Substance misuse is strongly linked to criminal behaviour (1). This association is described as a ‘dynamic relationship’, as the literature recognises that the links between drug-use and criminal behaviour are multi-directional (2, 3). A history of substance misuse has been identified in the literature as a strong predictor of violence and recidivism in both the general offender population (4) and for offenders with mental health disorders (5, 6). In a systematic review of 18,388 newly admitted inmates across 24 studies, the authors found that approximately 30% of men and 51% of women had a drug misuse disorder, and a quarter of both sexes had an alcohol misuse disorder (7).

Given the prevalence of substance use disorders in the criminal justice settings (CJS), various efforts aim to treat or divert people with substance use disorders away from traditional criminal justice disposals (8). Alternative approaches include the use of drug courts, diversion schemes, mandatory or voluntary treatment programmes, pharmacological and psychosocial interventions, provided across a range of settings including secure hospitals, prisons, and various forms of support or monitoring in the community (1). A comprehensive meta-analysis of 28 studies by Holloway et al. (9) concluded that the likelihood of reoffending by individuals assigned to some form of treatment was 41% lower than controls not receiving such an intervention.

Evidence as to the effectiveness of different types of interventions is mixed (1). For instance, although some studies suggest that for non-substance abusing offenders, community sentences lead to lower reoffending rates than penal sentences (3), interventions targeting offenders with substance use disorders in the community are less effective than intensive programmes in secure environments (9). Research indicates that programmes combining pharmacological and psychosocial elements are more effective than either delivered individually, and that therapeutic communities, which embody a secure holistic treatment approach, receive the most empirical support (1). A review of 43 trials using non-pharmacological interventions for criminal justice-involved persons found that on average these led to a significant reduction in re-incarceration and a near significant reduction in substance use, but not in re-arrests; therapeutic communities were the only modality that led to significant reductions in re-arrests (10). Further, interventions are differentially effective; non-white, young men are more likely to benefit than other groups (9).

Mandatory treatment is defined as ‘any form of drug treatment that is ordered, motivated, or supervised by the criminal justice system’ [(11), 2]. In Germany, Austria and Switzerland, treatment can be ordered without requiring an individual’s consent.

The evidence is mixed on the question of whether legal compulsion as compared to voluntary programmes is associated with better outcomes (12). A review of the effectiveness of mandatory treatment interventions for people with substance use disorders found nine relevant studies (13). Of these, three studies reported no difference between mandatory treatment and controls on substance use outcomes, two studies found ‘equivocal’ links to positive substance use outcomes but there were no control groups, two studies reported positive outcomes, and two studies observed negative impacts on criminal recidivism (i.e., higher reoffending rates).

One of the reasons for heterogeneity in outcomes is the wide variation in programme intensity, modality, setting, and patient population. A review of coerced treatment programmes for substance using offenders identified programme- and individual-level factors that were associated with more successful outcomes (14). At the programme-level, these included, among other things, programmes that: lasted longer than 90 days, had compliance measures to enforce requirements, incorporated motivation and reinforcement techniques, ensured a continuum of care across the criminal justice system, and employed staff who had accreditation and training. At the individual-level, the authors report that greater motivation, treatment readiness, higher levels of religion and faith, and lower baseline stress levels were associated with better substance use and offending outcomes. Co-occurring mental disorders were linked with higher recidivism rates, and criminal thinking styles with poor engagement in treatment.

The German criminal justice system provides for a form of mandatory treatment for offenders whose convictions are considered by a court to relate to substance use. §64 of the German Penal Code (Strafsgesetzbuch; StGB) stipulates that where an offence can be linked to a substance use disorder or is committed in a state of intoxication, there is risk of future substance use-related offending, and there are reasonable prospects that treatment might be successful, an individual can be ordered into a forensic mental health hospital specialising in the treatment of substance use disorders. These hospitals are secure forensic-psychiatric settings, the vast majority of which have secure perimeters. Some such hospitals specialise in treating individuals detained under §64 StGB while others also treat offender patients with other disorders having been committed under a hospital order with unlimited duration by the courts. There is no distinction of hospitals by security levels in Germany. The vast majority of forensic-psychiatric hospitals offer care at all levels of security but differentiate security levels internally within the same institution. Treatment models vary across these sites, though typically multidisciplinary, holistic treatment programmes involving a range of psychological, social, occupational, pharmacological, and substitution treatments are offered (15).

The use of §64 StGB is independent of criminal responsibility; in practice about 60% of the individuals detained have full, and most of the remainder diminished, criminal responsibility. Both these groups receive a parallel prison sentence and could be returned to prison to serve this sentence if treatment does not appear to be successful. Detention under §64 StGB is ordered for a period of 2 years, but can be extended by up to 2/3 of the length of the prison sentence ordered in parallel (16). In case of long prison sentences, individuals will be ordered to serve part of the prison sentence prior to mandatory treatment in order to avoid them having to return to prison after successful completion of the treatment (15). Placement is reviewed every 6 months. As is common internationally, most patients are men (87.9%); the average age is 37 years (17, 18). Average treatment length is 3.5 years (15). There were 4,300 §64 German Penal Code patients in Germany in 2019, a marked increase from 1,657 in 1999 (18). The reasons for this increase are not fully understood. However, a number of factors have been identified accounting for this observation: 1. There has been an increase in the average lengths of parallel prison sentences (by 9 months over a period of 20 years) due to an increase in severe (violent) offences. 2. As a consequence, average duration of treatment has also increased (by 6 months). 3. The characteristics of patients have changed, in particular, the percentage of patients admitted with full criminal responsibility has increased due to a wider interpretation of the entry criteria by the courts. This has led some to argue that there are fewer ill and a greater number of dissocial patients with limited treatment motivation and poor prognosis now residing in forensic psychiatric care. Some have suggested that the possibility of early release provides a false incentive for this group (for a summary of the discussions see Bundesministerium für Justiz, 2022).

Approximately half of all patients do not successfully complete treatment (15). This has raised questions about whether the right individuals are being selected for mandatory treatment in the first instance and led to calls, including from the German Association for Psychiatry, Psychotherapy and Psychosomatics (DGPPN), to significantly reform the admission criteria for treatment under this legal provision, including making such treatment voluntary (19).

Knowledge of the outcomes for the §64 German Penal Code treatment population and the factors predictive or associated with these outcomes in the international (English language) literature is sparse. A recent study found that patients successfully completing treatment took significantly longer to reoffend (N = 110, M = 46 months) than those transferred to prison (N = 151, M = 25 months) (20). A review by Fries et al. (21) of nine papers published between 1999 and 2009 in German reported that young age, history of offending, lack of educational and vocational attainment, and comorbid personality disorders were the most robust predictors of treatment completion. A more recent review of 16 studies published after 1999 in German looking at predictors of treatment completion found that age, past criminal convictions, comorbid personality disorders, use of substances other than alcohol, educational and vocational attainment, withdrawal from previous substitution programmes, and needs mostly identified as substance-related instead of criminogenic, are the strongest predictors (22). A second review conducted at the same time by the same authors address predictors of reoffending after discharge (23). The authors report mixed findings but suggest that patient motivation, living circumstances and resources, and offending history were the most reliable predictors.

The present review aimed to synthesise the literature describing studies of the mandatory substance use treatment patient population in Germany (§64 StGB) in relation to three primary outcomes: reconviction/reoffending, treatment completion and the recurrence of substance use. It summarises the reported prevalence rates for these outcomes and describes the evidence-base for factors/predictor variables associated with these outcomes. This review extends the findings of the previous reviews by additionally reporting on reconviction/reoffending during treatment and the recurrence of substance use during treatment and after discharge. It also is the first English-language review of this literature. The German system is worthy of study as the largest European jurisdiction to have mandatory substance use treatment, is currently reviewing laws relating to this, and little has been published on this system in the international literature.

Key study data were extracted and recorded in a spreadsheet. These data include: authors, location of study, journal, year, type of study design, participants, description of secure setting (e.g., level of security), sample size, type of intervention if any, outcome variables, predictor variables, method of data analysis and the study findings. Data were extracted by JT, EM, PW and JW.

Our initial search returned 17,769 results (see PRISMA flow diagram in Figure 1); 5,572 results were removed after de-duplication. The titles and abstracts of 12,197 results were screened, with 11,971 removed; 226 results were sought for retrieval; 16 of these could not be found due to lack of access (authors were contacted where contact details were available). The results of 210 publications were assessed for eligibility and 192 of these were excluded in line with the review inclusion and exclusion criteria, leaving 18 hits. The reference lists of these 18 remaining papers were screened and 245 possible additional results were identified, 71 of these could not be accessed due to lack of access. Of these additional results, 174 were assessed for eligibility, of which 152 were excluded according to the review inclusion and exclusion criteria, leaving 23 hits. This left a total of 40 publications reporting on 32 studies. An updated search returned 23 possible hits for full text screening after deduplication. Five of these additional papers were added following eligibility screening.

Publications included in the review are indicated with an asterisk (*) in the reference list.

Most publications reported on studies with cohort designs, of which 13 were retrospective and 21 prospective; there were six publications reporting cross-sectional studies; two retrospective case-controlled studies; and one case study. Sample sizes ranged from N = 2 (24) to N = 2,483 (25). Date of publication ranged from 1988 to 2023. There was a clear increase in the quality of reporting in the later studies rendering data extraction in more recent publications much easier. Five publications were in the English language, the rest in German.

Meta-analysis would not have produced meaningful insights given the number of variables measured in the literature and heterogeneous operationalisations [for example, number of past criminal offences was variously measured as: low, medium, and high severity in Gericke and Kallert (26); total number of officially recorded convictions in, for example, Querengässer et al. (27); and number of self-reported offences in Butz (28)]. Therefore, we have chosen to present the findings as a narrative synthesis in line with the general framework proposed by the Centre for Reviews and Dissemination (29). We do not articulate a theory of change as the treatment programmes, settings, and organisational cultures of the different forensic psychiatric hospitals that deliver mandatory treatment for this population over the studied time period vary widely. We develop a preliminary synthesis in the tables and text presented in our results section. We explore the relationships between studies by commenting on whether studies that investigate relationships between the same or similar variables draw convergent or divergent conclusions. We finally assess the robustness of our synthesis in the discussion section.

To strengthen the findings of the review, the sections below are a synthesis of findings relating to predictors that were investigated in more than one study. Significant and non-significant findings are reported. We present the findings relating to each primary outcome thematically (with the exception of the recurrence of substance use, for which there were fewer studies, and we present these findings in one section): legal and criminological; treatment and mental health; and social and demographic categories. Median values were the preferred measure of central tendency given the heterogeneity in outcome variables such as sample sizes and follow-up periods. Only publications reporting on studies using a cohort or cross-sectional design are included in the summary tables to increase the generalisability of findings. Case studies or studies of interventions are not included in these tables or used to contribute to median primary outcome values (e.g., reconviction rates); the findings from these studies are included in the textual description of results.

On average, reconviction rates (Mdn = 41% over Mdn = 30 months) were higher than those reported for forensic patients in Germany with major mental disorders [receiving treatment under §63 German Penal Code; e.g., 35.2% across a mean follow-up of 16.5 years reported in Seifert et al. (66)]. Franke et al. (40) directly compared the reconviction/reoffending rates of these two groups in a single federal state and confirmed higher recidivism rates in patients with substance use disorders (SUD). The rates found in our review are also higher than reconviction rates for non-forensic patients leaving prison settings in Germany: approximately 34% of people leaving prisons reoffend within 3 years (67). While indicative of the outcomes for the forensic SUD population, these groups are not entirely comparable as individuals have not been matched on factors like age, index offence type, and gender as part of a single prospective study.

Our finding that predictors of reoffending differed in some ways for successful and unsuccessful completers (e.g., number of past convictions, completed job training, age, marital status) might reflect wider national debates about the appropriateness of the §64 German Penal Code treatment order for a large number of people (19). A higher number of past convictions, a lack of job training completion prior to their current forensic treatment, a lower age at admission, and being single were significantly positively correlated with reconviction/offending for patients transferred to prison; while no significant associations were reported between these variables and reconviction/offending for patients transferred to the community. The unsuccessful completers might more closely resemble general (i.e., non-forensic) population offenders, for whom most models predicting reoffending have been developed and thus explain why some predictors were more predictive of reconviction in this group.

Across 22 publications in our review, patients were on average more likely to have a negative treatment outcome than a positive treatment outcome (47 vs. 45% respectively; though note the limitations section when considering this median value). This pattern is repeated in several of the largest studies. The importance of this is compounded by studies demonstrating that unsuccessful completers reoffend at higher rates than patients completing treatment successfully (40). One possible explanatory factor here might be systemic, that courts are sentencing people to mandatory SUD treatment based on incorrect assumptions about the helpfulness of treatment or broad interpretations of the §64 StGB treatment order by courts or psychiatric experts (40). A second factor might point to the absence of readiness/motivation as an important predictor of engagement and success. Motivation has been identified as a key antecedent to positive outcomes in substance use programme participation and recidivism in the international literature (12, 14, 68); in our review (see section 3.2.3); and is viewed by staff as an important factor in treatment outcome in this population (69, 70). A third factor might be methodological: treatment effects could be moderated by evaluation methods, program features, and treatment contexts.

A recent study highlights the possible role of demographics in the effectiveness of treatment. Berthold and Riedemann (17) used data collected in a large national survey of 2046 §64 StGB patients. A multivariate regression (N = 326) indicated that three variables were significant predictors of treatment completion: being in work in the 6 months prior to the index offence and intoxication at time of the index offence were associated with greater odds of treatment completion; having a migration background was associated with greater odds of treatment failure.

Explaining this latter finding, the authors propose that treatment engagement might by affected by lack of knowledge of the German language and provision of support in a range of languages, not migration background per se. Ross et al. (71) note that forensic patients with a migration background in Germany face a range of factors that complicate their health and social needs, including socioeconomic status, racism, access to education, fleeing from war or persecution, factors linked to precarious travel from home countries, and the cultural and linguistic adaptations to life in Germany. These factors are linked to challenges in assessment and treatment engagement (72). It should be noted that Kemper (52) found no significant link between either nationality or language difficulties and treatment outcome, however. The regression in Berthold & Riedemann (17) accounted for 24% of outcome variance, thus leaving considerable room for further explanatory factors. The only variables consistently associated with successful treatment completion across multiple studies in our review was the occurrence of escapes/absconding events and number of past convictions (successful treatment completers had significantly fewer escapes/absconding events and past convictions).

Studies of the recurrence of substance use included in our review found that approximately half of all patients reused substances during treatment, and around 60% reused after discharge, with follow-up periods ranging from 12 to 48 months. Reviews have highlighted high substance use recurrence rates across a range of substances in the general population during and after treatment (73). These recurrences can occur soon after initiating treatment, with several studies reporting recurrence of substance use within weeks to months in two-thirds of some study participants (73). A comparative study of outcomes for (a) criminal justice-mandated patients (n = 141), (b) justice-involved, non-mandated patients (n = 235), and (c) patients not involved in the CJS following substance use treatment in the US (n = 1,719) found that 53.9, 45.3 and 39.9% of people in these groups, respectively, remained abstinent after 1 year (74). Higher abstinence rates for the criminal justice-mandated patients were not explained by before-treatment differences.

Overall, the findings of our review suggest that although a large number of studies have been conducted to identify the predictors of successful treatment and criminological outcomes for this population, there remain three problem areas. The first is methodological; although many studies in our review tested the same relationships between x and y variables (e.g., vocational training and reconviction), the operationalisations of these varied widely, particularly at the predictor level. This renders meta-analysis unsuitable. The second is practical; most variables are static and historical as these are easier to routinely collect. This neglects important dynamic treatment and risk factors that are known to be important for predicting outcomes (75–77). The third is conceptual; study authors [with a few exceptions, e.g., speaking patterns in a therapeutic programme as an indicator of engagement in Querengässer (46)] did not choose variables/factors that fall firmly within recent developments in offender rehabilitation, namely strength- and recovery-based approaches, or risk assessment that incorporates protective factors or shared risk assessment. None of the included studies explicitly measured domains of personal recovery, protective factors, fundamental goods as identified in the Good Lives Model (78), or patient participation in treatment, for example in shared risk assessment (79). A growing body of literature supports the claim that strength- and recovery-based models of treatment are valuable concomitants to medical, risk, or deficit-based approaches (80).

Researchers should therefore agree upon variable operationalisations and methods of data imputation and collection. This can be achieved by the widespread adoption, expansion and harmonisation of pre-existing surveys and databases. Berthold and Riedemann (17, 81) annually collect data from across Germany: the ‘§64 StGB Stichtagserhebung’ (‘cut-off date survey’). A nationwide effort should be undertaken to adopt this in all forensic hospitals with forensic patients receiving §64 StGB care for SUDs. In furtherance to this nation-wide survey/database, a working group should be established to choose routinely used measures of strength- and recovery-based treatment. Given the relevance of therapeutic milieu and the efficacy of therapeutic community (TC) approaches, measures of ward atmosphere would also be beneficial (82). Further, researchers should ensure they consider gender differences in outcomes as some recent research of this population found that differences in outcomes across sex can be observed: a similar proportion of men (55%) and women (55%) completed treatment, whereas men (42%) were more likely to be released to prison than women (29%); rehospitalisation was significantly linked to homelessness and unemployment in women but not men (40).

Several limitations of this review should be acknowledged. First, literature for this review was sought by using search terms in the English language. This returned a reasonable number of relevant hits. However, the number of eligible publications found following hand searches of the reference lists of the initially returned hits suggests that (A) the initial search should have been conducted in the German language too, but (B) the low number of relevant results was compensated for as 245 potentially relevant publications were subject to full-text review and 26 publications were included. Second, it was not always possible to discern how authors operationalised certain variables, such as treatment ‘success’ or ‘failure’. We have attempted to be as consistent and clear as possible in reporting results and to inform the reader of these differences across studies; we also chose not to undertake a meta-analysis for this reason. Third, it was not always clear across different studies whether the same or overlapping samples were reported upon. Where this was clearer (e.g., Essener Evaluation Study) we indicated this. However, it is possible some patient data have been reported on more than once. Fourth, median values presented in this review are taken from a range of studies with various sampling methods (e.g., cohort, cross-sectional). Therefore, median values are indicative only and cannot be said to be definitive and generalisable for all patients at all hospitals in Germany. Finally, reoffending and reconviction offence categories and types of substances were not widely reported in studies. Predictors of reoffending, reconviction, and substance use will likely be different for different offence categories and substance types, so this should also be considered in future research.