Predicting criminal offence in adolescents who exhibit antisocial behaviour: a machine learning study using data from a large randomised controlled trial of multisystemic therapy.

INTRODUCTION: Accurate prediction of short-term offending in young people exhibiting antisocial behaviour could support targeted interventions. Here we develop a set of machine learning (ML) models that predict offending status with good accuracy; furthermore, we show interpretable ML analyses can complement models to inform clinical decision-making. METHODS: This study included 679 individuals aged 11-17 years who displayed moderate-to-severe antisocial behaviour, from a controlled trial of Multisystemic-therapy in England. The outcome was any criminal offence in the 18 months after study baseline. Four types of ML algorithms were trained: logistic regression, elastic net regression, random forest, and gradient boosting machine (GBM). Prediction models were developed (1) using predictors readily available to clinicians (e.g. sociodemographics, previous convictions), and (2) with additional information (e.g. parenting). Model agnostic feature importance values were calculated and the most important predictors identified. Nested cross-validation with 100 iterations of random data splits and 10-fold cross-validation within each iteration was employed, and the average predictive performance was reported. RESULTS: Among the ML models using readily available predictors, the GBM is the strongest model (AUC 0.85, 95% CI 0.85-0.86); the other models have average AUCs of 0.82. This performance was better than using only the total number of previous offences as the predictor (0.67, 0.66-0.68), and the model simply assuming past offending status as the prediction (0.81, 0.80-0.81). Additional predictors slightly increased the performance of logistic regression and random forest models but decreased the performance of elastic net regression and gradient boosting machine-based models. CONCLUSION: The potential utility of ML approaches for accurately predicting criminal offences in high-risk youth is demonstrated. Interpretable ML-based predictive models could be utilised in youth services or research to help develop and deliver effective interventions.

Antisocial cognition as a mediator of the peer influence effect and peer selection effect in antisocial adolescents.

The peer influence and peer selection effects are two widely replicated findings in the criminological literature that refer to the predictive relationship between antisocial behaviour and delinquent peer association as well as between delinquent peer association and antisocial behaviour, respectively. Research suggests that antisocial cognition might constitute a causal mechanism underlying part of these effects. This study investigated the extent that the peer influence and peer selection effects are mediated by one key aspect of antisocial cognition-beliefs and attitudes supporting peer conflict. This study examined whether beliefs and attitudes supporting peer conflict mediated the relationship between delinquent peer association and volume of self-reported antisocial behaviour and vice-versa, across a 1-year follow-up period, in 683 (433 male, 250 female) British adolescents (mean age: 13.8 years) with a history of serious antisocial behaviour. Participants completed measures at baseline and 6, 12 and 18 months thereafter. Findings indicated that beliefs and attitudes supporting peer conflict partially mediated the peer influence and peer selection effects, explaining a substantial proportion of the total effect in the peer influence (i.e., 26%) and peer selection (i.e., 17%) models. These results suggest that beliefs and attitudes supporting peer conflict could explain part of the mechanism underlying the peer influence and peer selection effects in adolescents with a history of serious antisocial behaviour.

Using an objective computer task (QbTest) to aid the identification of attention deficit hyperactivity disorder (ADHD) in the Children and Young People Secure Estate (CYPSE): a feasibility randomised controlled trial.

OBJECTIVES: QbTest has been shown to improve time to decision/diagnosis for young people with attention deficit hyperactivity disorder (ADHD). The aim was to assess the feasibility of QbTest for young people in prison. DESIGN: Single-centre feasibility randomised controlled trial (RCT), with 1:1 allocation. Concealed random allocation using an online pseudorandom list with random permuted blocks of varying sizes. SETTING: One Young Offenders Institution in England. PARTICIPANTS: 355 young people aged 15-18 years displaying possible symptoms of ADHD were assessed for eligibility, 69 were eligible to take part and 60 were randomised. INTERVENTION: QbTest-a computer task measuring attention, activity and impulsivity. MAIN OUTCOME MEASURES: Eligibility, recruitment and retention rates and acceptability of randomisation and trial participation. RESULTS: Of the 355 young people assessed for eligibility, 69 were eligible and 60 were randomised (n=30 QbTest plus usual care; n=30 usual care alone). The study achieved the specified recruitment target. Trial participation and randomisation were deemed acceptable by the majority of participants. 78% of young people were followed up at 3 months, but only 32% at 6 months, although this was also affected by COVID-19 restrictions. Secondary outcomes were mixed. Participants including clinical staff were mostly supportive of the study and QbTest; however, some young people found QbTest hard and there were issues with implementation of the ADHD care pathway. There were no serious adverse events secondary to the study or intervention and no one was withdrawn from the study due to an adverse event. CONCLUSIONS: With adaptations, a fully powered RCT may be achievable to evaluate the effectiveness of QbTest in the assessment of ADHD in the Children and Young People Secure Estate, with time to decision (days) as the primary outcome measure. However, further programme developmental work is required to address some of the challenges highlighted prior to a larger trial. TRIAL REGISTRATION NUMBER: ISRCTN17402196.

FACT: a randomised controlled trial to assess the feasibility of QbTest in the assessment process of attention deficit hyperactivity disorder (ADHD) for young people in prison-a feasibility trial protocol.

INTRODUCTION: The prevalence of attention deficit hyperactivity disorder (ADHD) within the Children and Young People Secure Estate (CYPSE) is much higher than seen in the general population. To make a diagnosis of ADHD, clinicians draw on information from multiple sources, including parents and teachers. However, obtaining these is particularly difficult for young people in the secure estate. There is increasing evidence in the community that QbTest is able to assist in the accurate and earlier diagnosis of ADHD. The objective of this study is to assess the feasibility and acceptability of QbTest in the assessment of ADHD within the CYPSE. METHODS AND ANALYSIS: A single-centre parallel group feasibility randomised controlled trial will be conducted. Sixty young people within the CYPSE identified as displaying possible symptoms of ADHD will be randomised to the intervention arm (n=30; QbTest plus usual care) or control arm (n=30; usual care). Primary analyses will be descriptive and a process evaluation will be conducted to assess the contexts involved in implementing the intervention. Interviews will be conducted to explore acceptability and thematic analysis will be used to analyse the data. ETHICS AND DISSEMINATION: This study was approved by National Health Service Wales research ethics committee 3 (18/WA/0347) on 15 February 2019. The findings will be published in peer-reviewed journals, presented at relevant conferences and disseminated to the public via summaries cocreated with our patient and public involvement group. TRIAL REGISTRATION NUMBER: ISRCTN17402196.

Multisystemic therapy versus management as usual in the treatment of adolescent antisocial behaviour (START): 5-year follow-up of a pragmatic, randomised, controlled, superiority trial.

BACKGROUND: Multisystemic therapy is a manualised treatment programme for young people aged 11-17 years who exhibit antisocial behaviour. To our knowledge, the Systemic Therapy for At Risk Teens (START) trial is the first large-scale randomised controlled trial of multisystemic therapy in the UK. Previous findings reported to 18 months after baseline (START-I study) did not indicate superiority of multisystemic therapy compared with management as usual. Here, we report outcomes of the trial to 60 months (START-II study). METHODS: In this pragmatic, randomised, controlled, superiority trial, young people (aged 11-17 years) with moderate-to-severe antisocial behaviour were recruited from social services, youth offending teams, schools, child and adolescent mental health services, and voluntary services across England, UK. Participants were eligible if they had at least three severity criteria indicating past difficulties across several settings and one of five general inclusion criteria for antisocial behaviour. Eligible families were randomly assigned (1:1), using stochastic minimisation and stratifying for treatment centre, sex, age at enrolment, and age at onset of antisocial behaviour, to management as usual or 3-5 months of multisystemic therapy followed by management as usual. Research assistants and investigators were masked to treatment allocation; the participants could not be masked. For this extension study, the primary outcome was the proportion of participants with offences with convictions in each group at 60 months after randomisation. This study is registered with ISRCTN, ISRCTN77132214, and is closed to accrual. FINDINGS: Between Feb 4, 2010, and Sept 1, 2012, 1076 young people and families were assessed for eligibility and 684 were randomly assigned to management as usual (n=342) or multisystemic therapy (n=342). By 60 months' of follow-up, 188 (55%) of 342 people in the multisystemic therapy group had at least one offence with a criminal conviction, compared with 180 (53%) of 341 in the management-as-usual group (odds ratio 1·13, 95% CI 0·82-1·56; p=0·44). INTERPRETATION: The results of the 5-year follow-up show no evidence of longer-term superiority for multisystemic therapy compared with management as usual. FUNDING: National Institute for Health Research Health Services and Delivery Research programme.

Changes in General and Specific Psychopathology Factors Over a Psychosocial Intervention.

OBJECTIVE: Recent research suggests that comorbidity in child and adolescent psychiatric symptoms can be summarized by a single latent dimension known as the p factor and more specific factors summarizing clusters of symptoms. This study investigated within- and between-person changes in general and specific psychopathology factors over a psychosocial intervention. METHOD: A secondary analysis was conducted of the Systemic Therapy for At-Risk Teens study, a pragmatic randomized controlled trial that compared the effects of multisystemic therapy with those of management as usual for decreasing antisocial behavior in 684 adolescents (82% boys; 11-18 years old at baseline) over an 18-month period. The general p factor and specific antisocial, attention, anxiety, and mood factors were estimated from a symptom-level analysis of a set of narrowband symptom scales measured repeatedly during the study. General and specific psychopathology factors were assessed for reliability, validity, and within- and between-person change using a parallel process multilevel growth model. RESULTS: A revised bi-factor model that included a general p factor and specific anxiety, mood, antisocial, and attention factors with cross-loadings fit the data best. Although the factor structure was multidimensional, the p factor accounted for most of the variance in total scores. The p factor, anxiety, and antisocial factors predicted within-person variation in external outcomes. Furthermore, the p factor and antisocial factors showed within-person declines, whereas anxiety showed within-person increases, over time. Despite individual variation in baseline factor scores, adolescents showed similar rates of change. CONCLUSION: The bi-factor model is useful for teasing apart general and specific therapeutic changes that are conflated in standard analyses of symptom scores. CLINICAL TRIAL REGISTRATION INFORMATION: START (Systemic Therapy for At Risk Teens): A National Randomised Controlled Trial to Evaluate Multisystemic Therapy in the UK Context; http://www.isrctn.com; ISRCTN77132214.

Multisystemic therapy versus management as usual in the treatment of adolescent antisocial behaviour (START): a pragmatic, randomised controlled, superiority trial.

BACKGROUND: Adolescent antisocial behaviour is a major health and social problem. Studies in the USA have shown that multisystemic therapy reduces such behaviour and the number of criminal offences committed by this group. However, findings outside the USA are equivocal. We aimed to assess the effectiveness and cost-effectiveness of multisystemic therapy versus management as usual in the treatment of adolescent antisocial behaviour. METHODS: We did an 18 month, multisite, pragmatic, randomised controlled, superiority trial in England. Eligible participants aged 11-17 years with moderate-to-severe antisocial behaviour had at least three severity criteria indicating past difficulties across several settings and one of five general inclusion criteria for antisocial behaviour. We randomly assigned families (1:1) using stochastic minimisation, stratifying for treatment centre, sex, age at enrolment to study, and age at onset of antisocial behaviour, to receive either management as usual or 3-5 months of multisystemic therapy followed by management as usual. Research assistants and investigators were masked to treatment allocation; the participants could not be masked. The primary outcome was out-of-home placement at 18 months. The primary analysis included all randomised participants for whom data were available. This trial is registered, number ISRCTN77132214. Follow-up of the trial is still ongoing. FINDINGS: Between Feb 4, 2010, and Sept 1, 2012, 1076 families were referred to nine multi-agency panels, 684 of whom were assigned to management as usual (n=342) or multisystemic therapy followed by management as usual (n=342). At 18 months, the proportion of participants in out-of-home placement was not significantly different between the groups (13% [43/340] in the multisystemic therapy group vs 11% [36/335] in the management-as-usual group; odds ratio 1·25, 95% CI 0·77-2·05; p=0·37). INTERPRETATION: The findings do not support that multisystemic therapy should be used over management as usual as the intervention of choice for adolescents with moderate-to-severe antisocial behaviour. FUNDING: Department for Children, Schools and Families, Department of Health.

Growing Up? A History of CAMHS (1987-2005).

A brief history of child and adolescent mental health services (CAMHS) from 1987 to 2005 is presented. Events of the last 20 years are organised around the two key themes of changes and development in policy and in clinical practice. This has been a period of great change in CAMHS and we attempt to highlight what we believe are the significant changes in what is, inevitably a personal view of recent history.

On the origin of the clinical standpoint in psychiatry, Dr Ewald Hecker in Görlitz.

Ewald Hecker (1843-1909), a friend and disciple of Karl Ludwig Kahlbaum (1828-1899), was a relentless advocate of his teacher's psychiatric nosology. This paper is an early manifesto of their ideas and sets the context for the following publications, namely Hecker's seminal paper on hebephrenia to be published in the same journal and in the same year (1871) and Kahlbaum's catatonia published in 1874. Their idea that age of onset and time course of an illness, together with close clinical observation, helps to delineate disease forms out of the mass of confusing psychiatric symptoms proved to be one of the most important paradigm shifts in middle to late nineteenth-century psychiatry. This had a strong influence on Kraepelin's dichotomy between dementia praecox and manic depressive insanity, and thus on our modern notions of schizophrenia and bipolar illness.

Evaluation of multisystemic therapy pilot services in the Systemic Therapy for At Risk Teens (START) trial: study protocol for a randomised controlled trial.

BACKGROUND: There is an urgent need for clinically effective and cost-effective methods to manage antisocial and criminal behaviour in adolescents. Youth conduct disorder is increasingly prevalent in the UK and is associated with a range of negative outcomes. Quantitative systematic reviews carried out for the National Institute for Health and Clinical Excellence have identified multisystemic therapy, an intensive, multimodal, home-based, family intervention for youth with serious antisocial behaviour, as one of the most promising interventions for reducing antisocial or offending behaviour and improving individual and family functioning. Previous international trials of multisystemic therapy have yielded mixed outcomes, and it is questionable to what extent positive US findings can be generalised to a wider UK mental health and juvenile justice context. This paper describes the protocol for the Systemic Therapy for At Risk Teens (START) trial, a multicentre UK-wide randomised controlled trial of multisystemic therapy in antisocial adolescents at high risk of out-of-home placement. METHODS/DESIGN: The trial is being conducted at 10 sites across the UK. Seven hundred participants and their families will be recruited and randomised on a 1:1 basis to multisystemic therapy or management as usual. Treatments are offered over a period of 3 to 5 months, with follow-up to 18 months post-randomisation. The primary outcome is out-of-home placement at 18 months. Secondary outcomes include offending rates, total service and criminal justice sector costs, and participant well-being and educational outcomes. Data will be gathered from police computer records, the National Pupil Database, and interview and self-report measures administered to adolescents, parents and teachers. Outcomes will be analysed on an intention-to-treat basis, using a logistic regression with random effects for the primary outcome and Cox regressions and linear mixed-effects models for secondary outcomes depending on whether the outcome is time-to-event or continuous. DISCUSSION: The START trial is a pragmatic national trial of sufficient size to evaluate multisystemic therapy, to inform policymakers, service commissioners, professionals, service users and their families about its potential in the UK. It will also provide data on the clinical and cost-effectiveness of usual services provided to youth with serious antisocial behaviour problems. TRIAL REGISTRATION: ISRCTN77132214.