Delinquency best treatments: how to divert youths from violence while saving lives and detention costs.

Youth development and violence prevention are two sides of the same public policy. The focus of much theoretical and empirical effort is identifying delinquency risks and intervening. Given the great costs of homicide and the historically high nationwide prison population, new policies must address increasing violence and rising expenses. Treatments of prenatal care, home visitation, bullying prevention, alcohol-substance abuse education, alternative thinking promotion, mentoring, life skills training, rewards for graduation and employment, functional family and multi-systemic therapy, and multi-dimensional foster care are effective, because they ameliorate age-specific risks for delinquency. At present, these interventions only yield a 10-40% diversion from crime however. Returns on investment (ROIs) vary from $1 to $98. Targeting empirical treatments to those determined to be most at risk, based on statistical models or actuarial testing, and using electronic surveillance for non-violent prisoners significantly diverted youth from violence, improving ROI, while simultaneously saving costs.

Finding workers, offenders, or students most at-risk for violence: actuarial tests save lives and resources.

147 adults (107 men, 40 women) and 89 adolescents (61 boys, 28 girls), selected randomly from referrals and volunteers, were given the Ammons Quick Test (QT), the Beck Suicide Scale (BSS), the Minnesota Multiphasic Personality Inventory Second (MMPI-2) or Adolescent Versions (MMPI-A), the Raven's Advanced Progressive Matrices, and the Standard Predictor (SP) of Violence Potential Adult or Adolescent Versions. The goals were to: (a) demonstrate computer and paper-and-pencil tests correlated; (b) validate tests to identify at-risk for violence; (c) show that identifying at-risk saves lives and resources; and (d) find which industries benefited from testing at-risk. Paper-and-pencil vs. computer test correlations (.83-.99), sensitivity (.97-.98), and specificity (.50-.97) were computed. Testing at-risk saves lives and resources. Critical industries for testing at-risk individuals may include airlines, energy generating industries, insurance, military, nonprofit-religious, prisoners, trucking or port workers, and veterans.

Early iron deficiency enhances stimulus-response learning of adult rats in the context of competing spatial information.

Iron deficiency early in life results in neurocognitive deficits that persist into adulthood despite iron treatment. The hippocampus is particularly vulnerable to iron deficiency during the fetal and neonatal periods as evidenced by poorer hippocampus-mediated spatial recognition learning. However, the extent to which early iron deficiency alters interactions between hippocampus-based and extra-hippocampus based learning systems remains undetermined. The present study used an ambiguous maze-learning task to examine the learning process in iron sufficient young adult rats that had recovered from iron deficiency in the fetal and neonatal period. Animals were presented with a stimulus response-learning task in the context of spatial information; a procedure designed to elicit competition between dorsal striatum- and hippocampus-based systems respectively. Formerly iron deficient adult rats showed enhanced stimulus-response learning in the context of competing spatial/distal cue information, a finding suggestive of reduced hippocampal functional influence. The study provides evidence that early iron deficiency alters how different learning systems develop and ultimately interact in adulthood. The potential unbalancing of activity among major memory systems during early life has been postulated by others as a relevant factor underlying the developmental origins of certain psychiatric disorders.

Delayed alternation performance in rats following recovery from early iron deficiency.

Early iron deficiency (ID) is one of the most common nutrient deficiencies in both developed and developing countries. This condition has been linked to perturbations in myelin formation, alterations of monoamine neurotransmitter systems particularly in the striatum, and deficits in energy metabolism particularly in the hippocampus (HP) and prefrontal cortex (PFC) in rats. Early ID has also been traced to long-term behavioral consequences in children in domains linked to these neuropathologies. The current experiment assesses formerly iron deficient (FID) adult rats on a delayed alternation (DA) task - a procedure thought to be sensitive to PFC dysfunction. Rat dams were started on an iron deficient chow at gestational day (G) 2 and maintained on this diet until postnatal day (P) 7; behavioral training began at P 65 when animals were iron replete. FID animals exhibited accelerated acquisition (p=0.002) and fewer errors (p=0.003) on the DA task compared to controls. These findings may reflect an imbalance between hippocampal and prefrontal modulation of this behavior most likely emanating from long-term hippocampal disinhibition by early ID that persists in spite of early iron treatment from P 7.

Multidimensional assessment of criminal recidivism: problems, pitfalls, and proposed solutions.

All states have statutes in place to civilly commit individuals at high risk for violence. The authors address difficulties in assessing such risk but use as an example the task of predicting sexual violence recidivism; the principles espoused here generalize to predicting all violence. As part of the commitment process, mental health professionals, who are often psychologists, evaluate an individual's risk of sexual recidivism. It is common for professionals conducting these risk assessments to use several actuarial risk prediction instruments (i.e., psychological tests). These tests rarely demonstrate close agreement in the risk figures they provide. Serious epistemological and psychometric problems in the multivariate assessment of recidivism risk are pointed out. Sound psychometric, or in some cases heuristic, solutions to these problems are proffered, in the hope of improving clinical practice. The authors focus on how to make these tests' outputs commensurable and discuss various ways to combine them in coherent, justifiable fashions.

Violence risk appraisal of male and female youth, adults, and individuals.

Data from youth (n = 1,127), adults (n = 1,595), and individuals (N = 2,722, combined youth and adults) were followed 3 to 12 years in records to develop items to predict abuse, violence, and homicide in these 3 relevant groups for risk appraisal or safety scales. Shao's bootstrapped logistic regression yielded 14 predictors for youth (AUC = .91), 11 for adults (AUC = .99), and 13 for individuals (AUC = .96). Three regression equations were cross-validated with in-bag and out-of-bag techniques. Pearson coefficients were computed with intelligence, achievement, adaptive behavior, and perception tests. Test-retest reliability was acceptable. Using case-control quasi-experimental design, this study extends probation-parole decision-making tests to infants and children as young as 3 years, with convergent and divergent validity and reliability with other tests. Sensitivity and specificity were high and minimized over- or under-identification challenges in identifying potentially violent persons in the general population.

Summary of studies of abused infants and children later homicidal, and homicidal, assaulting later homicidal, and sexual homicidal youth and adults.

To study the risks of abuse, violence, and homicide, 5 studies of groups at risk for violence are summarized. 192 Abused Infants, 181 Abused Children, 127 Homicidal Youth, 425 Assaulters, 223 Rapists, and 223 Molesters were randomly selected and tracked in court, probation, medical, and school records, then compared with carefully matched groups of Controls and (in older groups) Nonviolent Delinquents. In adolescence or adulthood, these groups were classified into Later Homicidal (N=234), Later Violent or Nonviolent Delinquent, and Later Nondelinquent subgroups for more detailed comparisons. Shao's bootstrapped logistic regressions were applied to identify risks for commission of homicide. Significant predictors for all homicidal cases in these samples were number of court contacts, poorer executive function, lower social maturity, alcohol abuse, and weapon possession. Predictors for the 373 Abused cases (Infants and Children) were court contacts, injury, burn, poisoning, fetal substance exposure, and parental alcohol abuse. Predictors for the 871 Violent Delinquent cases (Assaulters, Rapists, Molesters) were court contacts, poorer executive function, and lower social maturity. Accuracies of prediction from the regressions ranged from 81% for homicidal sex offenders to 87 to 99% for other homicidal groups.

Looking forward in records of youth abused as infants: risks for homicidal, violent, and delinquent offenses.

To study risks for abuse and later homicidal behavior, 192 abused infants (M age = 3.12 yr., SD = 1.48; 82 girls, 110 boys) and 192 controls were matched on demographics and examined; data discriminating abused and later homicidal cases were analyzed with Shao's bootstrapped logistic regression. Predictors of Abused status were injury, burn, poisoning, fetal substance exposure (OR = 2.47), later parental or youth court contacts (OR = 1.86e+12), and parental alcohol abuse (OR = .54; AUC = .99; 95% CI = .96-.99). Youth tracked through records 12 years (to M age = 15.17 [corrected] yr., SD = 1.89) were classified into Abused Later Homicidal (11%, n = 21), Abused Later Violent (14%, n = 27), Abused Later Delinquent (31%, n = 60), Abused Later Nondelinquent (n = 44), and Control groups (n = 192). Data were analyzed similarly. When the Abused Later Homicidal was contrasted with the Control group, predictors of homicide were three or more home/school moves (OR = .78), illnesses (OR = .90), and later court contacts (OR = 1.75e+07; AUC = .99; 95% CI = .90-.98). When the Abused Later Homicidal was compared with the Abused Later Nondelinquent group, predictors of homicide were poorer executive function (OR = 2.29) and later court contacts (OR = 7.78e+06; AUC = .94; 95% CI = .90-.98).

Looking forward in records of youth abused as children: risks for homicidal, violent, and delinquent offenses.

To study risks of abuse, violence, and homicide, 181 Abused Children (M age = 12.85 yr., SD = 2.74; 58 girls, 123 boys) were matched with 181 clinic-referred Controls. Data analysis was Shao's bootstrapped logistic regression with area under curve (AUC) and odds ratios (OR). Predictors of abused status were court contacts (OR = 2.04e+22) and poorer executive function (OR = .81; AUC = .99; 95% CI = .97-.99). Groups were tracked forward in records for 9 years (M = 8.78 yr., SD = 1.41). Looking forward, youth (M age = 21.63 yr., SD = 2.07) were classified into Abused Children Later Homicidal (5%, n = 10), Abused Children Later Violent (23%, n = 41), Abused Children Later Delinquent (28%, n = 50), Abused Children Later Nondelinquent (44%, n = 80), and Controls (n = 181). Data were analyzed with two more logistic regressions. Predictors of Abused Children Later Homicidal compared with Controls were number of court contacts (OR = 50,398.78) and poorer executive function (OR = 79.72; AUC = .91; 95% CI = .80-.95). The predictor of Abused Children Later Homicidal contrasted with Abused Children Later Nondelinquent was court contacts (OR = 2,077,089,352; AUC = .87; 95% CI = .65-.95). The common predictor for Abused Children and Abused Children Later Homicidal groups was court contacts.

Looking forward in records of young adults who were convicted of homicide or assault as youth: risks for reoffending.

To assess risks of violent offending in young adults, 425 Delinquent Assaulters (M age = 14.1 yr., SD = 1.7; 77 girls, 348 boys) were matched with 425 Nonviolent Delinquents. Analysis of data from court, school, and medical records used Shao's bootstrapped logistic regressions. Predictors of Assaulter status were poorer executive function (OR = 0.97) and prior court contacts for violent offenses (OR = 3.5e+ 23; AUC = .97; 95% CI = .82-.99). Looking in records backward 4 years (M = 4.1, SD = 2.6) and forward 10 years to mean age 24.5 yr. (SD = 2.1), adults were classified as Homicidal (8%, n = 69); Delinquent Assaulters Later Adult Assaulters (10%, n = 86); Delinquent Assaulters Later Noncriminals (32%, n = 270); Nonviolent Delinquents Later Nonviolent Criminals (10%, n = 87); and Nonviolent Delinquents Later Noncriminals (40%, n = 338). The Homicidal group (n = 69) was compared to matched Control and Nonviolent Delinquent groups (n = 69) using logistic regression. Predictors of Homicidal versus Control were poorer executive function and alcohol or substance abuse (AUC = .97; 95% CI = .93-.99). Predictors of Homicidal versus Nonviolent Delinquents were unemployment, poorer executive function, and prior court contacts (AUC = .98; 95% CI = .95-.99).

Looking forward and backward in records for risks among homicidal youth.

To identify risks for commission of homicide, 26 convicted Homicidal Youth (M age = 14.9 yr., SD = 1.4; n = 26; 1 girl, 25 boys) were matched with 26 Nonviolent Delinquents and 26 clinic-referred Controls. Youth were tracked backward 8 years (M = 7.7 yr., SD = 15) and forward 3 years (M = 3.1 yr., SD = 1.2) in records. Data analysis was Shao's bootstrapped logistic regression yielding area under the curve (AUC) and odds ratios (OR). Predictors of homicide were poorer executive function (OR = 7.04e+40), violent family (OR = 4.01e-16), and alcohol abuse (OR = 7.33e-17; AUC = .97, 95% CI = .77-.99). From earlier studies, 101 Homicidal Youth and their Controls were reanalyzed similarly. Predictors were poorer executive function (OR = 6.51), lower social maturity (OR = 0.28), weapon possession (OR = 26.10), and gang membership (OR = 4.14; AUC = .98, 95% CI = .96-.99). Groups were combined, i.e., 26 and 101 Homicidal; 127 Homicidal Youth (7 girls, 120 boys) and their matched Controls were tracked in records. The predictor was poorer executive function (OR = 3.34e-21; AUC = .98, 95% CI = .96-.97). When 127 Homicidal Youth were compared with 127 matched Nonviolent Delinquents, predictors were poorer executive function (OR = 2.83e-02), weapon possession (OR = 1.63e-10), lower social maturity (OR = 1.15), and use of special education services (OR = .94; AUC = .94, 95% CI = .37-.99).

Looking forward in records of young adults convicted of sexual homicide, rape, or molestation as youth: risks for reoffending.

To assess the risks predicting reoffense, 223 Rapists (M age = 14.2 yr., SD = 1.5; 25 girls, 198 boys) were matched with 223 Nonviolent Delinquents; risks were analyzed using logistic regression. The one predictor was prior court contacts (OR = 1.55e+12; AUC = .99, 95% CI = .98-.99). 223 Molesters were similarly matched with 223 Nonviolent Delinquents; this comparison yielded three predictors: previous court contacts (OR = 4.55e+23), poorer executive function (OR = 2.01), and lower social maturity (OR = .97; AUC = .98, 95% CI = .97-.99). Records for all cases (now M age = 24.2 yr., SD = 1.4) were reviewed forward 10 years and youth were classified into groups: Sexual Homicidal (1%, n = 7), Delinquent Rapists Later Adult Rapists (11%, n = 73), Delinquent Rapists (21%, n = 144), Delinquent Molesters Later Adult Molesters (10%, n = 69), Delinquent Molesters (23%, n = 153), Nonviolent Delinquent Later Nonviolent Adult Criminals (7%, n = 45), and Nonviolent Delinquents (27%, n = 178). Comparison of Sexual Homicidal cases (n = 7) with their matched Controls (n = 7) yielded one predictor, poorer executive function (AUC = .89, 95% CI = .71-.93). When Sexual Homicidal cases were matched with Nonviolent Delinquents, predictors were low social maturity and prior court contacts (AUC = .81, 95% CI = .64-.93).

Can violent (re)offense be predicted? Review of the role of the clinician and use of actuarial tests in light of new data.

This article is an evaluation of established actuarial probation-parole tests in light of new data on violent and homicidal behavior. Probation-parole tests originally were developed by observing risks related with recidivism or return to court after release (i.e., the "danger" of releasing an individual) by following offenders in court records for up to 10 years. Commonly used probation-parole tests together comprise 82 distinct items related to characteristics of the offender: home, school, peers, job, family, individual-medical, community, and court contacts. The risks for violence and homicide found by Zagar, et al. were compared with prior meta-analyses of risks with the criterion of violent delinquency. Bootstrapped logistic regressions in Zagar and colleagues' new data yielded highly accurate predictions of violence, showing that improved methods and sampling can lead to still higher accuracy than had been achieved by established probation-parole tests. A general discussion of the usefulness of actuarial tests and answers to challenges of their validity for decision making are provided.

An exploration of the base rate scores of the Millon Clinical Multiaxial Inventory-III.

The Millon Clinical Multiaxial Inventory (3rd ed.; MCMI-III) is a widely used psychological assessment of clinical and personality disorders. Unlike typical tests, the MCMI-III uses a base-rate score transformation to incorporate prior probabilities of disorder (i.e., base rates) in test output and diagnostic thresholds. The authors describe the base rate transformation and contend that its supporting documentation in the MCMI-III manual is incomplete and fails to meet interdisciplinary test documentation standards. They show that the MCMI-III's base rate transformation is not optimal, and they derive an optimal alternative transformation using Bayes' theorem. Bayes transformation formulae for 7 exponential family distributions are given. The authors discuss the effect of the base rate transformation and further argue that the MCMI-III's use of a single base rate per diagnostic category is ill-advised. They argue that base rates differ among clinical settings and that tests like the MCMI-III should flexibly incorporate a base rate of disorder pertinent to the examinee's characteristics, such as demographics, chief complaint, clinical history, or other variables. They explain how this can readily be accomplished. (PsycINFO Database Record (c) 2009 APA, all rights reserved).

Development and initial validation of a statistical prediction instrument for assessing combat-related posttraumatic stress disorder.

Statistical analyses were used to derive and validate a statistical prediction instrument to determine combat-related posttraumatic stress disorder (PTSD) status. Participants were 1081 Vietnam veterans with and without combat-related PTSD. The statistical prediction instrument, which consisted of 12 well-known risk and resilience variables associated with PTSD, proved to be an accurate and efficient means of detecting PTSD among participants and compared well against other existing self-report measures of PTSD. The instrument's practical applications and its use in clinical appraisals of PTSD are discussed.

The taxon concept: is it taxonic?

The question of whether the concept of a "taxon" (a nonarbitrary latent category) is itself categorical, or is a matter of degree, has lain dormant within taxometrics. I analyze the problem conceptually. Part of the meaning of "taxon," I hold, goes beyond the manifest statistical properties of admixed probability distributions; any of certain forms of causal nexus leading to admixed distributions are involved as well. I defend the thesis that the threshold question, "Is there a taxon?" has a yes-or-no answer, establishing that "taxonicity" is categorical (taxonic), but also that as some taxa are much more readily distinguished from their complement classes than are others, the taxon question is also quantitative. It is on this basis that Meehl and others held the view that taxonicity is a matter of degree; however, they were wrong to hold that it is only a matter of degree.

Predicting sex offender recidivism. I. Correcting for item overselection and accuracy overestimation in scale development. II. Sampling error-induced attenuation of predictive validity over base rate information.

The authors demonstrate a statistical bootstrapping method for obtaining unbiased item selection and predictive validity estimates from a scale development sample, using data (N = 256) of Epperson et al. [2003 Minnesota Sex Offender Screening Tool-Revised (MnSOST-R) technical paper: Development, validation, and recommended risk level cut scores. Retrieved November 18, 2006 from Iowa State University Department of Psychology web site: http://www.psychology.iastate.edu/ approximately dle/mnsost_download.htm] from which the Minnesota Sex Offender Screening Tool-Revised (MnSOST-R) was developed. Validity (area under receiver operating characteristic curve) reported by Epperson et al. was .77 with 16 items selected. The present analysis yielded an asymptotically unbiased estimator AUC = .58. The present article also focused on the degree to which sampling error renders estimated cutting scores (appropriate to local [varying] recidivism base rates) nonoptimal, so that the long-run performance (measured by correct fraction, the total proportion of correct classifications) of these estimated cutting scores is poor, when they are applied to their parent populations (having assumed values for AUC and recidivism rate). This was investigated by Monte Carlo simulation over a range of AUC and recidivism rate values. Results indicate that, except for the AUC values higher than have ever been cross-validated, in combination with recidivism base rates severalfold higher than the literature average [Hanson and Morton-Bourgon, 2004, Predictors of sexual recidivism: An updated meta-analysis. (User report 2004-02.). Ottawa: Public Safety and Emergency Preparedness Canada], the user of an instrument similar in performance to the MnSOST-R cannot expect to achieve correct fraction performance notably in excess of what is achievable from knowing the population recidivism rate alone. The authors discuss the legal implications of their findings for procedural and substantive due process in relation to state sexually violent person commitment statutes and the Supreme Court's Kansas v. Hendricks decision regarding the constitutionality of such statutes.

Dissociating the long-term effects of fetal/neonatal iron deficiency on three types of learning in the rat.

Iron deficiency (ID) is a common nutrient deficiency worldwide. This condition is linked to changes in myelin formation, dopaminergic function, and energy metabolism. Early ID results in persistent long-term cognitive and behavioral disturbances in children, despite a return to normal iron status. The present study assesses formerly ID adult rats on maze learning tasks that depend on specific brain regions related to learning, specifically the hippocampus, striatum, and amygdala. Rat dams were fed ID chow starting on gestational Day 2 through postnatal Day 7, and behavioral testing began at postnatal Day 65--following a return to normal iron status. Formerly ID rats exhibited delayed acquisition of the hippocampus-dependant task and no differences from controls on the striatum- and amygdala-dependent tasks. These findings likely reflect long-term reduction in but not abolition of hippocampus-dependent learning and preserved function in other brain structures (e.g., striatum and amygdala).

Dimensional analyses of taxonic data.

Principal component analysis (PCA) and common factor analysis are often used to model latent data structures. Typically, such analyses assume a single population whose correlation or covariance matrix is modelled. However, data may sometimes be unwittingly sampled from mixed populations containing a taxon (nonarbitrary subpopulation) and its complement class. One derives relations between values of PCA parameters within subpopulations and their values in the mixed population. These results are then extended to factor analysis in mixed populations. As relationships between subpopulation and mixed-population principal components and factors sensitively depend on within-subpopulation structures and between-subpopulation differences, naive interpretation of PCA or factor analytic findings can potentially mislead. Several analyses, better suited to the dimensional analysis of admixture data structures, are presented and compared.

Meehl's contribution to clinical versus statistical prediction.

Paul E. Meehl's work on the clinical versus statistical prediction controversy is reviewed. His contributions included the following: putting the controversy center stage in applied psychology; clarifying concepts underpinning the debate (especially his crucial distinction between ways of gathering data and ways of combining them) as well as establishing that the controversy was real and not concocted, analyzing clinical inference from both theoretical and probabilistic points of view, and reviewing studies that compared the accuracy of these 2 methods of data combination. Meehl's (1954/1996) conclusion that statistical prediction consistently outperforms clinical judgment has stood up extremely well for half a century. His conceptual analyses have not been significantly improved since he published them in the 1950s and 1960s. His work in this area contains several citation classics, which are part of the working knowledge of all competent applied psychologists today.