Multisubstance-Using Probationers and the Odds of Arrest While in the Community

Abstract

Failed drug tests commonly lead to technical violations and revocation hearings for probationers. The current study extends these observations by examining whether multisubstance-using probationers also have increased odds of arrest in the community. This is important as multisubstance-using probationers may present unique public safety challenges to community corrections agencies and require intensive treatment resources and additional monitoring. Using data from a county-level probation cohort (N = 2,257) from 2009 to 2010, a series of logistic regression analyses estimated the effects of multisubstance use on the odds of being arrested for a new offense while in the community. The findings revealed that multisubstance use and the frequency of multisubstance use increased the odds of arrest while on probation when compared with single-substance users. We discuss how agencies may best supervise multisubstance-using probationers and suggest directions for further examination.

Keywords

recidivism community corrections policy drug testing risk

Scholars have begun differentiating between individuals who intentionally blend and/or simultaneously use multiple illicit substances (i.e., multisubstance users) and those who use a single substance. Such a distinction stems from a growing body of literature demonstrating that multisubstance users are more likely to be impulsive and sensation seeking (Bechara, 2005; Bickel & Marsch, 2001; Conway et al., 2003; Galizio & Stein, 1983; Lacey & Evans, 1986; Preston et al., 2017; Quirk & McCormick, 1998; Sinha, 2008; Smith & Stoops, 2019). As a consequence, multisubstance users have different health, mental, and physical consequences, and motivations for substance use when compared with single-substance users. Research has demonstrated a host of negative associations related to multisubstance use, including having more nonfatal overdoses (Riley et al., 2016), seeking behavioral health care at higher rates (Calcaterra et al., 2015), having higher rates of mental health disorders (Salom et al., 2016), and being more prone to anxiety, hostility, and paranoia (Parrott et al., 2000). Furthermore, individuals who are experiencing acute life stressors are more likely to use multiple substances (Conway et al., 2003; Preston et al., 2017; Sinha, 2008; Smith & Stoops, 2019).

Changes in criminal justice policy over the past several decades have led to more substance-using and dependent individuals in the criminal justice system (Moore & Elkavich, 2008). The “War on Drugs,” beginning in the 1970s, prioritized aggressive law enforcement of controlled and illicit substances and brought increasing numbers of substance users into the justice system (Moore & Elkavich, 2008). Currently, substance users are more involved in the criminal justice system than nonsubstance users (Winkelman et al., 2018).

Understanding multisubstance use by probationers is important given the current policy shift away from incarceration and toward community-centered sentences. As more drug-dependent individuals enter community corrections, it is important to understand the effects of multisubstance use on this population. There is recent evidence that multisubstance users are becoming increasingly more common in community corrections populations (Denman et al., 2018; Hakansson et al., 2011; Winkelman et al., 2018). There is also evidence that multisubstance use in community correctional settings is associated with violent behavior, suicidal ideation, and cognitive issues (Hakansson et al., 2011). Furthermore, if allowed to remain in the community, this population can access illegal drug markets more easily and be exposed to law enforcement activity that intensely patrols known drug selling and using locations. This may lead to higher rates of arrest for multisubstance users. These factors are compounded by societal increases in opioid and novel psychoactive substance use over the previous decade (Smith & Stoops, 2019). The conditions of probation and the threat of punishment for violations can also exacerbate the stress placed on individual probationers. In turn, this may be a contributing factor for continued multisubstance use. As such, multisubstance use by probationers may be linked to increased public safety threats and recidivism.

Although researchers have documented the increase in multisubstance use in community corrections, there is still little known about what effects, if any, it has on supervision outcomes. The current study aims to clarify whether multisubstance-using probationers are more likely to recidivate (i.e., be arrested for a new offense) while in the community. A methodological contribution is to examine whether there is utility in distinguishing between single- and multisubstance users in understanding probation recidivism. A demonstrated link between multisubstance-using probationers and recidivism can help community corrections personnel focus resources on these individuals. Doing so may disrupt the cycle of criminal justice involvement that keeps individuals in the system for an extended length of time. Furthermore, we aim to clarify community correctional policy toward handling multisubstance drug users identified through drug testing. Multisubstance-using probationers may benefit from intensive intervention and treatment services in conjunction with traditional supervision.

Literature Review

Drug Use and Probation

Probation is the most common form of correctional punishment, with over 4.5 million persons under this form of supervision in 2016 (Kaeble, 2018). Drug use, abuse, and addiction among community correctional populations are high, as many probationers commit drug-related crimes and report prior drug use. In 2014, for example, approximately 25% of probationers were under community correctional supervision for a drug-related offense (Kaeble, 2018). Similarly, in a study comparing Illinois probationers with residents, the authors found that 43% of probationers compared with 13% of residents met Diagnostic and Statistical Manual of Mental Disorders (3rd ed., rev.; DSM-III-R; American Psychiatric Association, 1987) criteria for being drug dependent (Lurigio et al., 2003). In a national sample of male probationers, Feucht and Gfroerer (2011) found that 31% of persons on probation reported illicit drug or alcohol dependence or abuse in the past year.

Multisubstance use among probation populations is also a growing concern. Using a large nationally representative sample of the United States from 2015 to 2016, a recent medical study by Winkelman and colleagues (2018) demonstrated that polysubstance opioid users were significantly more likely to have been arrested and/or have probation sentences within the past 12 months when compared with nonopioid users. In a statewide study of probationers and parolees in New Mexico from 2004 to 2016, the authors found that 21% of their sample were multisubstance users (Denman et al., 2018). Furthermore, nearly half (42%) of those screened for substance use were multisubstance users (Denman et al., 2018). Although prior research on multisubstance users is sparse, there is a rising concern that multisubstance users are increasingly being sentenced to probation.

Research examining the relationship between any drug use and probation outcomes has shown that those with drug use histories are at an increased risk for recidivism. Olson and Lurigio (2000) examined probation outcomes for adult probationers in Illinois (n = 2,400) and found that probationers with a history of drug abuse were more than twice as likely to have a technical violation while on probation and were 60% more likely to be rearrested than probationers without drug abuse histories. In a study on the effects of drug use and treatment completion, Huebner and Cobbina (2007) found that 64% of their sample had a drug use history and were more likely to be arrested and failed more quickly while on probation.

While these relationships are complex, drug-using probationers may frequent illicit drug markets more often and are exposed to law enforcement activity that aggressively enforces drug laws (Kleiman et al., 2003). Furthermore, the negative effects of drug use can radiate out into communities (Boardman et al., 2001) by straining relationships (Fairbairn et al., 2018), taxing local resources (French et al., 2000), exacerbating physical and mental health problems (Hall, 2015), and decreasing residential and employment stability (Henkel, 2011). Many of these factors heighten law enforcement attention and can result in the incarceration of probationers along with a sizable number of other community residents (Clear et al., 2003; Harcourt, 2008; Meares, 1997). There is further evidence that drug-using probationers are more likely to recidivate and have their sentences revoked, which may result in subsequent incarceration (Gray et al., 2001; Olson & Lurigio, 2000). Probation agencies, therefore, have an interest in addressing drug use among their clients.

Probation agencies can address drug use by referring clients to substance treatment. This approach, however, can be expensive and rests on initial detection and identification of drug-using probationers. In addition, some recent work suggests that not all drug users need treatment nor meet the criteria for clinical drug dependence (DeMatteo et al., 2009; Marlowe, 2011; Werb et al., 2016). Another approach is to rely on abstinence through adherence to the rules and conditions of probation. The threat of punishment coerces probationers to abstain from drug use. This is a common condition of probation and makes drug testing possible. In accordance with deterrence theory, if drug testing is frequent and sanctions swift and predictable, then the assumption is that agencies can reduce drug use among clients without the use of treatment (Kleiman et al., 2003). Furthermore, if an agency identifies a severely dependent probationer through drug testing, then the agency can refer the individual to treatment. Mandatory drug testing, therefore, has become commonplace among community supervision agencies (Caudy et al., 2014; Friedmann et al., 2007). Although drug testing probationers is common practice, there are relatively few studies investigating the complex links between probationer drug use, let alone multisubstance use, and the increased risk of arrest while in the community.

Drug Testing and Probation Supervision Outcomes

Community correctional authorities commonly use drug testing for arrestees, pretrial detainees, probationers, parolees, and incarcerated individuals (Harrell & Kleiman, 2002). There are several purposes of drug testing those under correctional supervision, including, but not limited to, screening for recent use, identifying chronic users, and monitoring and deterring subsequent drug use (Harrell & Kleiman, 2002). As both prior and future drug use poses a challenge to the successful completion of probation, drug testing, typically in the form of urinalysis, is a common condition of probation.

There is evidence that probationers frequently fail drug tests, and that drug test failure can lead to technical violations, revocation, and incarceration. In a study examining the timing of probation violations, Gray and colleagues (2001) found that failed drug tests were the second most common form of probation violation, representing 22.4% of violations. Furthermore, the authors found that probationers with drug use histories were more likely to have subsequent technical violations. The results suggest that probationers with histories of drug use were more likely to be drug tested. Drug testing, in turn, then increased the risk of detection and subsequent technical violation (Gray et al., 2001).

Research on the specific relationship between drug testing and arrest shows a more complex relationship. In an experiment using pretrial defendants, the researchers randomly assigned them to receive drug testing prior to case disposition in two counties. The authors found varied effects. Britt and colleagues (1992) found that drug testing was associated with a slight decrease in pretrial failure in one sample, but drug testing was associated with a higher rate of pretrial failure in the second sample. Similarly, a study examining time to pretrial rearrest failed to reveal a relationship between drug testing and predicted pretrial misconduct (Rhodes et al., 1996). In another study of probationers, Deschenes and colleagues (1996) randomly assigned offenders to different supervision types: (a) standard with no drug tests, (b) standard with random drug tests, (c) standard with regularly scheduled drug tests (2 weeks), and (d) drug court with an integrated treatment. The authors found that the frequency of drug testing did not have a significant impact on arrest at 12 months (Deschenes et al., 1996). Other researchers, however, have demonstrated a link between drug testing and arrest. Olson and Lurigio (2000), for instance, found that probationers who tested positive for drugs during probation were 3 times more likely to be revoked, nearly 2 times more likely to be arrested, and nearly 3 times more likely to have a technical violation when compared with probationers who did not have a positive drug test. Although there is mixed evidence regarding the relationship between drug testing and arrest among probationers, there is some evidence that multisubstance use may affect the likelihood of being arrested.

Drug testing can be viewed as a double-edged instrument that reflects the conflicting goals of probation supervision, public safety, and rehabilitation (Whetzel et al., 2011). On one hand, drug testing is a way to increase the surveillance-focused goal of probation drug testing by monitoring individuals for continued criminal behavior. Alternatively, drug testing achieves the rehabilitation goal by identifying at-risk individuals that require treatment. Perhaps this explains the mixed relationship findings between arrest and drug use while on community supervision.

Current Study

There is some evidence that using multiple substances is linked to an increase in offending (Bennett & Holloway, 2005). To date, however, it is unclear whether multisubstance drug users are more or less likely to recidivate while under community supervision. Prior research has examined the relationship between drug use frequency and drug type, but no study has assessed probationer multisubstance use. This is an important oversight because probation violations can also result in a revocation hearing. In turn, violations are a common reason for incarceration or back-door sentencing practices (Lin et al., 2010). The primary focus of this research is to examine whether multisubstance use by probationers is related to arrest and whether multisubstance users should be considered a high-risk population that requires more treatment resources while on probation. Should multisubstance users and single-substance users differ in recidivism, then it supports disaggregating substance user types in future community correctional research. The current study seeks to answer the following research questions:

Research Question 1: Does multisubstance use increase the likelihood of arrest compared with single-substance use?

Research Question 2: Does the frequency of multisubstance use increase the likelihood of arrest?

Method

The study site is Philadelphia County, Pennsylvania, which has a combined county-level adult probation and parole agency. The agency oversees the supervision of adults sentenced to probation of any length and incarceration sentences (with a parole option) that are less than 2 years. A separate state parole agency supervises individuals sentenced to incarceration for more than 2 years.

Data on these probationers came from the supervising agency’s electronic case management system from 2009 to 2014. This system is central to the administration of supervision and record keeping and contains probationer information from several criminal justice agencies and the supervising officers. For instance, court records include the client’s offense that resulted in their probation sentence, community sentence length, and supervision conditions. The Philadelphia Police Department regularly sends arrest notices to the agency. New arrest data occurring in Philadelphia County are then updated in the case management system to alert supervising officers of offenders’ new probable criminal activity. The primary contributors and managers of the electronic case management system are the supervising probation officers. Officer notations include clients’ demographic and social features, in addition to several risk/needs assessments. The officers’ notations also record details about technical violations, like positive drug tests. Taken together, the case management system is an array of data relating to the clients’ background, behavior, social conditions, and interactions with criminal justice agents before and during supervision.

It is important to note that the probation agency structured client supervision during the study period around the clients’ risk of committing any crime within 2 years from the start of their community supervision. To do so, the agency used a custom risk assessment instrument that classified individuals into low-, moderate-, and high-risk supervision units (see Barnes & Hyatt, 2012, and Berk et al., 2009, for a complete description of the risk instrument). This is an important organizational feature because risk prediction structured the requirements for drug testing. During the study period, the agency assigned approximately 56% of clients to low-risk divisions that required twice a year officer–client contact but no drug testing. Conversely, the agency assigned approximately 12% of clients to high-risk supervision divisions. Officers supervised those individuals in small caseload units, similar to Intensive Supervision Probation practices (Latessa, Travis, Fulton, & Stichman, 1998; Morris & Tonry, 1991). Those individuals were required to drug test frequently by court order or discretion of their supervising officer. Per agency protocol, individuals supervised in high-risk units were drug tested frequently at the beginning of supervision but this frequency could be relaxed with several consecutive negative drug tests. Furthermore, officers had the discretion to order additional drug screenings, if there was information or suspicion of drug use. The agency assigned the remaining clients (32%) to moderate-risk divisions that allowed officers to make discretionary decisions about whether to drug test beyond what was required in a court order. Supervising officers in moderate-risk units had discretion over drug screenings based on circumstances, such as the nature of the instant offense or suspicion of drug use.

Study Sample

The agency initially provided data from the electronic case management system on all probation supervision cases generated from August 2009 through July 2010. Data from this 1-year cohort included all static and time-varying measures up to the clients’ case closure or through the follow-up date of August 15, 2014. The total number of cases generated during this 1 year was 25,052. After de-identifying and merging procedures, we restructured the data to group supervision cases by the probationer. This restructured data set contained each probationer’s earliest generated supervision case during the study period (n = 12,320). As we were interested in the effects of drug testing on the odds of arrest, we excluded low-risk clients because they were not required to be drug tested unless it was court ordered. We also excluded individuals supervised in specialized units (e.g., Sex Offender Unit) because their drug testing varied widely by unit and court-ordered conditions. Furthermore, individuals in these units represented unique subpopulations within the agency, based on their criminal offenses, court orders, and needs. We excluded parolees from analyses because it was not possible from these data to parse out parolees who received an incarceration sentence with immediate parole versus those who experienced a period of incarceration before their eventual release. This could have affected whether the individual received a drug test. Finally, we also excluded probationers who did not receive a drug test. The final sample (n = 2,257) included moderate- and high-risk probationers who received at least one drug test until their arrest or case closure.

It is also important to note that in the current study, a “case closure” is not equivalent to a sentence revocation. The final sample only included the probationer’s earliest supervision case, even if they had concurrent and consecutive cases. Our decision was to be parsimonious by excluding individuals who had their original sentences revoked and resentenced. Including those individuals could have skewed the results and led to erroneous conclusions. Case closures in the current study, therefore, is the termination of a supervision case that occurred for a variety of reasons, including, but not limited to, natural sentence expiration or revocation.

Dependent Variable

Arrest for a new offense

The outcome variable was the odds of a probationer being arrested by the police for a new offense in Philadelphia County (=1). An arrest represents interactions between the probationer and Philadelphia area law enforcement. An arrest, therefore, is different from a community corrections officer detaining a probationer for violating the rules and conditions of supervision, often labeled a “technical violation.” Specifically, an arrest occurs when the police take a probationer into custody for having enough probable cause that the individual committed a crime in Philadelphia. Police arrest data, incorporated into the probation/parole agency’s electronic case management system, indicated whether a client had been arrested during the supervision period and a municipal court docket was generated. An arrest for criminal activity while under community supervision is a common community-based correction measure of recidivism and is used as the dependent variable in this study as it poses a serious type of failure with a potential risk to public safety. If substantiated by a conviction, this represents a “direct violation” of the community supervision conditions and can result in a probation revocation hearing. This can result in prolonged criminal justice involvement for probationers and additional resources for the criminal justice system.

Independent Variables

Substance use type and frequency of multisubstance use

The agency screened for the presence of marijuana, cocaine, methamphetamine, benzodiazepine, phencyclidine, and opiates. Urinalysis screening occurred in the probation office and under the supervision of a lab technician. A third-party contractor analyzed and processed the urine samples, who then reported the results to the supervising officer through the electronic case management system. The number of administered drug tests was the number of urinalysis screenings given to the probationer during the study period or their arrest, whichever came first. The number of positive drug tests was the number of failed (i.e., tested positive for the presence of a substance) administered during the study period or first arrest.

Variables captured whether probationers were consistent single- or multisubstance users. A multisubstance drug user was a probationer who consistently tested positive for multiple substances or fluctuated between a single substance and multiple substances (marijuana, cocaine, methamphetamine, benzodiazepine, phencyclidine, and opiates) (=1) during the study period. Conversely, a single-substance drug user included individuals who consistently tested positive for a single substance across all drug tests (=1) during the study period. The reference group were probationers who were drug tested but never tested positive for the presence of any substance during the study period.

Dummy variables isolated the frequency of positive multisubstance drug tests during the study period. One positive multisubstance test included probationers who tested positive only once for multiple substances (=1). Two positive multisubstance tests included probationers who tested positive twice for multiple substances (=1). Three+ positive multisubstance tests included probationers who tested positive 3 or more times for multiple substances (=1). The reference category included probationers who did not test positive for multiple substances during the study period.

Client features

Prior research has demonstrated the importance of demographic characteristics of probationers on recidivism (Morgan, 1994). One consistent factor related to a higher likelihood of recidivism is being male (e.g., Langan & Levin, 2002; Olson et al., 2000). In the current study, we control for gender, with male probationers (=1). We also included the probationer’s age in years at the start of their probation sentence, as prior research has revealed that younger probationers are more likely to fail supervision compared with older probationers (Huebner & Berg, 2011). To control for the effects of racial and ethnic differences in outcomes, the current study included non-White (=1; African American, Asian, “Other,” and “Unknown” as categorized by the probation agency) and Latino (=1). The race and Latino ethnicity variables were self-reported by the probationer to the supervising officer. Employment status was also included and measured as whether the probationer was unemployed at the start of his or her probation sentence (=1).

Another important client feature was the predicted risk of recidivism. Prior research has demonstrated that risk assessment instruments are valid predictors of rearrests (Gendreau et al., 1996). The current study used the agency’s risk classification score developed by Berk and colleagues (2009) to classify clients into low-, moderate-, or high-risk supervision units. Calculation of the risk score accounted for the probationers’ criminal history (Berk et al., 2009). The current study examined clients in moderate- (=0) and high-risk (=1) units who were eligible for drug use screening.

Case and supervision features

Several aspects of the probationers’ instant offense were also included. Crime severity was captured by a measure the offense level, felony (=1), misdemeanor (=0), and the total number of charges associated with the criminal event. Furthermore, we included the offense type. These were coded as a series of dichotomous variables, including drug- (=1), violent- (=1), property-related (=1). The reference category included those who were convicted of a violent-related offense (=1). To control for the length of supervision, we included the probation sentence length in months. Generally, probationers with longer sentences have a higher risk of violating conditions of supervision. Finally, we included the number of open community supervision cases with the probation agency as a proxy for the level of involvement with the local court and community corrections agency.

Analysis Plan

Data were analyzed using logistic regression in Stata version 14. Although observed independent variables need not be normally distributed as an assumption of logistic regression, the distribution of their expected probabilities across the dependent variable should follow a likelihood function (distribution) that is linear and S-shaped (Cabrera, 1994; Menard, 2002). Outliers can influence this relationship between the independent and dependent variables, so we took precaution to minimize those effects. We winsorized (i.e., shortened the observed distribution tails by percentages) the following independent variables, the number of administered drug tests (3.37%; 52 to “25+”), the number of positive drug tests (1.15%; 34 to “15+”), the number of open supervision cases (0.44%; 12 to “6+”), the number of instant offense charges (0.80%; 18 to “9+”), and age (0.53%; 63 to “56+”), to reduce skewness. Prior to conducting multivariate analyses, we examined the variance inflation factors (VIFs) for the variables in our models. As a rule of thumb, VIFs over 10 are a sign of collinearity. The largest VIF in our models was having a drug-related instant offense that led to the probation sentence (VIF = 3.57), and the average VIF was 1.77, which gave us some reassurance that collinearity was not affecting these results. We also examined bivariate correlations to assess for collinearity among the predictor variables and did not detect any instances of multicollinearity in these data (see the appendix). We handled missing data from seven cases through listwise deletion. A test logit model included only the outcome measure arrest with no predictors, which was accurate in prediction 51% of the time (−2LL = −1,563.62).

Results

We first compared the final sample (n = 2,257) with the 1-year cohort of probationers (n =12,340) to compare demographic features, substance use, and recidivism. The entire cohort, which included low-risk and special unit probationers, was predominately male (80%), non-White (79%), and averaged 35.65 years of age. The full cohort was drug tested on average 1.86 times (SD = 5.00) during the study period. During this same time, the entire probationer cohort tested positive for substance use on average 0.68 times (SD = 2.02). Stated differently, about 21% tested positive for any substance use. In terms of recidivism, about 31% of the entire cohort had been arrested during the study period.

Table 1 shows the descriptive statistics for the final sample. The sample was also predominately male (85%), non-White (81%), and averaged 30.57 years of age. The final sample of moderate- to high-risk probationers averaged 6.24 drug tests (SD = 6.27). The study sample tested positive for substance use on average 2.37 times (SD = 3.02). About a third of sampled probationers (34%) were multisubstance users. Of all those who received a drug test, 18% tested positive once for multiple substances. About 15% of the sample’s multisubstance users had two or more positive multisubstance tests. Almost half (49%) of the probationers included in the analyses were arrested during the study period. Furthermore, about 38% of all multisubstance users were arrested. In short, the study sample received tests for substance use more often and failed drug tests more frequently compared with the full cohort. A higher proportion of the study sample was arrested compared with the cohort. The study sample appears to be a distinct subgroup from the larger 1-year cohort in terms of substance use and recidivism but somewhat similar in terms of demographic features. This makes sense considering that the study sample was composed of moderate- to high-risk probationers who were drug tested more frequently. The first research question addresses multisubstance within this sample.

Table 1.

Descriptive Statistics for Probationers in the Sample, 2009–2014 (N = 2,257).

Variable	M	SD	Minimum	Maximum	Skewness	Kurtosis
Arrest while on probation	0.49	0.50	0 = No	1 = Yes	—	—
Multisubstance user	0.34	0.47	0	1	—	—
Single-substance user	0.40	0.49	0	1	—	—
One positive multisubstance test	0.18	0.38	0	1	—	—
Two positive multisubstance tests	0.16	0.37	0	1	—	—
Three+ positive multisubstance tests	0.09	0.29	0	1	—	—
Number of administered drug tests	6.24	6.27	0	25+	1.47	4.35
Number of positive drug tests	2.37	3.02	0	15+	2.11	7.84
Drug instant offense	0.59	0.49	0	1	—	—
Weapon/other instant offense	0.17	0.38	0	1	—	—
Property instant offense	0.15	0.36	0	1	—	—
Number of open supervision cases	1.53	0.92	1	6+	2.23	8.93
Number of instant offense charges	2.22	1.61	1	9+	1.90	7.05
Latino	0.11	0.31	0	1	—	—
Nonwhite	0.81	0.39	0	1	—	—
Felony	0.24	0.43	0	1	—	—
Male	0.85	0.36	0	1	—	—
High risk	0.40	0.49	0	1	—	—
Age	30.57	9.51	18	56+	0.76	2.63
Unemployed	0.35	0.48	0	1	—	—
Sentence length (months)	22.59	18.02	1.03	121.77	1.88	8.11

Note. Except number of instant offense charges = 2,252.

Research Question 1 asked whether multisubstance use increased the odds of arrest compared with single-substance use. Table 2 reports the results of that logit model examining the effects of probationer multisubstance use on the odds of arrest. Global statistics initially indicated the data were a poor fit, so we removed the number of drug tests and the number of positive drug tests as predictors from this model. The improved fitting model explained about 6% of the variation in arrests—Hosmer–Lemeshow (df = 8) = 14.84 (ns); Pseudo R² = .06; Model χ² = 191.05, p < .001. Overall, being a multisubstance user increased the odds of arrest by a factor of two (odds ratio [OR] = 2.05, p < .0001) compared with probationers who never tested positive for substance use, and holding other factors constant. The results also suggest that multisubstance-using probationers were more likely to be arrested in the community compared with single-substance-using probationers. Although the OR was not as high, single-substance-using probationers were about 50% (OR = 1.52, p < .0001) more likely to be arrested in the community compared with nonsubstance users. There are several plausible explanations for these findings, including that multisubstance drug use increases the likelihood of aggressive and violent behavior in public situations that are more likely to be targeted for drug enforcement. A more likely explanation, however, may be that multisubstance users seek to purchase or use drugs in places that are more frequently patrolled and targeted by law enforcement even compared with single-substance users. Furthermore, multisubstance users may be more likely to be in possession of drugs or paraphernalia if searched by police, which may lead to an arrest.

Table 2.

Logistic Regression Predicting a Probationers’ New Arrest by Single- and Multisubstance Use, 2009–2014 (N = 2,250).

Variable	b	SE	OR	p < z	CI^a
Multisubstance user^b	0.720***	0.117	2.054	.001	[0.490, 0.949]
Single-substance user^b	0.418***	0.112	1.518	.001	[0.198, 0.638]
Drug instant offense^c	−0.014	0.169	0.986	.933	[−0.345, 0.317]
Weapon/other instant offense^c	0.196	0.190	1.217	.301	[−0.176, 0.568]
Property instant offense^c	0.378*	0.192	1.459	.049	[0.002, 0.753]
No. of open supervision cases	0.151*	0.065	1.163	.020	[0.023, 0.279]
No. of instant offense charges	−0.049	0.035	0.952	.154	[−0.117, 0.019]
Latino	0.041	0.147	1.042	.778	[−0.246, 0.329]
Non-White	0.155	0.116	1.168	.182	[−0.073, 0.384]
Felony	0.168	0.118	1.183	.154	[−0.063, 0.399]
Male	0.604***	0.133	1.829	.001	[0.343, 0.865]
High risk	−0.512***	0.099	0.599	.001	[−0.705, −0.318]
Age	−0.021***	0.005	0.980	.001	[−0.030, −0.011]
Unemployed	0.120	0.093	1.127	.197	[−0.062, 0.301]
Sentence length	0.017***	0.003	1.017	.001	[0.011, 0.023]

Note. Model χ²(16) = 191.05***; log likelihood = −1,463.31; Pseudo R² = .06; Area under curve = 0.67; Hosmer–Lemeshow χ²(8) = 14.84 (ns). OR = odds ratio; CI = confidence interval.

Lower and upper bound CIs.

Reference category = probationers who were drug tested but did not test positive for substance use.

Reference category = violent instant offense.

p < .05. **p < .01. ***p < .001.

A second logit model included the dummy variables for the number of multisubstance drug tests on the odds of arrest for probationers. Global fit measures indicated this model predicted about 6% of the variation in the outcome and the data were a good fit—Hosmer–Lemeshow (df =8) = 13.10 (ns); Pseudo R² = .06; Model χ² = 178.14, p < .001. The results, shown in Table 3, indicate that multisubstance use was related to increasingly higher odds of arrest to a point, holding other factors constant. The effect of having one positive multisubstance test increased the odds of arrest by nearly 50%, holding other factors constant (OR = 1.482, p < .001), compared with those who only tested positive for single-substance use or who never tested positive for substance use. The odds that a probationer would receive a new arrest increased with two positive multisubstance drug tests. The odds of a new arrest increased over 60% for probationers with two positive multisubstance drug tests (OR = 1.612, p < .01), when compared with single-substance and no substance users. These results may speak to the larger negative psychological effects of multisubstance drug use on an individual’s living and economic conditions and social relationships. The odds of arrest, however, were attenuated, for additional (three or more) positive multisubstance tests. Although we cannot say with these data, supervising officers may be more likely to refer chronic multisubstance users to treatment programs, thus reducing their likelihood of encountering police in the community.

Table 3.

Logistic Regression Predicting a Probationers’ New Arrest by the Number of Positive Multisubstance Use Drug Tests, 2009–2014 (N = 2,250).

Variable	b	SE	OR	p < z	CI^a
One positive multisubstance test^b	0.394***	0.120	1.482	.001	[0.158, 0.630]
Two positive multisubstance tests^b	0.477**	0.186	1.612	.010	[0.112, 0.842]
Three+ positive multisubstance tests^b	0.060	0.229	1.061	.795	[−0.389, 0.509]
Number of administered drug tests	−0.012	0.010	0.988	.220	[−0.031, 0.007]
Number of positive drug tests	0.011	0.020	1.011	.598	[−0.029, 0.050]
Drug instant offense^c	−0.027	0.169	0.973	.871	[−0.358, 0.304]
Weapon/other instant offense^c	0.177	0.190	1.194	.349	[−0.194, 0.549]
Property instant offense^c	0.357	0.192	1.428	.063	[−0.019, 0.732]
No. of open supervision cases	0.172*	0.069	1.188	.013	[0.036, 0.308]
No. of instant offense charges	−0.048	0.039	0.953	.214	[−0.125, 0.028]
Latino	0.032	0.146	1.033	.826	[−0.254, 0.319]
Non-White	0.194	0.117	1.214	.097	[−0.035, 0.424]
Felony	0.161	0.118	1.175	.171	[−0.070, 0.392]
Male	0.611***	0.133	1.842	.001	[0.351, 0.871]
High risk	−0.446***	0.107	0.640	.001	[−0.656, −0.236]
Age	−0.021***	0.005	0.979	.001	[−0.030, −0.011]
Unemployed	0.118	0.093	1.126	.201	[−0.063, 0.300]
Sentence length	0.001***	0.001	1.001	.001	[0.000, 0.001]

Note. Model χ²(18) = 179.41***; log likelihood = −1,469.13; Pseudo R² = .06; Area under curve = 0.67; Hosmer–Lemeshow χ²(8) = 12.73 (ns). OR = odds ratio; CI = confidence interval.

Lower and upper bound CIs.

Reference category = probationers who had no positive multisubstance tests.

Reference category = violent instant offense.

p < .05. **p < .01. ***p < .001.

These findings prompted additional exploratory analyses for older and female probationers as show in Table 4. Although substance types vary, an important predictor of multisubstance use is being younger in age (Beswick et al., 2001; Denman et al., 2018; Hakansson et al., 2011; Martinotti et al., 2009). We decided to examine the odds of arrest for older probationers by creating a dichotomous variable that categorized probationers into younger (18–34 years old) and older (35–56 years old) groups. We hypothesized that the effect size between polysubstance use would be weak or attenuated in the older age group (n = 685). We fit a logit model with just older probationers aged 35 to 56 years predicting the odds of arrest. Interestingly, the results showed that multisubstance use (OR = 1.93, p < .01) and single-substance use (OR = 1.69, p < .05) were both statistically related to the odds of arrest for the older age group compared with nonsubstance users. Substance use of any kind was related to recidivism across age type.

Table 4.

Logistic Regressions Predicting a Probationer’s New Arrest by Gender (N = 344) and Age (N = 685), 2009–2014.

Variable	Females only			Older only (ages 35–56)
Variable	OR	SE	p < z	OR	SE	p < z
Multisubstance user^a	2.457*	0.363	.013	1.926**	0.477	.008
Single-substance user^a	1.840	0.346	.078	1.691**	0.365	.015
Number of administered drug tests	0.944	0.034	.096	0.970	0.017	.091
Number of positive drug tests	0.999	0.068	.984	1.050	0.041	.213
Drug instant offense^b	0.947	0.548	.920	1.035	0.344	.919
Weapon/other instant offense^b	1.479	0.550	.477	1.776	0.639	.110
Property instant offense^b	2.296	0.573	.147	1.985	0.718	.058
No. of open supervision cases	1.068	0.121	.585	1.263*	0.149	.048
No. of instant offense charges	0.997	0.077	.972	0.940	0.059	.326
Latino	0.677	0.445	.381	1.151	0.333	.628
Non-White	0.904	0.291	.728	1.093	0.239	.685
Felony	0.858	0.478	.750	0.846	0.214	.508
Male	—	(omitted)	—	2.071***	0.464	.001
High risk	0.566	0.380	.135	0.626*	0.133	.027
Age	0.989	0.014	.454	—	(omitted)	—
Unemployed	0.979	0.253	.933	0.854	0.148	.363
Sentence length	1.035***	0.010	.001	1.017**	0.006	.002

Note. Female only model: Model χ²(16) = 43.28***; −2LL= −205.42; Pseudo R² = .09; Area under curve = 0.63; Hosmer–Lemeshow χ²(8) = 6.82 (ns).Older only model: Model χ²(16) = 60.14***; −2LL= −438.15; Pseudo R² = .06; Area under curve = 0.64; Hosmer–Lemeshow χ²(8) = 6.40 (ns). OR = odds ratio; LL = log likelihood.

Reference category = probationers who were drug tested but did not test positive for substance use.

Reference category = violent instant offense.

p < .05. **p < .01. ***p < .001.

Similarly, we fit a logit model using only female probationers (n = 344). There is evidence suggesting that females and males differ in motivation for multisubstance use and substance type, which can lead to gender varying outcomes (Beswick et al., 2001; Kendall et al., 1995; Lex, 1991). This was supported in these results showing that multisubstance-using females had more than 2 times the odd of arrest (OR = 2.46, p < .05). This risk, however, was attenuated for female single-substance users. Multisubstance-using females are more likely to be arrested compared with single-substance-using probationers. With these findings in mind, we turn to a discussion of the results.

Discussion

The current study examined the relationships between multisubstance use and arrest while under probation supervision. The results showed that multisubstance use while on probation increased the odds of recidivism. The frequency of multisubstance use also increased the odds of an arrest in the community, albeit there is much variation left in the outcome to be explained. Therefore, future research should explore other factors that are associated with an arrest that was not available in these data. Exploratory analyses also revealed age and gender risks in multisubstance use. In essence, the odds of arrest increase for multisubstance users across age groups and sex. In the future, these differences need to be examined further. Although not conclusive, our findings lend support to the complex relationship between drug use, crime, and law enforcement attention in the community. Unexpectedly, probationers who were on high-risk supervision consistently had a decreased odds of arrest in the community by about a third, net other effects. This is an intriguing finding, considering that these individuals were classified as being more at risk of recidivism. From a policy perspective, at least for this agency, this particular finding reveals a potentially useful approach for handling multisubstance users.

Within the sample, multisubstance- and single-substance-using probationers had a significantly higher likelihood of being arrested compared with nonsubstance users, with multisubstance users having the highest risk of arrest. Furthermore, high-risk offenders were less likely to be arrested while on probation. These findings have implications for probation policy. It may be plausible that the supervising agency is adequately assessing and monitoring the clients’ risk of recidivism by focusing resources and supervision efforts on the highest risk offenders. There may be value in classifying multisubstance users as a high-risk population. Thus, one policy recommendation could be to supervise multisubstance users in high-risk units. At the very least, multisubstance users could be monitored closely as an at-risk subpopulation that require intensive treatment. Future research, however, should examine whether or not this finding is generalizable.

Agencies should develop policies specifically aimed at addressing multisubstance use during supervision. Prior responses to drug use, in general, have included compulsory treatment (e.g., Farabee et al., 1998) and therapeutic jurisprudence, such as the use of drug courts (e.g., Gottfredson et al., 2003). There has been some support for the use of compulsory drug treatment, especially when coupled with frequent contact between officers and treatment staff (Young et al., 2004). There is also support for the use of the drug court treatment model on recidivism (Gottfredson et al., 2003); however, this approach requires diverting the correct defendants into drug court before sentencing. It is also unclear whether these strategies would be effective with multisubstance users. While not addressing multisubstance use specifically, Harrell and Roman (2001) did find that graduated sanctions coupled with judicial monitoring significantly reduced drug use and arrests in the sample they examined. In short, a tailored policy response to drug use may include treatment, additional sanctions, diversion, or combinations of these strategies. Although agencies across the country have adopted these policy approaches, it may be beneficial for agencies to continue investigating why multisubstance users are more likely to be arrested and who comprises this population.

The increased odds of arrest for multisubstance users may also be explained by the typologies of the drug–crime relationship outlined by Bennett and Holloway (2009). Three rationalizations emerged from interviewing drug offenders about their recent crimes. These offenders reported that (a) they felt compelled to generate money to support their habit, (b) they committed their crimes during an altered state, or (c) they committed a crime as part of a larger distribution and use process (Bennett & Holloway, 2009). Although their study did not focus on multisubstance users specifically, it is reasonable to assume that multisubstance use may compound these factors, increasing the risk of arrest. Future analysis should examine these typologies with a sample of multisubstance users to examine whether these remain relevant.

This finding also raises theoretical questions about the relationship between multisubstance drug use and crime. The current sample was time-ordered to include only detected multisubstance use before an arrest or case closure, but it is beyond the scope of this research to establish causation. The findings, however, do fit within the narrative examining whether criminality leads to drug use or whether drug use leads to criminality. There is support for a reinforcing relationship between criminal behavior and drug use. That is, we make no claims about the causal order of this relationship but we do observe that crime and drug use link in complex ways.

This study is not without its limitations. The results of the analyses reflect one county in the northeastern United States. Recidivism (i.e., arrests) data were only available within Philadelphia County; therefore, it is possible that all recidivism was not captured. As such, this research provides a starting point for subsequent investigations into the relationship between multiple drug use and arrest. The findings may also be useful to other similarly sized agencies or agencies with multisubstance-using probationers. Another limitation of the current study is that we did not include other factors related to success on probation. One such factor, for example, would be marital status (Morgan, 1994). Unfortunately, not all measures were available in these data. Related, some features are dynamic but we were only able to collect data at one time, so in essence, we are analyzing a “snapshot” of what was occurring at the time the agency provided the data. For example, a probationer’s employment status may change throughout supervision; however, we could not account for any changes in employment in our analyses. Future research should account for these potentially confounding and time-varying factors.

Another limitation concerns drug testing for opiates, as it was not possible to identify the opiate source. The use of prescription opiate medication, for instance, could result in a positive opiate drug test. In such a circumstance, a supervising officer should have verified the authenticity of the medical prescription. This process of prescription verification was handled on a case-by-case basis between the supervising officer and probationer and recorded in notes that were not linked to specific drug tests. It was not possible in the current study to parse out the number of positive opiate tests that resulted from prescription medication. In addition, it was not possible to calculate the number of individuals who were ordered to undergo a urinalysis drug screening but did not comply. It is possible that a probationer who was instructed to take a drug test never complied with those instructions and instead “walked away.” Identifying those who tested positive for opiate medication and those who walked away from drug tests could have changed the percentage of drug test failures in these data.

Limitations aside, the current research is important because it supports the theoretical and practical arguments that drug use is not monolithic and the effects of drug use on arrest may vary by substance type and frequency. Thus, it would be beneficial for community corrections to consider substance type when calculating criminogenic risk and needs. This would allow probation officers to identify a more tailored treatment policy to address the needs of multisubstance users. Future research should also investigate how multisubstance use might affect other probation outcomes, such as the likelihood of receiving a revocation. Future research could also investigate whether multisubstance use varies across important other subgroups, including specialized supervsion type.

Footnotes

Appendix

Zero-Order Correlations Between Variables in the Models.

S. no.	Variable	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16	17	18	19	20	21
1	Arrest while on probation	1
2	Multisubstance user	.09*	1
3	Single-substance user	.01	−.58*	1
4	One positive multisubstance test	.05*	.65*	−.38*	1
5	Two positive multisubstance tests	.06*	.61*	−.35*	−.20*	1
6	Three + positive multisubstance tests	.05*	.45*	−.26*	−.15*	.75*	1
7	Number of administered drug tests	.02	.05*	.07*	−.02	.08*	.10*	1
8	Number of positive drug tests	.05*	.33*	.06*	.08*	.35*	.36*	.59*	1
9	Drug instant offense	−.02	−.03	.02	−.01	−.02	−.02	−.13*	−.07*	1
10	Weapon/other instant offense	−.01	.02	−.03	.01	.01	−.01	.02	−.01	−.54*	1
11	Property instant offense	.07*	.03	−.02	.03	.01	.04	.07*	.04*	−.51*	−.19*	1
12	Number of open supervision cases	.06*	.06*	.01	.01	.07*	.06*	.00	.02	−.06*	.08*	.05*	1
13	Number of instant offense charges	.03	.06*	−.01	.01	.07*	.06*	.06*	.07*	−.05*	.05*	.02	.69*	1
14	Latino	.04	.03	−.03	.03	.01	.00	−.02	−.02	.09*	−.06*	−.03	.05*	.01	1
15	Non-White	.03	−.12*	.10*	−.02	−.13*	−.14*	.08*	.02	.05*	−.05*	−.05*	−.07*	−.05*	.15*	1
16	Felony	.12*	−.05*	.01	−.04	−.02	−.01	.11*	.04	.03	−.16*	.14*	−.07*	−.04	.04	.09*	1
17	Male	.10*	−.06*	.05*	−.01	−.07*	−.06*	.08*	.04*	.17*	−.23*	.00	−.14*	−.05*	.02	.09*	.14*	1
18	High risk	−.09*	.01	.02	.00	.00	−.01	.37*	.17*	−.22*	.02	.10*	−.10*	−.01	−.11*	.02	−.06*	.12*	1
19	Age	−.09*	.00	−.01	.01	−.01	.01	−.12*	−.08*	−.06*	.07*	.05*	.04*	.01	−.03	−.01	−.16*	−.15*	−.18*	1
20	Unemployed	.04	.00	.00	−.01	.02	.03	.02	.02	.00	.05*	−.03	−.01	.01	−.02	.02	−.03	.00	−.01	−.06*	1.00
21	Sentence length (months)	.18*	−.05*	.03	−.04	−.02	.00	.25*	.12*	−.12*	−.03	.14*	.11*	.14*	.04*	.06*	.40*	.08*	−.06*	−.04	.05*	1.00

p < 05.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Justin C. Medina

References

American Psychiatric Association. (1987). Diagnostic and statistical manual of mental disorders (3rd ed., Rev.). Washington, DC: Author.

Barnes

Hyatt

(2012). Using random forest risk prediction in the Philadelphia probation department. U.S. Department of Justice.

Bechara

(2005). Decision making, impulse control and loss of willpower to resist drugs: A neurocognitive perspective. Nature Neuroscience, 8(11), 1458–1463.

Bennett

Holloway

(2005). The association between multiple drug misuse and crime. International Journal of Offender Therapy and Comparative Criminology, 49(1), 63–81.

Bennett

Holloway

(2009). The causal connection between drug misuse and crime. The British Journal of Criminology, 49(4), 513–531.

Berk

Sherman

Barnes

Kurtz

Ahlman

(2009). Forecasting murder within a population of probationers and parolees: A high stakes application of statistical learning. Journal of the Royal Statistical Society: Series A (Statistics in Society), 172(1), 191–211.

Beswick

Best

Rees

Coomber

Gossop

Strang

(2001). Multiple drug use: Patterns and practices of heroin and crack use in a population of opiate addicts in treatment. Drug and Alcohol Review, 20(2), 201–204.

Bickel

W. K.

Marsch

L. A.

(2001). Toward a behavioral economic understanding of drug dependence: Delay discounting processes. Addiction, 96(1), 73–86.

Boardman

J. D.

Finch

B. K.

Ellison

C. G.

Williams

D. R.

Jackson

J. S.

(2001). Neighborhood disadvantage, stress, and drug use among adults. Journal of Health and Social Behavior, 42, 151–165.

10.

Britt

C. L.

Gottfredson

M. R.

Goldkamp

J. S.

(1992). Drug testing and pretrial misconduct: An experiment on the specific deterrent effects of drug monitoring defendants on pretrial release. Journal of Research in Crime and Delinquency, 29(1), 62–78.

11.

Cabrera

A. F.

(1994). Logistic regression analysis in higher education: An applied perspective. In Smart

J. C.

(Ed.), Higher education: Handbook of theory and research (Vol. 10, pp. 225–256). New York, NY: Agathon Press.

12.

Calcaterra

S. L.

Keniston

Blum

Crume

Binswanger

I. A.

(2015). The association between stimulant, opioid, and multiple drug use on behavioral health care utilization in a safety-net health system. Substance Abuse, 36(4), 407–412.

13.

Caudy

M. S.

Tang

Wooditch

Taxman

F. S.

(2014). Short-term trajectories of substance use in a sample of drug-involved probationers. Journal of Substance Abuse Treatment, 46(2), 202–213.

14.

Clear

T. R.

Rose

D. R.

Waring

Scully

(2003). Coercive mobility and crime: A preliminary examination of concentrated incarceration and social disorganization. Justice Quarterly, 20(1), 33–64.

15.

Conway

K. P.

Kane

R. J.

Ball

S. A.

Poling

J. C.

Rounsaville

B. J.

(2003). Personality, substance of choice, and polysubstance involvement among substance dependent patients. Drug and Alcohol Dependence, 71(1), 65–75.

16.

DeMatteo

Marlowe

D. B.

Festinger

D. S.

Arabia

P. L.

(2009). Outcome trajectories in drug court: Do all participants have drug problems? Criminal Justice and Behavior, 36(4), 354–368.

17.

Denman

Beal

Carrion

Dole

Medaris

Robinson

. . . Wilson-Fetrow

(2018). Substance use among New Mexico probationers and parolees. New Mexico Statistical Analysis Center.

18.

Deschenes

E. P.

Turner

Greenwood

P. W.

Chiesa

(1996). An experimental evaluation of drug testing and treatment intervention for probationers in Maricopa County, Arizona. RAND Corporation.

19.

Fairbairn

C. E.

Briley

D. A.

Kang

Fraley

R. C.

Hankin

B. L.

Ariss

(2018). A meta-analysis of longitudinal associations between substance use and interpersonal attachment security. Psychological Bulletin, 144(5), 532–555.

20.

Farabee

Prendergast

Anglin

M. D.

(1998). The effectiveness of coerced treatment for drug-abusing offenders. Federal Probation, 62, 3–10.

21.

Feucht

T. E.

Gfroerer

(2011). Mental and substance use disorders among adult men on probation or parole: Some success against a persistent challenge. Substance Abuse and Mental Health Services Administration.

22.

French

M. T.

McGeary

K. A.

Chitwood

D. D.

McCoy

C. B.

Inciardi

J. A.

McBride

(2000). Chronic drug use and crime. Substance Abuse, 21(2), 95–109.

23.

Friedmann

P. D.

Taxman

F. S.

Henderson

C. E.

(2007). Evidence-based treatment practices for drug-involved adults in the criminal justice system. Journal of Substance Abuse Treatment, 32(3), 267–277.

24.

Galizio

Stein

F. S.

(1983). Sensation seeking and drug choice. International Journal of the Addictions, 18(8), 1039–1048.

25.

Gendreau

Little

Goggin

(1996). A meta-analysis of the predictors of adult offender recidivism: What works! Criminology, 34(4), 575–608.

26.

Gottfredson

D. C.

Najaka

S. S.

Kearley

(2003). Effectiveness of drug treatment courts: Evidence from a randomized trial. Criminology & Public Policy, 2(2), 171–196.

27.

Gray

M. K.

Fields

Maxwell

S. R.

(2001). Examining probation violations: Who, what, and when. Crime & Delinquency, 47(4), 537–557.

28.

Hakansson

Schlyter

Berglund

(2011). Associations between polysubstance use and psychiatric problems in a criminal justice population in Sweden. Drug and Alcohol Dependence, 118(1), 5–11.

29.

Hall

(2015). What has research over the past two decades revealed about the adverse health effects of recreational cannabis use? Addiction, 110(1), 19–35.

30.

Harcourt

B. E

. (2008). Against prediction: Profiling, policing, and punishing in an actuarial age. The University of Chicago Press.

31.

Harrell

Kleiman

(2002). Drug testing in criminal justice settings. In Leukefeld

Tims

(Eds.), Clinical and policy responses to drug offenders: Policies and issues (p. 149–171). Springer.

32.

Harrell

Roman

(2001). Reducing drug use and crime among offenders: The impact of graduated sanctions. Journal of Drug Issues, 31(1), 207–232.

33.

Henkel

(2011). Unemployment and substance use: A review of the literature (1990-2010). Current Drug Abuse Reviews, 4(1), 4–27.

34.

Huebner

B. M.

Berg

M. T.

(2011). Examining the sources of variation in risk for recidivism. Justice Quarterly, 28(1), 146–173.

35.

Huebner

B. M.

Cobbina

(2007). The effect of drug use, drug treatment participation, and treatment completion on probationer recidivism. Journal of Drug Issues, 37(3), 619–641.

36.

Kaeble

(2018). Probation and parole in the United States, 2016. Bureau of Justice Statistics, Office of Justice Programs, U.S. Department of Justice.

37.

Kendall

J. C.

Sherman

M. F.

Bigelow

G. E.

(1995). Psychiatric symptoms in polysubstance abusers: Relationship to race, sex, and age. Addictive Behaviors, 20(5), 685–690.

38.

Kleiman

M. A.

Tran

T. H.

Fishbein

Magula

M. T.

Allen

Lacy

(2003). Opportunities and barriers in probation reform: A case study of drug testing and sanctions. California Policy Research Center.

39.

Lacey

J. H.

Evans

C. D. H.

(1986). The impulsivist: A multi-impulsive personality disorder. British Journal of Addiction, 81(5), 641–649.

40.

Langan

P. A.

Levin

D. J.

(2002). Recidivism of prisoners released in 1994. Federal Sentencing Reporter, 15(1), 58–65.

41.

Latessa

Travis

Fulton

Stichman

(1998). Evaluating the prototypical ISP. Report Submitted to the National Institute of Justice (95-IJ-CX-0032).

42.

Lex

B. W.

(1991). Some gender differences in alcohol and polysubstance users. Health Psychology, 10(2), 121–132.

43.

Lin

Grattet

Petersilia

(2010). “Back-end sentencing” and reimprisonment: Individual, organizational, and community predictors of parole sanctioning decisions. Criminology, 48(3), 759–795.

44.

Lurigio

Cho

Swartz

Johnson

Graf

Pickup

(2003). Standardized assessment of substance-related, other psychiatric, and comorbid disorders among probationers. International Journal of Offender Therapy and Comparative Criminology, 47, 630–652.

45.

Marlowe

D. B.

(2011). Evidence-based policies and practices for drug-involved offenders. The Prison Journal, 91(Suppl. 3), 27S–47S.

46.

Martinotti

Carli

Tedeschi

Di Giannantonio

Roy

Janiri

Sarchiapone

(2009). Mono- and polysubstance dependent subjects differ on social factors, childhood trauma, personality, suicidal behaviour, and comorbid Axis I diagnoses. Addictive Behaviors, 34(9), 790–793.

47.

Meares

T. L.

(1997). Social organization and drug law enforcement. American Criminal Law Review, 35, 191–277.

48.

Menard

(2002). Applied logistic regression analysis (Vol. 106). SAGE.

49.

Moore

L. D.

Elkavich

(2008). Who’s using and who’s doing time: Incarceration, the war on drugs, and public health. American Journal of Public Health, 98(Suppl. 1), S176–S180.

50.

Morgan

K. D.

(1994). Factors associated with probation outcome. Journal of Criminal Justice, 22(4), 341–353.

51.

Morris

Tonry

(1991). Between prison and probation: Intermediate punishments in a rational sentencing system. Oxford University Press.

52.

Olson

D. E.

Lurigio

A. J.

(2000). Predicting probation outcomes: Factors associated with probation rearrest, revocations, and technical violations during supervision. Justice Research and Policy, 2(1), 73–86.

53.

Olson

D. E.

Lurigio

A. J.

Seng

(2000). A comparison of female and male probationers: Characteristics and case outcomes. Women & Criminal Justice, 11(4), 65–79.

54.

Parrott

A. C.

Sisk

Turner

J. J. D.

(2000). Psychobiological problems in heavy “ecstasy” (MDMA) polydrug users. Drug and Alcohol Dependence, 60(1), 105–110.

55.

Preston

K. L.

Kowalczyk

W. J.

Phillips

K. A.

Jobes

M. L.

Vahabzadeh

Lin

J. L.

. . . Epstein

D. H.

(2017). Context and craving during stressful events in the daily lives of drug-dependent patients. Psychopharmacology, 234(17), 2631–2642.

56.

Quirk

S. W.

McCormick

R. A.

(1998). Personality subtypes, coping styles, symptom correlates, and substances of choice among a cohort of substance abusers. Assessment, 5(2), 157–169.

57.

Rhodes

Hyatt

Scheiman

(1996). Predicting pretrial misconduct with drug tests of arrestees: Evidence from eight settings. Journal of Quantitative Criminology, 12(3), 315–348.

58.

Riley

E. D.

Evans

J. L.

Hahn

J. A.

Briceno

Davidson

P. J.

Lum

P. J.

Page

(2016). A longitudinal study of multiple drug use and overdose among young people who inject drugs. American Journal of Public Health, 106(5), 915–917.

59.

Salom

C. L.

Betts

K. S.

Williams

G. M.

Najman

J. M.

Alati

(2016). Predictors of comorbid polysubstance use and mental health disorders in young adults—A latent class analysis. Addiction, 111(1), 156–164.

60.

Sinha

(2008). Chronic stress, drug use, and vulnerability to addiction. Annals of the New York Academy of Sciences, 1141(1), 105–130.

61.

Smith

K. E.

Stoops

W. W.

(2019). Synthetic cathinone use among polysubstance users: Indirect indicator of indiscriminate drug taking or preferred drug of abuse? Journal of Drug Issues, 49(2), 369–386.

62.

Werb

Kamarulzaman

Meacham

M. C.

Rafful

Fischer

Strathdee

S. A.

Wood

(2016). The effectiveness of compulsory drug treatment: A systematic review. International Journal of Drug Policy, 28, 1–9.

63.

Whetzel

Paparozzi

Alexander

Lowenkamp

C. T.

(2011). Goodbye to a worn-out dichotomy: Law enforcement, social work, and a balanced approach (a survey of federal probation officer attitudes). Federal Probation, 75(2), 7–12.

64.

Winkelman

T. N.

Chang

V. W.

Binswanger

I. A.

(2018). Health, polysubstance use, and criminal justice involvement among adults with varying levels of opioid use. JAMA Network Open, 1(3), e180558.

65.

Young

Fluellen

Belenko

(2004). Criminal recidivism in three models of mandatory drug treatment. Journal of Substance Abuse Treatment, 27(4), 313–323.