Towards more accurate classification of risk of arrest among offenders on community supervision: An application of machine learning versus logistic regression

BACKGROUND

Although there is general consensus about the behavioural, clinical and sociodemographic variables that are risk factors for reoffending, optimal statistical modelling of these variables is less clear. Machine learning techniques offer an approach that may provide greater accuracy than traditional methods.

AIM

To compare the performance of advanced machine learning techniques (classification trees and random forests) to logistic regression in classifying correlates of rearrest among adult probationers and parolees in the United States.

METHOD

Data were from the subgroup of people on probation or parole who had taken part in the National Survey on Drug Use and Health for the years 2015-2019. We compared the performance of logistic regression, classification trees and random forests, using receiver operating characteristic curves, to examine the correlates of arrest within the past 12 months.

RESULTS

We found that machine learning techniques, specifically random forests, possessed significantly greater accuracy than logistic regression in classifying correlates of arrest.

CONCLUSIONS

Our findings suggest the potential for enhanced risk classification. The next step would be to develop applications for criminal justice and clinical practice to inform better support and management strategies for former offenders in the community.

Correlates of Homelessness Among Adults with Personality Disorder

Although personality disorders (PDs) are more common among persons experiencing homelessness than the general population, few studies have investigated the risk of experiencing homelessness among persons with PDs. This study seeks to identify the demographic, socioeconomic, and behavioral health correlates of past-year homelessness among persons with antisocial, borderline, and schizotypal PDs. Nationally representative data of the civilian, noninstitutionalized population of the United States was used to identify correlates of homelessness. Descriptive statistics and bivariate associations between variables and homeless status were summarized prior to conducting several multivariate logistic regression models to identify correlates of homelessness. Main findings revealed positive associations between poverty, relationship dysfunction, and lifetime suicide attempt with homelessness. In the antisocial PD (ASPD) and borderline PD (BPD) models, comorbid BPD and ASPD, respectively, were associated with higher odds of past-year homelessness. Findings underscore the importance of poverty, interpersonal difficulties, and behavioral health comorbidities on homelessness among persons with ASPD, BPD, and schizotypal PD. Strategies to promote economic security, stable relationships, and interpersonal functioning may buffer against the effects of economic volatility and other systemic factors that could contribute to homelessness and persons with PD.

Self-management interventions for reducing challenging behaviors among school-age students: A systematic review

BACKGROUND

Challenging classroom behaviors can interfere with student social and academic functioning and may be harmful to everyone in schools. Self-management interventions within schools can address these concerns by helping students develop necessary social, emotional, and behavioral skills. Thus, the current systematic review synthesized and analyzed school-based self-management interventions used to address challenging classroom behaviors.

OBJECTIVES

The current study aimed to inform practice and policy by (a) evaluating the effectiveness of self-management interventions at improving classroom behaviors and academic outcomes and (b) examining the state of research for self-management interventions based on existing literature.

SEARCH METHODS

Comprehensive search procedures included electronically searching online databases (e.g., EBSCO Academic Search Premier, MEDLINE, ERIC, PsycINFO), hand-searching 19 relevant journals (e.g., School Mental Health, Journal of School Psychology), reference-list searching 21 relevant reviews, and searching gray literature (e.g., contacting authors, searching online dissertation/theses databases and national government clearinghouses/websites). Searches were completed through December of 2020.

SELECTION CRITERIA

Included studies employed either a multiple group-design (i.e., experimental or quasi-experimental) or single-case experimental research design and met the following criteria: (a) utilized a self-management intervention, (b) conducted in a school setting, (c) included school-aged students, and (d) assessed classroom behaviors.

DATA COLLECTION AND ANALYSIS

Standard data collection procedures expected by the Campbell Collaboration were used in the current study. Analyses for single-case design studies incorporated three-level hierarchical models to synthesize main effects, and meta-regression for moderation. Further, robust variance estimation was applied to both single-case design and group-design studies to account for dependency issues.

MAIN RESULTS

Our final single-case design sample included 75 studies, 236 participants, and 456 effects (i.e., 351 behavioral outcomes and 105 academic outcomes). Our final group-design sample included 4 studies, 422 participants, and 11 total behavioral effects. Most studies occurred in the United States, in urban communities, in public schools, and in elementary settings. Single-case design results indicated that self-management interventions significantly and positively impacted both student classroom behaviors (LRRi = 0.69, 95% confidence interval [CI] [0.59, 0.78]) and academic outcomes (LRRi = 0.58, 95% CI [0.41, 0.76]). Single-case results were found to be moderated by student race and special education status, whereas intervention effects were more pronounced for African American students (F = 5.56, p = 0.02) and students receiving special education services (F = 6.87, p = 0.01). Single-case results were not found to be moderated by intervention characteristics (i.e., intervention duration, fidelity assessment, fidelity method, or training). Despite positive findings for single-case design studies, risk of bias assessment indicated methodological shortcomings that should be considered when interpreting findings. A significant main effect of self-management interventions for improving classroom behaviors was also revealed for group-design studies (g = 0.63, 95% CI [0.08, 1.17]). However, these results should be interpreted with caution given the small number of included group-design studies.

IMPLICATIONS FOR POLICY PRACTICE AND RESEARCH

The current study, conducted using comprehensive search/screening procedures and advanced meta-analytic techniques, adds to the large amount of evidence indicating that self-management interventions can be successfully used to address student behaviors and academic outcomes. In particular, the use specific self-management elements (i.e., self-determining a performance goal, self-observing and recording progress, reflecting on a target behavior, and administering primary reinforcers) should be considered within current interventions as well as in the development of future interventions. Future research should aim to assess the implementation and effects of self-management at the group or classroom-level within randomized controlled trials.

Helping Survivors of Human Trafficking: A Systematic Review of Exit and Postexit Interventions

BACKGROUND

Human trafficking is a global problem and results in deleterious psychological, social, and physical effects on the lives of those who are trafficked; however, it is not clear how to best intervene with survivors. The purpose of this review was to synthesize the evidence of exit and postexit intervention programs for survivors of human trafficking to inform practice and research.

METHOD

Systematic review methods were used to search, select, and extract data from published and unpublished experimental, quasi-experimental, and preexperimental studies that assessed the effects of any exit or postexit interventions for victims of human trafficking. The authors searched eight databases, reviewed bibliographies, and conducted forward citation searches from relevant reports and prior reviews to find studies authored between 2005 and 2015.

RESULTS

The search yielded six eligible studies that included 155 female and 6 male survivors from four countries. Interventions were diverse, with three using a trauma-informed approach. Authors measured a myriad of outcomes, including mental health, social network, community reintegration, and employment; however, the quality of most studies was poor.

DISCUSSION

Evidence of effects of exit and postexit interventions is sparse, and much of the research is poorly designed and executed; however, the needs of trafficking survivors are complex and effective interventions are desperately needed. Implications for practice and research are discussed.

Truancy in the United States: Examining Temporal Trends and Correlates by Race, Age, and Gender

BACKGROUND

Truancy has long been regarded a common problem in urgent need of effective intervention. Knowledge about factors associated with truancy can guide the development and implementation of interventions.

METHOD

This paper examined trends in truancy rates between 2002-2014 and correlates of truancy across racial/ethnic groups. Variables of interest included sociodemographic factors (e.g., age, gender, socio-economic background), behavioral factors (e.g., substance use, violence), and psychosocial factors (e.g., academic engagement, grades, parental control). Using data from a large sample of adolescents (n=209,393; 12-17 years) we estimated truancy prevalence rates and examined trends and correlates via regression analyses.

RESULTS

Truancy rates remained constant between 2002 (10.8%) and 2014 (11.1%). Rates were highest among older youth, females, and Hispanic youth. For all racial/ethnic groups, truancy was significantly correlated with alcohol and marijuana use, fighting, the propensity to take risks, and lower academic engagement and school grades. Other factors were differentially associated with racial/ethnic groups. This divergence in risk patterns for different racial/ethnic groups points to some heterogeneity amongst truant youth.

DISCUSSION

Despite truancy reduction efforts, truancy rates have remained stable. Efforts to prevent truancy and to intervene with truant youth may need to target risk factors more prevalent in specific racial/ethnic groups.

Bullying Victimization Among School-Aged Immigrant Youth in the United States

PURPOSE

Bullying is a serious sociodevelopmental issue associated with a range of short- and long-term problems among youth who are bullied. Although race and ethnicity have been studied, less attention has been paid to examining prevalence and correlates of bullying victimization among immigrant youth.

METHODS

Using data from the Health Behavior in School-Aged Children (N = 12,098), we examined prevalence and correlates of bullying victimization among U.S. immigrant youth.

RESULTS

After controlling for several demographic variables, findings indicate that immigrant youth are more likely to experience bullying victimization than native-born youth. Furthermore, immigrant youth who experience bullying victimization were more likely to report interpersonal, socioemotional, health, and substance use problems.

CONCLUSIONS

Given the greater risk and unique challenges experienced by immigrant youth, prevention and intervention programs may need to be tailored to their specific needs and circumstances. Further research is needed to understand the specific factors and mechanisms involved in bullying victimization among immigrant youth.

Are homeschooled adolescents less likely to use alcohol, tobacco, and other drugs?

BACKGROUND

Nearly two million school-aged children in US are currently homeschooled. This study seeks to examine homeschooled adolescents' attitudes toward, access to, and use of alcohol, tobacco, and other drugs (ATOD) compared to their non-homeschooled peers.

METHODS

The study uses data between 2002 and 2013 from the National Survey on Drug Use and Health (NSDUH) for school-attending respondents aged 12-17 (n=200,824). Participants were questioned regarding peer use of licit and illicit substances, ease of accessing illicit substances, and past 12-month substance use. Survey adjusted binary logistic regression analyses were systematically executed to compare non-homeschooled adolescents with homeschooled adolescents with respect to views toward, access to, and use of substances.

RESULTS

Findings indicate that homeschooled adolescents were significantly more likely to strongly disapprove of their peers drinking (AOR=1.23) and trying (AOR=1.47) and routinely using (AOR=1.59) marijuana. Homeschooled adolescents were significantly less likely to report using tobacco (AOR=0.76), alcohol (AOR=0.50), cannabis (AOR=0.56) and other illicit drugs and to be diagnosed with an alcohol (AOR=0.65) or marijuana (AOR=0.60) use disorder. Finally, homeschooled adolescents were also less likely to report easier access to illicit drugs and to be approached by someone trying to sell drugs compared to non-homeschooled peers.

CONCLUSIONS

Homeschooled adolescents' views, access, use and abuse of ATOD are uniquely different from those of non-homeschooled adolescents. Findings point to the need to more extensively examine the underlying mechanisms that may account for these differences.

Are immigrants more likely than native-born Americans to perpetrate intimate partner violence?

Despite an emerging body of research indicating that immigrants are less likely than native-born Americans to engage in crime and antisocial behavior, less attention has focused specifically on intimate partner violence (IPV) perpetration among immigrant populations. We address this gap by using data from Wave II of the National Epidemiologic Survey on Alcohol and Related Conditions (NESARC) and compare immigrants from Asia, Africa, Europe, and Latin America to native-born Americans with respect to multiple forms of IPV. After controlling for an extensive array of confounds, results indicate that in the aggregate, immigrants are significantly more likely to perpetrate IPV. However, examination of major world regions indicates these results are driven by Latin American immigrants. Immigrants from Asia, Africa, and Europe report a lower prevalence of IPV perpetration than native-born Americans. This study extends prior research on the immigrant paradox and suggests that future studies take into account regional heterogeneity when examining IPV and other forms of violence in immigrant populations.

Prevalence and correlates of psychiatric disorders among former juvenile detainees in the United States

OBJECTIVE

Juvenile offenders face increased liability for psychiatric disorders and greater psychopathology, but little is known about the psychiatric status of former juvenile delinquents as adults.

METHOD

Drawing on data from Wave 1 and Wave 2 of the NESARC, logistic regression models examine correlates of psychiatric disorders in a large nationally representative sample of former juvenile detainees in adulthood (n=1177) compared to adults who did not have a history of juvenile offending (n=33,193). Further, we explored the psychosocial correlates associated with the increased likelihood of psychiatric disorders among former juvenile detainees.

RESULTS

Nearly half of former juvenile detainees met criteria for one or more psychiatric disorders in the past twelve months and approximately two-thirds meet criteria for any lifetime personality disorder. Compared to the general population, former juvenile detainees not only denote greater psychiatric comorbidity across a range of affective, personality, and substance use disorders but are also more likely to report childhood adversity.

CONCLUSIONS

Former juvenile detainees experience significantly greater and more varied psychiatric problems across adulthood.

High school dropouts in emerging adulthood: substance use, mental health problems, and crime

This study examined the distribution of substance use, mental health, and criminal behavior among dropouts derived from a nationally representative sample of 18-25 year old (N = 19,312) emerging adults in the United States. Using public-use data from the 2010 National Survey on Drug Use and Health, this study employed multiple logistic regression with adjustments for complex survey sampling and compared high school dropouts with graduates with respect to substance use, mental health, and criminal behavior. After controlling for the effects of age, gender, race/ethnicity, family income, receipt of government assistance, employment status, and metropolitan population density, dropouts were more likely to meet criteria for nicotine dependence and report daily cigarette use, and more likely to report having attempted suicide in the previous year, been arrested for larceny, assault, drug possession or drug sales relative to their high school graduate counterparts. The findings of this study provide important insights and an initial epidemiologic portrait of mental health, substance use, and criminal behaviors of dropouts during emerging adulthood.

Practitioner perspectives of implementing check & connect

While there is a growing reserve of evidence-based practices (EBPs) available to practitioners, much can be learned about how to implement EBPs in real-world settings. Evidence of the effects of a widely disseminated student engagement intervention, Check & Connect (C&C), is emerging yet little is known about the implementation of C&C in community-based settings. The purpose of the authors in this study was to examine practitioner attitudes and perspectives related to the C&C intervention and implementation to gain an understanding of core implementation components that facilitated or impeded implementation. A researcher-developed survey instrument was used to assess practitioner attitudes related to the C&C model and implementation among 14 school-based practitioners working in a dropout prevention program. Findings indicate that practitioners were highly positive about the C&C intervention and in their attitudes about implementing EBPs. Benefits of C&C identified by practitioners included increased relationship building with students, tracking students on a consistent and timely basis, and addressing attendance issues as a main focus of treatment. The most common implementation challenges were time constraints, paperwork, and targeting absentee students. These findings contribute to the emerging literature on C&C and the implementation of EBPs in schools and community-based settings.

Gambling prevalence rates among immigrants: a multigenerational examination

INTRODUCTION

The present study employed data from Waves I and II of the National Epidemiologic Survey of Alcohol and Related Conditions (NESARC) to compare gambling prevalence rates across gender and world regions (e.g., Africa, Asia, Europe, and Latin America).

METHODS

Responses from first generation (n=5363), second generation (n=4826), third generation (n=4746), and native-born Americans (n=19,715) were subjected to a series of multinomial regression analyses, after controlling for sociodemographic variables such as age, gender, race/ethnicity, household income, education level, region of the United States, and urbanicity.

RESULTS

The prevalence of gambling and problem gambling was markedly lower among first-generation immigrants than that of native-born Americans and second and third-generation immigrants. Results also point to inter- and intra-generational dynamics related to gender, age of arrival and duration in the United States, and world region from which participants emigrated. Additionally, we found that second-generation immigrants and nonimmigrants were significantly more likely to meet criteria for disordered gambling compared to first-generation immigrants in general.

CONCLUSIONS

Compared to first-generation immigrants, male and female immigrants of subsequent generations and nonimmigrants were significantly more likely to report involvement in all problem gambling behaviors examined. Findings suggest that gambling prevalence rates increase across subsequent generations, and are more likely to occur in women than among men.

Effects of after-school programs with at-risk youth on attendance and externalizing behaviors: a systematic review and meta-analysis

The popularity, demand, and increased federal and private funding for after-school programs have resulted in a marked increase in after-school programs over the past two decades. After-school programs are used to prevent adverse outcomes, decrease risks, or improve functioning with at-risk youth in several areas, including academic achievement, crime and behavioral problems, socio-emotional functioning, and school engagement and attendance; however, the evidence of effects of after-school programs remains equivocal. This systematic review and meta-analysis, following Campbell Collaboration guidelines, examined the effects of after-school programs on externalizing behaviors and school attendance with at-risk students. A systematic search for published and unpublished literature resulted in the inclusion of 24 studies. A total of 64 effect sizes (16 for attendance outcomes; 49 for externalizing behavior outcomes) extracted from 31 reports were included in the meta-analysis using robust variance estimation to handle dependencies among effect sizes. Mean effects were small and non-significant for attendance and externalizing behaviors. A moderate to large amount of heterogeneity was present; however, no moderator variable tested explained the variance between studies. Significant methodological shortcomings were identified across the corpus of studies included in this review. Implications for practice, policy and research are discussed.

Toward a Bioecological Model of School Engagement: A Biometric Analysis of Gene and Environmental Factors

School disengagement is associated with poor academic achievement, dropout, and risk behaviors such as truancy, delinquency, and substance use. Despite empirical research identifying risk correlates of school disengagement across the ecology, it is unclear from which domain these correlates arise. To redress this issue, the current study used intraclass correlation and DeFries-Fulker analyses to longitudinally decompose variance in three domains of engagement (academic, behavioral, and emotional) using data from the National Longitudinal Study of Adolescent Health. Findings suggest that nonshared environmental factors (that is, environmental contexts and experiences that are unique to each sibling) account for approximately half of the variance in indicators of school disengagement when controlling for genetic influences, and that this variance increases as adolescents grow older and rely less on their immediate family. The present study contributes new evidence on the biosocial underpinnings of school engagement and highlights the importance of interventions targeting factors in the nonshared environment.

Caution or warning? A validity study of the MAYSI-2 with juvenile offenders

Given the high prevalence of mental health and substance use disorders in the juvenile justice system and the emphasis on efficient screening and referrals, the current study sought to further validate the use of the Massachusetts Youth Screening Instrument version 2 (MAYSI-2). Using data from institutionalized delinquents (N=836) committed to the California Youth Authority, finite mixture modeling was employed to identify distinct latent classes based on MAYSI-2 scale scores. Identified classes were then compared across a range of covariates, including prior offenses, official records of misconduct, and multiple measures of mental health and psychological well-being. Findings revealed a three-class, gradient-based structure: low distress (n=380), moderate distress (n=327), and high distress (n=129). Overall, the MAYSI-2 identified juvenile offenders with mental health and substance use issues, but it did not differentiate youths with respect to offending patterns.

The immigrant paradox: immigrants are less antisocial than native-born Americans

PURPOSE

Although recent research on crime and violence among immigrants suggests a paradox--where immigrants are more socially disadvantaged yet less likely to commit crime--previous research is limited by issues of generalizability and assessment of the full depth of antisocial behavior.

METHODS

We surmount these limitations using data from waves I and II of the National Epidemiologic Survey on Alcohol and Related Conditions (NESARC) and compare immigrants (N = 7,320) from Asia, Africa, Europe, and Latin America to native-born Americans (N = 34,622) with respect to violent and nonviolent forms of antisocial behavior.

RESULTS

After controlling for an extensive array of confounds, results indicate that immigrants are significantly less antisocial despite being more likely to have lower levels of income, less education, and reside in urban areas. These findings hold for immigrants from major regions of the world including Africa, Asia, Europe, and Latin America.

CONCLUSIONS

This study confirms and extends prior research on crime and antisocial behavior, but suggests that it is premature however to think of immigrants as a policy intervention for treating high crime areas.

Alcohol use among Hispanic early adolescents in the United States: an examination of behavioral risk and protective profiles

Few studies have examined the behavioral and protective correlates of alcohol use among young Hispanics. Using a national sample (N = 7,606), logistic regression and latent profile analysis (LPA) are employed to examine the relationships between alcohol use, psychosocial factors, and externalizing behavior among Hispanics during early adolescence. Early drinkers are more likely to report truancy, fighting, smoking, and drug use. LPA results revealed a three class solution. Classes identified included: psychosocial risk (41.11%), moderate protection (39.44%), and highly religious (19.44%). Alcohol use is clearly associated with externalizing behavior; however, an important degree of psychosocial and behavioral heterogeneity nevertheless exists.

Religiosity and violence among adolescents in the United States: findings from the national survey on drug use and health 2006-2010

There is a pressing need to develop a more nuanced understanding of the relationships between particular expressions of religiosity and the various manifestations of violence among youth. This study examines these relationships among adolescents in the general population as well as across racial/ethnic, gender, and family income differences. Using a nationally representative sample of adolescents (N = 90,202) from the National Survey on Drug Use and Health (2006-2010), logistic regression is used to examine the relationships between religiosity and violence. Results indicate that multiple components of adolescent religiosity are associated with the decreased likelihood of fighting, group fighting, and, to a lesser extent, violent attacks. A number of noteworthy differences were identified across race/ethnicity, gender, and family income. Findings from this investigation shed light on the relationship between particular facets of religiosity and violence that may be useful for violence prevention organizations seeking to integrate religious components into intervention efforts.

Dropping out of school and chronic disease in the United States

AIM

Despite links between education and health, it is unclear to what extent dropping out of school is associated with major chronic health conditions.

SUBJECTS AND METHODS

Data from the 2006-2010 United States National Survey on Drug Use and Health (NSDUH) was employed and examined associations between dropping out of school and major chronic health conditions among individuals 18 years or older (N=189,896).

RESULTS

Analyses show that dropout status is associated with increased odds of reporting a major chronic health condition; however, the effect is stronger for Whites and African-Americans than Hispanics.

CONCLUSIONS

Study findings suggest that one important strategy to prevent and reduce health disparities is increased education in general including high school completion. Policies and practices that reduce dropout may in turn impact the prevalence of chronic disease.

Prevalence and correlates of truancy in the US: results from a national sample

Truancy has been a persistent problem in the United States for more than 100 years. Although truancy is commonly reported as a risk factor for substance use, delinquency, dropout, and a host of other negative outcomes for youth, there has been surprisingly little empirical investigation into understanding the causes and correlates of truancy using large, nationally representative samples. Using the adolescent sample (N = 17,482) of the 2009 National Survey on Drug Use and Health (NSDUH), this study presents the prevalence of truancy and examines individual, school engagement, parental, and behavioral correlates of truancy. Overall, 11% of adolescents between the ages of 12-17 reported skipping school in the past 30 days. Results from multinomial logistic regression models indicate skipping school was robustly associated with an increased probability of reporting externalizing behaviors, less parental involvement, and engagement and lower grades in school. Implications for theory, prevention, and policy are discussed.

Who are truant youth? Examining distinctive profiles of truant youth using latent profile analysis

The present study explored the heterogeneity of truant youth to provide a more nuanced examination of the nature of adolescent truancy and examine distinct profiles of truant youth as they relate to externalizing behaviors. Latent profile analysis was employed to examine the heterogeneity of truant youth by using a nationally representative sample of 1,646 truant adolescents (49.8 % female) from the 2010 National Survey on Drug Use and Health. Five key indicator variables were utilized to identify latent classes: school engagement, participation in school-based activities, grades, parental academic involvement, and number of school days skipped. Additionally, multinomial regression was employed to examine the relationship between latent truant youth classes and externalizing behaviors. Four classes of truant youth were identified: achievers (28.55 %), moderate students (24.30 %), academically disengaged (40.89 %), and chronic skippers (6.26 %). Additionally, group membership was found to be associated differentially with marijuana use, fighting, theft and selling drugs. Results from the present study suggest that truant youth are not a homogenous group, but rather present with different risk profiles as they relate to key indicators, demographic characteristics and externalizing behaviors. Implications for practice, policy and future research are discussed.

Transitional cell carcinoma of the fossa navicularis of the male urethra

We report a case of transitional cell carcinoma arising in the fossa navicularis, a portion of the male urethra ordinarily lined by squamous epithelium. This is the third reported case of transitional cell carcinoma of the distal urethra. The neoplasm metastasized to superficial inguinal lymph nodes.