Bayesian hierarchical vector autoregressive models for patient-level predictive modeling.

Predicting health outcomes from longitudinal health histories is of central importance to healthcare. Observational healthcare databases such as patient diary databases provide a rich resource for patient-level predictive modeling. In this paper, we propose a Bayesian hierarchical vector autoregressive (VAR) model to predict medical and psychological conditions using multivariate time series data. Compared to the existing patient-specific predictive VAR models, our model demonstrated higher accuracy in predicting future observations in terms of both point and interval estimates due to the pooling effect of the hierarchical model specification. In addition, by adopting an elastic-net prior, our model offers greater interpretability about the associations between variables of interest on both the population level and the patient level, as well as between-patient heterogeneity. We apply the model to two examples: 1) predicting substance use craving, negative affect and tobacco use among college students, and 2) predicting functional somatic symptoms and psychological discomforts.

Auditory tasks for assessment of sensory function and affective prosody in schizophrenia.

Schizophrenia patients exhibit impairments in auditory-based social cognition, indicated by deficits in detection of prosody, such as affective prosody and basic pitch perception. However, little is known about the psychometric properties of behavioral tests used to assess these functions. The goal of this paper is to characterize the properties of prosody and pitch perception tasks and to investigate whether they can be shortened. The pitch perception test evaluated is a tone-matching task developed by Javitt and colleagues (J-TMT). The prosody test evaluated is the auditory emotion recognition task developed by Juslin and Laukka (JL-AER). The sample includes 124 schizophrenia patients (SZ) and 131 healthy controls (HC). Properties, including facility and discrimination, of each item were assessed. Effects of item characteristics (e.g., emotion) were also evaluated. Shortened versions of the tests are proposed based on facility, discrimination, and/or ability of item characteristics to discriminate between patients and controls. Test-retest reliability is high for patients and controls for both the original and short forms of the J-TMT and JL-AER. Thus, the original as well as short forms of the J-TMT and JL-AER are suggested for inclusion in clinical trials of social cognitive and perceptual treatments. The development of short forms further increases the utility of these auditory tasks in clinical trials and clinical practice. The large SZ vs. HC differences reported here also highlight the profound nature of auditory deficits and a need for remediation.

A comparative study of variable selection methods in the context of developing psychiatric screening instruments.

The development of screening instruments for psychiatric disorders involves item selection from a pool of items in existing questionnaires assessing clinical and behavioral phenotypes. A screening instrument should consist of only a few items and have good accuracy in classifying cases and non-cases. Variable/item selection methods such as Least Absolute Shrinkage and Selection Operator (LASSO), Elastic Net, Classification and Regression Tree, Random Forest, and the two-sample t-test can be used in such context. Unlike situations where variable selection methods are most commonly applied (e.g., ultra high-dimensional genetic or imaging data), psychiatric data usually have lower dimensions and are characterized by the following factors: correlations and possible interactions among predictors, unobservability of important variables (i.e., true variables not measured by available questionnaires), amount and pattern of missing values in the predictors, and prevalence of cases in the training data. We investigate how these factors affect the performance of several variable selection methods and compare them with respect to selection performance and prediction error rate via simulations. Our results demonstrated that: (1) for complete data, LASSO and Elastic Net outperformed other methods with respect to variable selection and future data prediction, and (2) for certain types of incomplete data, Random Forest induced bias in imputation, leading to incorrect ranking of variable importance. We propose the Imputed-LASSO combining Random Forest imputation and LASSO; this approach offsets the bias in Random Forest and offers a simple yet efficient item selection approach for missing data. As an illustration, we apply the methods to items from the standard Autism Diagnostic Interview-Revised version.

Designing a new proof-of-principle trial for treatment of partial seizures to demonstrate efficacy with minimal sample size and duration-a case study.

The ideal proof-of-principle study design provides a strong efficacy signal over the shortest duration, while exposing the fewest patients possible. Data from a large database (Pfizer Inc) which studied add-on pregabalin for the treatment of partial seizures was used to model how duration of baseline, post-randomization treatment period, and number of subjects impact the likelihood of an interpretable efficacy signal. Data from four double-blind, randomized, placebo-controlled, phase III studies that had at least one 600 mg/day treatment arm were combined. The common 6-week baseline period was divided into weekly intervals, as was the 12-week post-randomization period. Two methods of analysis were used: logistic regression performed on 50% responder rate and the Hodges-Lehmann estimate on percentage reduction from baseline seizure rate. A simulation-based re-sampling approach was used to determine sufficient sample size. Four weeks of baseline with 3 weeks of treatment were determined to be clinically and statistically sufficient. A reasonable sample size was estimated to be 40-50 patients per group, if a highly efficacious drug was used. These modeling results indicate that the efficacy of an antiepileptic drug can be demonstrated in a relatively short period of time with a reasonable sample size.

Remediating organizational functioning in children with ADHD: immediate and long-term effects from a randomized controlled trial.

OBJECTIVE: The study compared the efficacy of 2 behavioral interventions to ameliorate organization, time management, and planning (OTMP) difficulties in 3rd- to 5th-grade children with attention-deficit/hyperactivity disorder (ADHD). METHOD: In a dual-site randomized controlled trial, 158 children were assigned to organizational skills training (OST; N = 64); PATHKO, a performance-based intervention that precluded skills training (N = 61); or a wait-list control (WL, N = 33). Treatments were 20 individual clinic-based sessions over 10-12 weeks. OST involved skills building provided primarily to the child. PATHKO trained parents and teachers to reinforce children contingently for meeting end-point target goals. Primary outcomes were the Children's Organizational Skills Scales (COSS-Parent, COSS-Teacher). Other relevant functional outcomes were assessed. Percentage of participants no longer meeting inclusion criteria for OTMP impairments informed on clinical significance. Assessments occurred at post-treatment, 1-month post-treatment, and twice in the following school year. RESULTS: OST was superior to WL on the COSS-P (Cohen's d = 2.77; p < .0001), COSS-T (d = 1.18; p < .0001), children's COSS self-ratings, academic performance and proficiency, homework, and family functioning. OST was significantly better than PATHKO only on the COSS-P (d = 0.63; p < .005). PATHKO was superior to WL on most outcomes but not on academic proficiency. Sixty percent of OST and PATHKO participants versus 3% of controls no longer met OTMP inclusion criteria. Significant maintenance effects were found for both treatments. CONCLUSIONS: Two distinct treatments targeting OTMP problems in children with ADHD generated robust, sustained functional improvements. The interventions show promise of clinical utility in children with ADHD and organizational deficits.

Paradoxical results of adaptive false discovery rate procedures in neuroimaging studies.

Adaptive false discovery rate (FDR) procedures, which offer greater power than the original FDR procedure of Benjamini and Hochberg, are often applied to statistical maps of the brain. When a large proportion of the null hypotheses are false, as in the case of widespread effects such as cortical thinning throughout much of the brain, adaptive FDR methods can surprisingly reject more null hypotheses than not accounting for multiple testing at all-i.e., using uncorrected p-values. A straightforward mathematical argument is presented to explain why this can occur with the q-value method of Storey and colleagues, and a simulation study shows that it can also occur, to a lesser extent, with a two-stage FDR procedure due to Benjamini and colleagues. We demonstrate the phenomenon with reference to a published data set documenting cortical thinning in attention deficit/hyperactivity disorder. The paper concludes with recommendations for how to proceed when adaptive FDR results of this kind are encountered in practice.

Self-Disclosure and Mental Health Service Use in Socially Anxious Adolescents.

Social anxiety disorder is highly prevalent in adolescence, persistent into adulthood, and associated with multiple impairments. Despite the development of efficacious treatments for socially anxious youth, few affected adolescents receive such treatment. This study examined service use in a sample of high school students (n = 1,574), as well as predictors of treatment delay and factors associated with adolescents' disclosure of social difficulties. Self-report measures of social anxiety and service utilization were administered by study staff to 10th- and 11th-grade classrooms across three public high schools. Consistent with the literature, results indicated low treatment utilization (14 %) and lengthy delays in treatment initiation. Symptom severity, impairment, and disclosing anxiety to school personnel were significant predictors of service utilization. Several demographic and illness-specific factors were associated with a higher likelihood of disclosing social discomfort. These findings underscore the important role of school personnel in identifying and referring youth with anxiety disorders. Implications are discussed for increasing access to services, including school-wide screenings and training of school personnel to recognize and provide intervention for anxious youth.

A multisite study of the clinical diagnosis of different autism spectrum disorders.

CONTEXT: Best-estimate clinical diagnoses of specific autism spectrum disorders (autistic disorder, pervasive developmental disorder-not otherwise specified, and Asperger syndrome) have been used as the diagnostic gold standard, even when information from standardized instruments is available. OBJECTIVE: To determine whether the relationships between behavioral phenotypes and clinical diagnoses of different autism spectrum disorders vary across 12 university-based sites. DESIGN: Multisite observational study collecting clinical phenotype data (diagnostic, developmental, and demographic) for genetic research. Classification trees were used to identify characteristics that predicted diagnosis across and within sites. SETTING: Participants were recruited through 12 university-based autism service providers into a genetic study of autism. PARTICIPANTS: A total of 2102 probands (1814 male probands) between 4 and 18 years of age (mean [SD] age, 8.93 [3.5] years) who met autism spectrum criteria on the Autism Diagnostic Interview-Revised and the Autism Diagnostic Observation Schedule and who had a clinical diagnosis of an autism spectrum disorder. MAIN OUTCOME MEASURE: Best-estimate clinical diagnoses predicted by standardized scores from diagnostic, cognitive, and behavioral measures. RESULTS: Although distributions of scores on standardized measures were similar across sites, significant site differences emerged in best-estimate clinical diagnoses of specific autism spectrum disorders. Relationships between clinical diagnoses and standardized scores, particularly verbal IQ, language level, and core diagnostic features, varied across sites in weighting of information and cutoffs. CONCLUSIONS: Clinical distinctions among categorical diagnostic subtypes of autism spectrum disorders were not reliable even across sites with well-documented fidelity using standardized diagnostic instruments. Results support the move from existing subgroupings of autism spectrum disorders to dimensional descriptions of core features of social affect and fixated, repetitive behaviors, together with characteristics such as language level and cognitive function.