national sample, more than half of individuals who reported opioid use had past-year criminal justice contact (Winkelman, Chang, & Binswanger, 2018) indicating the inclusion of justice-involved persons in responses to the opioid crisis is essential. Among a sample of African American justice-involved individuals, NMPO use was estimated at 23.3% in 2014, a rate that significantly increased from 17.6% in 2010 (Knighton, Stevens-Watkins, Staton, & Pangburn, 2018). The same study found that older age was a significant protective factor of NMPO use; whereas, greater years of education and recurrent mental health symptoms increased the odds of NMPO use (Knighton et al., 2018). Despite this work, it remains unclear how justice-involved individuals have been comparatively impacted by the opioid crisis based on gender, race, age, and geography.

Current Research

A recent study by the CDC suggested that research aimed at examining NMPO and heroin use trends among institutionalized populations is needed to accurately understand drug use in ways that are representative of the U.S. general population (Jones, Logan, Gladden, & Bohm, 2015). The current research responds to this call by examining NMPO and heroin use trends that coincide with recent national (2010, OxyContin reformulation)- and state (2012, PDMP; 2015, expanded treatment)-policy initiatives to curb the use and diversion of both substances among a criminal justice sample in a state with high opioid use. The purpose of this research is to 1) analyze NMPO and heroin trends during key policy years from 2008 to 2016 among an incarcerated population and; 2) determine demographic correlates of NMPO and heroin use among an incarcerated population.

Methods

Yearly trends of NMPO and heroin use were examined using secondary data collected from an ongoing evaluation, the Criminal Justice Kentucky Treatment Outcome Study (CJKTOS) between July 1, 2011, and June 30, 2017. Data in the current study is from incarcerated individuals in the FY2011 to 2017 cohorts and was restructured by incarceration year. Thus, all included individuals entered Kentucky corrections-based substance abuse programs (SAP) between 2011 and 2017 but were included in analyses only if they began their incarceration between 2008 and 2016. Individuals could only be in one study time-period (i.e., there are no duplicates across years). This time frame includes the three-key time periods associated with policy changes (2010, OxyContin reformulation; 2012, PDMP; 2015, expanded treatment).

CJKTOS data is collected when individuals enter SAP, which are based on a modified therapeutic community modality. In the first two weeks of an individual’s entrance to SAP, they are provided a baseline assessment by treatment providers. The assessments are available through a web-based computer assisted interviewing program. The information analyzed in the current analyses rely on self-report portions of the assessment. SAP is available to any individual with a history of substance use, who is within 24 months of their release date, but still had 6-months remaining sentence before parole or release. Given the aims of the current paper to focus on incarcerated individuals, the sample is limited to only prison and jail participants, excluding 102 (0.29%) of the sample who were in community custody programs. Participants’ submission of the data to the research team was part of their consent to treatment in accordance with Department of Corrections standards.

The data were collected by treatment providers in electronic format using a web-based program and uploaded to University of Kentucky secure servers. All data were collected and stored in compliance with HIPAA regulations, including the use of encrypted identification numbers and abbreviated birthdays (month and year) to secure the confidentiality of protected health information.

Measures

The dependent variables of interest were heroin and NMPO use in the 12 months prior to incarceration. NMPO use included any form of opioids not prescribed (e.g. morphine, Percocet, oxycodone, OxyContin, Lortab, hydrocodone, Dilaudid, Opana). Specifically, heroin use was measured by asking participants, “In the 12 months prior to this incarceration did you use heroin (e.g. smack, H, junk, skag)?” Use of NMPO was measured by asking participants, “In the 12 months prior to this incarceration did you use any form of opioids that were not prescribed for you (e.g. morphine, Percocet, oxycodone, Oxycontin, Lortab, hydrocodone, Dilaudid, opana)?” The dependent variables were coded as dichotomous with 1=“Yes” and 0=“No”. Independent variables include demographic variables of gender (1=male), race (1=white), and age (continuous). Transgender was excluded from analyses due to low sample size (n=14). Rural and urban variables were based on the county of conviction that participants reported at baseline, classified with the National Center for Health Statistics’ urban-rural classification scheme (Ingram & Franco, 2014). IDU was measured dichotomously (1=yes) as behavior 12-months prior to incarceration.

Analytic Approach

First, prevalence of heroin and NMPO use by year were examined via descriptive statistics and ANOVA analyses (Table 1). In addition to the original variables detailed in the methods section, rates per 100 persons were calculated utilizing a standard epidemiological formula for incidence rate to account for differences in population sizes each year. Percent change from the year prior was also calculated.

Table 1.

Trend Analysis of Heroin and NMPO Use (N=34,542)

	2008	2009	2010	2011	2012	2013	2014	2015	2016
N	1,676	3,483	4,759	4,905	4,919	5,256	4,487	3,551	1,506
Heroin Use (N)	188	401	590	854	1,211	1,544	1,411	1,153	514
Heroin Use Rate (per 100)	11.22	11.51	12.40	17.41	24.62	29.38	31.45	32.47	34.13
P from year prior use rate	--------	.81	.33	<.001	<.001	<.001	.01	.27	.19
Heroin Rate Change (%)	--------	2.58	7.73	40.40	41.41	19.33	7.04	3.24	5.11
NMPO Use (N)	772	1,674	2,339	2,410	2,357	2,332	2,004	1,684	698
NMPO Use Rate (per 100)	46.06	48.06	49.15	49.13	47.92	44.37	44.66	47.42	46.35
P from year prior use rate	--------	.18	.33	.99	.22	<.001	.77	.01	.48
NMPO Rate Change (%)	--------	4.34	2.27	−0.04	−2.46	−7.41	0.65	6.18	−2.26
IDU Rate (per 100)	31.98	33.85	34.92	37.80	42.35	44.84	47.56	52.07	51.53
P from year prior use rate	--------	.20	.32	.004	<.001	.01	.01	<.001	.72

Note: Significance from year prior determined through ANOVA analyses.

Second, descriptive statistics were examined by incarceration year for the CJKTOS population overall as well as among users of heroin and NMPO (Table 2). Third, multivariate logistic regression was utilized to examine significant yearly trends of heroin and NMPO (Tables 3 & 4). Year was included as an independent variable with 2008 serving as the reference. Then, interaction effects of year with gender, race, rural, and age (all demographic variables) were examined. Significant effects of interactions indicated a significance beyond main effects, to allow for consideration of yearly trends. Results of the logistic regression report the adjusted odds ratios, model chi-square, and pseudo R². Tests for collinearity revealed no issues. Data analyses were conducted using Stata/SE version 15.1.

Table 2.

Demographics (N=34,542)

	2008 (n=1,676)	2009 (n=3,483)	2010 (n=4,759)	2011 (n=4,905)	2012 (n=4,919)	2013 (n=5,256)	2014 (n=4,487)	2015 (n=3,551)	2016 (n=1,506)	Total (n=34,542)
Female	9.01	11.97	12.08	15.21	17.42	17.81	16.63	17.54	17.13	15.37
Male	90.99	88.03	87.92	84.79	82.58	82.19	83.37	82.46	82.87	84.63
Race (%)
White	71.84	73.50	73.52	72.68	64.20	48.48	55.61	83.84	85.26	67.43
Non-White	28.16	26.50	26.48	27.32	35.80	51.52	44.39	16.16	14.74	32.57
County (%)
Rural	48.93	49.70	49.42	47.99	46.21	43.76	46.67	47.93	44.49	47.18
Urban	51.07	50.30	50.58	52.01	53.79	56.24	53.33	52.07	55.51	52.82
Age ($\stackrel{-}{x}$)	34.04	33.98	33.73	33.55	33.34	33.74	34.06	34.27	33.85	33.79

Table 3.

Logistic Regression of Associations and Trend with NMPO Use by Year of Incarceration (N=34,542)

	Model 1	Model 2	Model 3	Model 4	Model 5
Heroin Use	3.13***	3.13***	3.10***	3.13***	3.16***
Male	0.75***	0.78	0.76***	0.75***	0.75***
White	1.62***	1.61***	2.51***	1.62***	1.62***
Rural	1.71***	1.71***	1.67***	1.68***	1.71***
Age	0.98***	0.98***	0.98***	0.98***	0.96***
IDU	2.07***	2.07***	2.02***	2.07***	2.08***
Year1
2009	1.05	1.25	1.09	1.05	1.03
2010	1.08	1.37	0.98	1.11	0.96
2011	1.00	1.18	1.13	1.03	0.58*
2012	0.88*	0.94	1.30*	0.81*	0.45***
2013	0.75***	0.80	1.25	0.74***	0.39***
2014	0.70***	0.63*	1.26*	0.69***	0.23***
2015	0.66***	0.50***	1.08	0.65***	0.28***
2016	0.62***	0.52***	0.88	0.63***	0.35***
Interaction effect	None	Year X Gender	Year X Race	Year X Rural	Year X Age
2009	--	0.83	0.95	1.00	1.00
2010	--	0.76	1.13	0.95	1.00
2011	--	0.82	0.86	0.95	1.02*
2012	--	0.91	0.59***	1.16	1.02**
2013	--	0.93	0.46***	1.05	1.02**
2014	--	1.12	0.41***	1.01	1.03***
2015	--	1.41	0.54***	1.03	1.02***
2016	--	1.22	0.63*	0.95	1.02
Constant	0.84***	0.82***	0.61***	0.84***	1.44***
$\mathrm{M}\mathrm{o}\mathrm{d}\mathrm{e}\mathrm{l}{x}^2$	5456.66***	5489.46***	5645.65***	5464.77***	5520.19***
Pseudo${\mathrm{R}}^2$	0.11	0.11	0.12	0.11	0.11

Note: Odds Ratios (OR) reported;

^*indicates significance at p<.05;

^**p<.01,

^***p<.0001

¹2008 is excluded baseline year.

Table 4.

Logistic Regression of Associations and Trend with Heroin Use by Year of Incarceration: (N=34,542)

	Model 6	Model 7	Model 8	Model 9	Model 10
NMPO Use	3.30***	3.30***	3.28***	3.31***	3.31***
Male	0.97	1.20	0.97	0.97	0.97
White	1.14***	1.14***	1.50	1.14***	1.14***
Rural	0.34***	0.34***	0.33***	041***	0.34***
Age1	0.97***	0.97***	0.97***	0.97***	0.99
IDU	8.87***	8.87***	8.61***	8.87***	8.87***
Year1
2009	0.96	1.15	0.65	1.03	1.10
2010	1.00	0.99	1.09	1.01	2.06
2011	1.54***	1.86**	1.79*	1.63***	4.35***
2012	2.40***	3.23***	3.08***	2.62***	5.68***
2013	3.32***	4.13***	4.59***	3.74***	7.32***
2014	3.72***	4.55***	5.37***	4.13***	8.72***
2015	3.39***	4.11***	2.52***	3.85***	5.59***
2016	3.62***	4.16***	3.12***	4.20***	4.27***
Interaction effect	None	Year X Gender	Year X Race	Year X Rural	Year X Age
2009	--	0.82	1.55	0.85	1.00
2010	--	1.03	0.91	0.98	0.98
2011	--	0.81	0.84	0.88	0.97**
2012	--	0.71	0.74	0.82	0.97*
2013	--	0.78	0.64	0.76	0.98*
2014	--	0.79	0.6	0.79	0.97*
2015	--	0.81	1.67	0.75	0.98
2016	--	0.86	1.16	0.71	0.99
Constant	0.09***	0.07***	0.07***	0.08***	0.04***
$\mathrm{M}\mathrm{o}\mathrm{d}\mathrm{e}\mathrm{l}{x}^2$	10347.97***	10353.79***	10399.84***	10356.89***	10370.33***
Pseudo${\mathrm{R}}^2$	0.28	0.28	0.28	0.28	0.28

Note: Odds Ratios (OR) reported;

^*indicates significance at p<.05;

^**p<.01,

^***p<.0001

¹2008 is excluded baseline year.

Results

Heroin use rate among the current population increased from 11.2 per 100 individuals in 2008 to 34.1 per 100 in 2016, with years 2011–2014 having statistically significant increases from prior years (Table 1). Since 2010 (i.e., year of Oxycontin reformulation), rates of use have continued to increase but growth is tapering, such that from 2011 to 2012 rates increased 41.4% and the most recent years (2015 to 2016) only saw a 5.1% increase.

Since 2008, NMPO use rates have remained consistently greater than heroin use rates (Figure 1). It is notable that the only significant drop in NMPO rates occurred between 2012 and 2013 (−7.41%; p<.001). The trend of injecting behavior somewhat mirrors that of heroin use, although at a greater rate (range 31.98–51.53 per 100) (Figure 1). Injecting behavior also follows a similar significance trend as heroin with significant changes in bivariate tests each year from 2011 to 2015.

Figure 1.

Trends in heroin, NMPO, and injection drug use among an incarcerated sample: Criminal Justice Kentucky Treatment Outcome Study;2008–2016 (n=34,542)

Note: NMPO= nonmedical prescription pain reliecer IDU=injection drug use; Rate is per 100 persons

Grey bars indicate significant policy reforms: 2010, Reformulation of OxyContin; 2012; KY H.B. 1, implements a prescription drug monitoring program; 2015, KY S.B. 192, expands access to treatment and naloxone

Independent Correlates of Heroin and NMPO Use

Table 2 contains demographic information for the study sample per year. In certain years, disproportionate rates of individuals who identified as a race other than white were enrolled in SAP (i.e., 2008–2014) whereas more recent years (i.e., 2015–2016) the rates of white and non-whites were more comparable to general population demographics in the state of Kentucky. The rate of females increased from 9.0% in 2008 to 17.1% in 2016. The sample was nearly split between urban and rural residency with slightly more individuals enrolled from urban areas each year. The mean age remained relatively stable across years with the average respondent being about 33 years old.

Odds ratios for associations between heroin use and NMPO use by year are presented in Table 3 and 4, respectively. Individuals who reported using heroin were more likely to report use of NMPO (Table 3). Further, significant associations were found among gender such that males were at decreased odds of using NMPO. Race was also significantly associated with use of NMPO in that individuals who identified as white were more likely to use NMPO. The likelihood of using NMPO was increased if an individual lived in a rural location or reported IDU. Older individuals were less likely to have used NMPO compared to younger persons.

Model 1 (Table 3) indicates that beginning in 2012 NMPO use statistically significantly declined. In 2012 the odds of using NMPO were reduced by 12% and by 2016 odds of using NMPO had reduced by 38%. There were no significant yearly trends associated with gender (Model 2) or rural location (Model 4). Significant yearly trends were found for race (Model 3) and age (Model 4). From 2012 to 2016, individuals who identified as white were more likely to use NMPO compared to those of other races. The magnitude of this probability of use decreased as the white-to-non-white ratio decreased significantly from 2012 to 2016 (see interaction effects).

Regarding age, from 2011 to 2015 older individuals were less likely to use NMPO; however, this trend increased such that the strength of the relationship between older age and NMPO use weakened. That is, age was a stronger protective factor from 2008–2010, such that for every year older an individual was, the likelihood of NMPO use decreased by 4%. This relationship weakened in later years, as individuals were only 1–2% less likely to use NMPO with increasing age (OR: 1.02–1.03). Overall, the models were statistically significant and provided a good fit for examining NMPO use among the current population.

Examining heroin use (Table 4, Model 6), individuals who reported using NMPO were statistically significantly more likely to report also using heroin. Gender was not a statistically significant predictor of heroin use. Race was a significant predictor, such that individuals who identify as white were more likely to use heroin. Rurality and older age were both protective factors against heroin use. Additionally, IDU was significantly associated with heroin use.

Examination of yearly trends in Model 6 indicate from 2011 to 2016 the likelihood of heroin use increased significantly. In 2011, individuals were 54% more likely to use heroin and this peaked in 2014 at an 272% increased probability. There were no significant yearly trends for gender, race, or rurality. However, age offered a pronounced protective effect from 2011 to 2014, such that older individuals were less likely to use heroin during these years (Model 10). Overall, the models examining heroin use were statistically significant and provided a good fit for examining heroin use among the current population.

Discussion

The present study examined heroin and NMPO use trends among a justice-involved, substance abuse treatment population over an eight-year period that includes changes in national and state policies designed to reduce the misuse and diversion/distribution of opioids. Although several studies have investigated the associations between NMPO use and heroin use among community- or treatment-based samples, much of the literature has excluded justice-involved populations. The findings of the current study have implications for addressing opioid-related use patterns and policy effects among the vulnerable population of justice-involved populations.

Although multivariate models indicate NMPO use to be on the decline among this justice-involved sample, heroin use has escalated significantly since 2011, the first year of reported use after the 2010 OxyContin reformulation. Prior studies suggest it is common to shift to heroin use after NMPO (Jones et al., 2015) and thus this finding is not unique to the current justice-involved population. What policymakers should consider, however, is that while general populations appear to be deterred- and remain deterred- following policy reforms such as 2010 OxyContin reformulation or 2012 PDMP (Kennedy-Hendricks et al., 2016; Victor et al., 2017), justice-involved populations use of NMPO remains high and yearly reductions in use appear to be shrinking.

Unlike previous policies, which demonstrated clear trend changes following enactment, KY S.B. 192 (2015, expanded treatment) does not appear to have had an impact on trends in the current criminal justice sample except that heroin, NMPO, and IDU rates did not significantly increase or decrease after its implementation. However, the probability of use in multivariate models indicates that heroin use continued to increase in 2016, despite the 2015 implementation. Policymakers were hopeful that this policy will be associated with a decrease in heroin use by expanding access to substance use treatment and harm-reduction resources (e.g., access to naloxone, needle exchange programs). It may be that roll out of treatment resources has slower effects compared to supply-side interventions effects which may be felt more immediately.

Considering demographic associations, race was a significant predictor of NMPO use and heroin use. In part, the current findings were consistent with others that show opioid use is most prevalent in white populations (Cicero et al., 2014). However, this effect was diminished in NMPO use from 2012– 2016, as the racial gap in NMPO use began to decrease. Knighton and colleagues (2018) found increases in NMPO use across the same years studied (i.e., 2010–2014) among justice-involved African American men. Their findings, along with the diminishing magnitude of the racial trends reported in this paper, suggest that the opioid crisis is multiracial among justice-involved populations. Yet, it remains unclear to what degree the unique social and structural determinants faced by justice-involved populations impact drug use patterns and how these determinants may differ by race. Future research should examine racial differences in opioid use to determine if the trends observed in the current research hold in other states and in the years to come.

In the current research, age offered a protective relationship regarding NMPO and heroin use. As the age of participants increased it became less likely that they would report use of opioids. This effect was significant beginning in 2011 for both NMPO and heroin use. It may be that older individuals are more risk adverse and did not want to make the switch to heroin after OxyContin reformulation in contrast to their younger drug-using counterparts. Recent research using community-based samples has suggested that younger individuals are using heroin at greater rates and in some cases initiating heroin use prior to NMPO use – due in part to supply-side policy interventions (Ciccarone, 2017). Among a justice-involved sample in Australia, younger age was also associated with IDU and with high-risk injecting behaviors whilst incarcerated (Cunningham et al., 2018).

Studies utilizing the National Survey of Drug Use and Health, a non-institutionalized population study, have found males are more likely to use NMPO and heroin; although, males in the current study were less likely to use NMPO overall, and no significant differences were found for heroin when controlling for relevant covariates (Back, Payne, Simpson, & Bray, 2010; SAMHSA, 2017). To contextualize greater NMPO use among women, it is relevant to consider that compared to justice-involved men, justice-involved women are uniquely vulnerable, with complex histories of trauma, higher rates of substance use and health morbidities (Aday, Dye, & Kaiser, 2014; Jones, 2013; Salisbury & Van Voorhis, 2009). The fact that gender was a significant correlate of overall NMPO use without significant yearly trends reflect important gender differences in opioid use. Women have been found to have a ‘telescoping’ or accelerated initiation to substance use, including NMPO (Back, Lawson, Singleton, & Brady, 2011), and this may also help to understand subsequent criminal justice involvement.

Limitations

The current study is not without limitations. First, individuals were asked about their drug use only once they entered the substance abuse treatment program. While most individuals were incarcerated for only an average of 1.38 years before their assessment, some individuals were incarcerated for up to 9 years (R: .08–9), which could impact their recall about substance use the year prior to their incarceration. A total of 18 (<.01%) were incarcerated 8–9 years. Additionally, self-report data could be inaccurate due to recall, bias, or lack of rapport. However, research has generally demonstrated that self-report data are a key methodology for substance use behaviors and reviews have found self-report of drug use among substance-using populations to be valid for study (Darke, 1998).

Second, this study was conducted in a state that has been particularly affected by the opioid epidemic and the high rates of use found in the current study may not generalize to other institutionalized populations in states where opioid use is not as prevalent. Third, the study was unable to control for other factors that may influence trends, such as drug market availability and other individual factors. Fourth, the demographic analysis may have also been influenced by a fairly homogeneous sample (predominantly white men). The study adds to a growing literature-base that demonstrated the “whiteness” of the opioid epidemic, however, (Back et al., 2011; Binswanger et al., 2007) as well as the rurality of use (Hansen & Netherland, 2016). These findings should be considered within the racial and geographic dynamics of the Kentucky SAP which do not reflect national prison demographics. In certain years (e.g., 2015), the proportion of white to other races in SAP was reflective of Kentucky’s general population demographics. However, in other years (e.g., 2013) races other than white were overrepresented.

Conclusions

Based on the trends observed in the present study and in other studies, attempts to reduce substance use through policy alone is ineffective among criminal justice populations and may have unintended consequences, such as switching from a more highly regulated substance to another (Compton et al., 2016; Muhuri et al., 2013; Slavova et al., 2017). The results in this study highlight the drug use trends prior to incarceration, but they may also underscore the potential health risks that incarcerated individuals face upon community reentry and points for intervention among public health researchers and practitioners.

To counteract the harmful consequences of the opioid crisis, policy changes that increase the availability of substance use disorder treatment and harm reduction resources (e.g., syringe exchange programs, naloxone) in community- and correctional-settings are warranted. Maintenance therapy is under-utilized in the United States, especially in prisons and jails, despite strong evidence that such treatments are effective in treating opioid use disorder (Belenko et al., 2013; Keyes, Cerdá, Brady, Havens, & Galea, 2014).

The findings in the present study demonstrate that trends in the types of opioid use among individuals who enter incarceration with a history of substance abuse are distinctly different from general populations in terms of prevalence and gender (Compton et al., 2016; Slavova et al., 2017; Victor et al., 2017). Targeting intervention strategies toward individuals who are involved in the criminal justice system offers a unique opportunity since justice-involved individuals report higher rates of substance use disorders than the general public and have a period of time where they are frequently in contact with service provides either while incarcerated or on probation/parole (Binswanger et al., 2007). It is critically important that future research continue to assess the prevalence, trends, and demographics of these populations, in order to better inform public health initiatives and policy.