Individual Mobility across Clusters: The Impact of Ignoring Cross-Classified Data Structures in Discrete-Time Survival Analysis.

A multilevel-discrete time survival model may be appropriate for purely hierarchical data, but when data are non-purely hierarchical due to individual mobility across clusters, a cross-classified discrete time survival model may be necessary. The purpose of this research was to investigate the performance of a cross-classified discrete-time survival model and assess the impact of ignoring a cross-classified data structure on the model parameters of a conventional discrete-time survival model and a multilevel discrete-time survival model. A Monte Carlo simulation was used to examine the performance of three discrete-time survival models when individuals are mobile across clusters. Simulation factors included the value of the between-clusters variance, number of clusters, within-cluster sample size, Weibull scale parameter, and mobility rate. The results suggest that substantial relative parameter bias, unacceptable coverage of the 95% confidence intervals, and severely biased standard errors are possible for all model parameters when a discrete-time survival model is used that ignores the cross-classified data structure. The findings presented in this study are useful for methodologists and practitioners in educational research, public health, and other social sciences where discrete-time survival analysis is a common methodological technique for analyzing event-history data.

Psychosis Symptom Trajectories Across Childhood and Adolescence in Three Longitudinal Studies: An Integrative Data Analysis with Mixture Modeling.

Psychotic-like experiences (PLEs) are common throughout childhood, and the presence of these experiences is a significant risk factor for poor mental health later in development. Given the association of PLEs with a broad number of mental health diagnoses, these experiences serve as an important malleable target for early preventive interventions. However, little is known about these experiences across childhood. While these experiences may be common, longitudinal measurement in non-clinical settings is not. Therefore, in order to explore longitudinal trajectories of PLEs in childhood, we harmonized three school-based randomized control trials with longitudinal follow-up to identify heterogeneity in trajectories of these experiences. In an integrative data analysis (IDA) using growth mixture modeling, we identified three latent trajectory classes. One trajectory class was characterized by persistent PLEs, one was characterized by high initial probabilities but improving across the analytic period, and one was characterized by no reports of PLEs. Compared to the class without PLEs, those in the improving class were more likely to be male and have higher levels of aggressive and disruptive behavior at baseline. In addition to the substantive impact this work has on PLE research, we also discuss the methodological innovation as it relates to IDA. This IDA demonstrates the complexity of pooling data across multiple studies to estimate longitudinal mixture models.

Dyadic effects of minority stress and problematic alcohol use on sexual intimate partner violence in same sex couples.

Little is known about the factors that facilitate the perpetration of sexual violence within the context of same-sex romantic relationships (sexual intimate partner violence perpetration [S-IPV]). The present study sought to identify the effects of external and internal minority stress and problematic drinking on perpetration of S-IPV within a dyadic framework. A community-based sample of 137 sexual and gender minority (SGM) couples (N = 274; 59 male assigned at birth and 78 female assigned at birth couples) completed self-report surveys about minority stressors, alcohol use, and S-IPV perpetration. Multilevel logistic regression analyses were conducted within an actor-partner interdependence framework. This approach accounted for both actor effects (e.g., how much one's S-IPV perpetration is predicted by their own risk factors) and partner effects (e.g., how much one's S-IPV perpetration is influenced by their partner's risk factors). Both Actor external minority stress and internal minority stress were positively associated with Actor S-IPV perpetration. Actor problematic drinking was not associated with Actor S-IPV perpetration; however, Partner problematic drinking was positively associated with Actor S-IPV perpetration. Observed effects were robust above the addition of other risk factors. This research innovatively extricates S-IPV perpetration from other forms of IPV and indicates that Actor minority stress and Partner problematic drinking increase S-IPV likelihood. Results serve as a starting point for development of etiological models to inform the design of culturally-informed interventions to reduce S-IPV among SGM couples.

Latent classes of aggression and peer victimization: Measurement invariance and differential item functioning across sex, race-ethnicity, cohort, and study site.

Peer victimization is common and linked to maladjustment. Prior research has typically identified four peer victimization subgroups: aggressors, victims, aggressive-victims, and uninvolved. However, findings related to sex and racial-ethnic differences in subgroup membership have been mixed. Using data collected in September of 2002 and 2003, this study conducted confirmatory latent class analysis of a racially-ethnically diverse sample of 5415 sixth graders (49% boys; 50.6% Black; 20.9% Hispanic) representing two cohorts from 37 schools in four U.S. communities to replicate the four subgroups and evaluate measurement invariance of latent class indicators across cohort, sex, race-ethnicity, and study site. Results replicated the four-class solution and illustrated that sociodemographic differences in subgroup membership were less evident after accounting for differential item functioning.

The Intersection of Alcohol Use, Gender Based Violence and HIV: Empirical Findings among Disadvantaged Service-Seeking Youth in Kampala, Uganda.

The SAVA syndemic is frequently used to describe the co-occurrence of HIV, gender-based violence (GBV), and substance use. In this study we determine the extent to which the typologies of the SAVA syndemic can be described and utilized for intervention strategies among youth living in the slums of Kampala, Uganda. We analyzed the "Kampala Youth Survey 2014," a cross-sectional survey conducted in the spring of 2014, consisting of a convenience sample (N = 1134) of urban youth (12-18 years of age). Descriptive statistics were computed for hypothesized risk factors and demographic variables among the 8 typologies of GBV, HIV, and alcohol use. Multinomial logistic regression was conducted to determine statistically significant correlates with each typology. The overall prevalence of GBV was 31.7%, whereas the overall prevalence of alcohol use in the past 12 months was 31.2%. HIV-Positive youth comprised 10.5% of the total sample. Females comprised the majority of the typology with no SAVA components compared to males (55% vs. 45%, respectively), as well as the SAVA syndemic typology (GBV + HIV + ALC +) (58% vs. 42%, respectively). Engaging in commercial sex work (36%), witnessing parental abuse (61%), and depression/suicidality (81%) were all highly prevalent among youth in the SAVA syndemic typology (GBV, HIV, and alcohol use). Sex work and observing parental abuse were associated with the SAVA syndemic typology in the multivariable model. In our study, alcohol rarely coexisted without GBV among the typologies. Therefore, prevention efforts including structural interventions may be particularly warranted in vulnerable populations to address alcohol use, which may directly or indirectly impact GBV and HIV.

Individual Differences in the Stability and Change of Childhood Depression: A Growth Mixture Model With Structured Residuals.

Studies of developmental trajectories of depression are important for understanding depression etiology. Existing studies have been limited by short time frames and no studies have explored a key factor: differential patterns of responding to life events. This article introduces a novel analytic technique, growth mixture modeling with structured residuals, to examine the course of youth depression in a large, prospective cohort (N = 11,641, ages 4-16.5, 96% White). Age-specific critical points were identified at ages 8 and 13 where depression symptoms spiked for a minority of children. Most depression risk was due to dynamic responses to environmental events, drawn not from a small pool of persistently depressed children, but a larger pool of children who varied across higher and lower symptom levels.

A Mixed-Methods Longitudinal Investigation of Mothers' Disclosure of HIV to Their Children.

Mothers living with HIV (MLH) must navigate disclosing their serostatus to their children, but the longitudinal impact on families remains unknown. This study examined HIV disclosure, parenting, parenting stress, and child adjustment among 174 MLH-child dyads (aged 6-14; 35% Latinx; 57% Black/African American). Quantitative data were collected over four waves spanning 15 months. Qualitative data were collected with 14 families in which disclosure had occurred. Latent change score modeling revealed that disclosure led to improvements in parenting stress, communication, and relationship quality. Disclosure did not predict child adjustment. Qualitative themes contextualized these findings, revealing stability and improvements in family functioning. MLH should be supported in disclosing their serostatus to their children to minimize parenting stress and bolster parenting skills.

Child maltreatment, adaptive functioning, and polygenic risk: A structural equation mixture model.

This study used a structural equation mixture model to examine associations between child maltreatment, polygenic risk, and indices of adaptive functioning. Children aged 6 to 13 years (N = 1,004), half maltreated, half nonmaltreated, were recruited to attend a research day camp. Multi-informant indicators of prosocial behavior, antisocial behavior, withdrawn behavior, and depression were collected and used in a latent class analysis. Four classes emerged, characterizing "well-adjusted," "externalizing," "internalizing," and "socially dominant" groups. Twelve genetic variants, previously reported in the Gene × Environment literature, were modeled as one weighted polygenic risk score. Large main effects between maltreatment and adaptive functioning were observed (Wald = 35.3, df = 3, p < .0001), along with evidence of a small Gene × Environment effect (Wald = 13.5, df = 3, p = .004), adjusting for sex, age, and covariate interaction effects.

How do collective student behavior and other classroom contextual factors relate to teachers' implementation of an evidence-based intervention? A multilevel structural equation model.

Building on prior work regarding the potential for peer contagion or deviance training in group delivered interventions (Dishion & Dodge, 2005, 2006; Dodge, Dishion, & Lansford, 2006), we leveraged data from a randomized trial, testing the integration of two preventive interventions (Promoting Alternative THinking Strategies and PAX Good Behavior Game), to explore the extent to which classroom contextual factors served as either a barrier to or a motivator for teachers to implement the evidence-based PAX Good Behavior Game with high frequency or dosage. We included students' baseline levels of behavior, measured with regard to both positive (i.e., engagement and social emotional skills) and negative (i.e., hyperactive and aggressive-disruptive) behaviors. Data were collected from 204 teachers in 18 urban elementary schools. A series of multilevel structural equation models were fit to the data. The analyses indicated that classrooms with higher classroom levels of aggressive behavior, on average, at baseline had teachers with lower implementation dosage (i.e., played fewer games) across the school year. In addition, teachers who reported higher baseline levels of emotional exhaustion, regardless of student behavior, also reported lower implementation dosage. Taken together, the results indicated that negative, but not positive, contextual factors at baseline were related to lower implementation dosage; this, in turn, suggests that negative contextual factors may serve as a barrier, rather than a motivator, of teachers' implementation dosage of classroom-based preventive interventions.

Polygenic Score × Intervention Moderation: an Application of Discrete-Time Survival Analysis to Model the Timing of First Marijuana Use Among Urban Youth.

The present study examines the interaction between a polygenic score and an elementary school-based universal preventive intervention trial and its effects on a discrete-time survival analysis of time to first smoking marijuana. Research has suggested that initiation of substances is both genetically and environmentally driven (Rhee et al., Archives of general psychiatry 60:1256-1264, 2003; Verweij et al., Addiction 105:417-430, 2010). A previous work has found a significant interaction between the polygenic score and the same elementary school-based intervention with tobacco smoking (Musci et al., in press). The polygenic score reflects the contribution of multiple genes and has been shown in prior research to be predictive of smoking cessation, tobacco use, and marijuana use (Uhl et al., Molecular Psychiatry 19:50-54, 2014). Using data from a longitudinal preventive intervention study (N = 678), we examined age of first marijuana use from sixth grade to age 18. Genetic data were collected during emerging adulthood and were genotyped using the Affymetrix 6.0 microarray (N = 545). The polygenic score was computed using these data. Discrete-time survival analysis was employed to test for intervention main and interaction effects with the polygenic score. We found main effect of the polygenic score approaching significance, with the participants with higher polygenic scores reporting their first smoking marijuana at an age significantly later than controls (p = .050). We also found a significant intervention × polygenic score interaction effect at p = .003, with participants at the higher end of the polygenic score benefiting the most from the intervention in terms of delayed age of first use. These results suggest that genetics may play an important role in the age of first use of marijuana and that differences in genetics may account for the differential effectiveness of classroom-based interventions in delaying substance use experimentation.

Developmental Latent Patterns of Identification as Mostly Heterosexual Versus Lesbian, Gay, or Bisexual.

Scant research exists on the development of mostly heterosexual identity, the largest sexual orientation minority subgroup. We used longitudinal latent class analysis to characterize the patterns of identification with lesbian, gay, bisexual (LGB), or mostly heterosexual identities from ages 12 to 23 in 13,859 youth (57% female) in a U.S. national cohort. Three classes emerged: completely heterosexual (88.2%), mostly heterosexual (9.5%), and LGB (2.4%). LGB class youth generally identified with sexual minority identities by ages 12-17. In contrast, mostly heterosexual class youth identified with sexual minority identities gradually, with steady increases in endorsement starting at the age of 14. Developmental implications of these differential patterns are discussed.

The effects of the interplay of genetics and early environmental risk on the course of internalizing symptoms from late childhood through adolescence.

Internalizing symptoms during adolescence and beyond is a major public health concern, particularly because severe symptoms can lead to the diagnosis of a number of serious psychiatric conditions. This study utilizes a unique sample with a complex statistical method in order to explore Gene × Environment interactions found in internalizing symptoms during adolescence. Data for this study were drawn from a longitudinal prevention intervention study (n = 798) of Baltimore city school children. Internalizing symptom data were collected using self-report and blood or saliva samples genotyped using Affymetrix 6.0 microarrays. A major depression polygenic score was created for each individual using information from the major depressive disorder Psychiatric Genetics Consortium and used as a predictor in a latent trait-state-occasion model. The major depressive disorder polygenic score was a significant predictor of the stable latent trait variable, which captures time-independent phenotypic variability. In addition, an early childhood stressor of death or divorce was a significant predictor of occasion-specific variables. A Gene × Environment interaction was not a significant predictor of the latent trait or occasion variables. These findings support the importance of genetics on the stable latent trait portion of internalizing symptoms across adolescence.

The Effects of Visual Attention Span and Phonological Decoding in Reading Comprehension in Dyslexia: A Path Analysis.

Increasing evidence has shown visual attention span to be a factor, distinct from phonological skills, that explains single-word identification (pseudo-word/word reading) performance in dyslexia. Yet, little is known about how well visual attention span explains text comprehension. Observing reading comprehension in a sample of 105 high school students with dyslexia, we used a pathway analysis to examine the direct and indirect path between visual attention span and reading comprehension while controlling for other factors such as phonological awareness, letter identification, short-term memory, IQ and age. Integrating phonemic decoding efficiency skills in the analytic model, this study aimed to disentangle how visual attention span and phonological skills work together in reading comprehension for readers with dyslexia. We found visual attention span to have a significant direct effect on more difficult reading comprehension but not on an easier level. It also had a significant direct effect on pseudo-word identification but not on word identification. In addition, we found that visual attention span indirectly explains reading comprehension through pseudo-word reading and word reading skills. This study supports the hypothesis that at least part of the dyslexic profile can be explained by visual attention abilities. Copyright © 2016 John Wiley & Sons, Ltd.

Modeling contextual effects using individual-level data and without aggregation: an illustration of multilevel factor analysis (MLFA) with collective efficacy.

Population health scientists increasingly study how contextual-level attributes affect individual health. A major challenge in this domain relates to measurement, i.e., how best to measure and create variables that capture characteristics of individuals and their embedded contexts. This paper presents an illustration of multilevel factor analysis (MLFA), an analytic method that enables researchers to model contextual effects using individual-level data without using derived variables. MLFA uses the shared variance in sets of observed items among individuals within the same context to estimate a measurement model for latent constructs; it does this by decomposing the total sample variance-covariance matrix into within-group (e.g., individual-level) and between-group (e.g., contextual-level) matrices and simultaneously modeling distinct latent factor structures at each level. We illustrate the MLFA method using items capturing collective efficacy, which were self-reported by 2,599 adults in 65 census tracts from the Los Angeles Family and Neighborhood Survey (LAFANS). MLFA identified two latent factors at the individual level and one factor at the neighborhood level. Indicators of collective efficacy performed differently at each level. The ability of MLFA to identify different latent factor structures at each level underscores the utility of this analytic tool to model and identify attributes of contexts relevant to health.

Polygenic score × intervention moderation: an application of discrete-time survival analysis to modeling the timing of first tobacco use among urban youth.

The present study examines the interaction between a polygenic score and an elementary school-based universal preventive intervention trial. The polygenic score reflects the contribution of multiple genes and has been shown in prior research to be predictive of smoking cessation and tobacco use (Uhl et al., 2014). Using data from a longitudinal preventive intervention study, we examined age of first tobacco use from sixth grade to age 18. Genetic data were collected during emerging adulthood and were genotyped using the Affymetrix 6.0 microarray. The polygenic score was computed using these data. Discrete-time survival analysis was employed to test for intervention main and interaction effects with the polygenic score. We found a main effect of the intervention, with the intervention participants reporting their first cigarette smoked at an age significantly later than controls. We also found an Intervention × Polygenic Score interaction, with participants at the higher end of the polygenic score benefitting the most from the intervention in terms of delayed age of first use. These results are consistent with Belsky and colleagues' (e.g., Belsky, Bakermans-Kranenburg, & van IJzendoorn, 2007; Belsky & Pleuss, 2009, 2013; Ellis, Boyce, Belsky, Bakermans-Kranenburg, & van IJzendoorn, 2011) differential susceptibility hypothesis and the concept of "for better or worse," wherein the expression of genetic variants are optimally realized in the context of an enriched environment, such as provided by a preventive intervention.

Measuring psychosocial environments using individual responses: an application of multilevel factor analysis to examining students in schools.

Interest in understanding how psychosocial environments shape youth outcomes has grown considerably. School environments are of particular interest to prevention scientists as many prevention interventions are school-based. Therefore, effective conceptualization and operationalization of the school environment is critical. This paper presents an illustration of an emerging analytic method called multilevel factor analysis (MLFA) that provides an alternative strategy to conceptualize, measure, and model environments. MLFA decomposes the total sample variance-covariance matrix for variables measured at the individual level into within-cluster (e.g., student level) and between-cluster (e.g., school level) matrices and simultaneously models potentially distinct latent factor structures at each level. Using data from 79,362 students from 126 schools in the National Longitudinal Study of Adolescent to Adult Health (formerly known as the National Longitudinal Study of Adolescent Health), we use MLFA to show how 20 items capturing student self-reported behaviors and emotions provide information about both students (within level) and their school environment (between level). We identified four latent factors at the within level: (1) school adjustment, (2) externalizing problems, (3) internalizing problems, and (4) self-esteem. Three factors were identified at the between level: (1) collective school adjustment, (2) psychosocial environment, and (3) collective self-esteem. The finding of different and substantively distinct latent factor structures at each level emphasizes the need for prevention theory and practice to separately consider and measure constructs at each level of analysis. The MLFA method can be applied to other nested relationships, such as youth in neighborhoods, and extended to a multilevel structural equation model to better understand associations between environments and individual outcomes and therefore how to best implement preventive interventions.

Translating multilevel theory into multilevel research: challenges and opportunities for understanding the social determinants of psychiatric disorders.

INTRODUCTION: The observation that features of the social environment, including family, school, and neighborhood characteristics, are associated with individual-level outcomes has spurred the development of dozens of multilevel or ecological theoretical frameworks in epidemiology, public health, psychology, and sociology, among other disciplines. Despite the widespread use of such theories in etiological, intervention, and policy studies, challenges remain in bridging multilevel theory and empirical research. METHODS: This paper set out to synthesize these challenges and provide specific examples of methodological and analytical strategies researchers are using to gain a more nuanced understanding of the social determinants of psychiatric disorders, with a focus on children's mental health. To accomplish this goal, we begin by describing multilevel theories, defining their core elements, and discussing what these theories suggest is needed in empirical work. In the second part, we outline the main challenges researchers face in translating multilevel theory into research. These challenges are presented for each stage of the research process. In the third section, we describe two methods being used as alternatives to traditional multilevel modeling techniques to better bridge multilevel theory and multilevel research. These are (1) multilevel factor analysis and multilevel structural equation modeling; and (2) dynamic systems approaches. CONCLUSIONS: Through its review of multilevel theory, assessment of existing strategies, and examination of emerging methodologies, this paper offers a framework to evaluate and guide empirical studies on the social determinants of child psychiatric disorders as well as health across the life course.

The intersectionality of discrimination attributes and bullying among youth: an applied latent class analysis.

Discrimination is commonly experienced among adolescents. However, little is known about the intersection of multiple attributes of discrimination and bullying. We used a latent class analysis (LCA) to illustrate the intersections of discrimination attributes and bullying, and to assess the associations of LCA membership to depressive symptoms, deliberate self harm and suicidal ideation among a sample of ethnically diverse adolescents. The data come from the 2006 Boston Youth Survey where students were asked whether they had experienced discrimination based on four attributes: race/ethnicity, immigration status, perceived sexual orientation and weight. They were also asked whether they had been bullied or assaulted for these attributes. A total of 965 (78%) students contributed to the LCA analytic sample (45% Non-Hispanic Black, 29% Hispanic, 58% Female). The LCA revealed that a 4-class solution had adequate relative and absolute fit. The 4-classes were characterized as: low discrimination (51%); racial discrimination (33%); sexual orientation discrimination (7%); racial and weight discrimination with high bullying (intersectional class) (7%). In multivariate models, compared to the low discrimination class, individuals in the sexual orientation discrimination class and the intersectional class had higher odds of engaging in deliberate self-harm. Students in the intersectional class also had higher odds of suicidal ideation. All three discrimination latent classes had significantly higher depressive symptoms compared to the low discrimination class. Multiple attributes of discrimination and bullying co-occur among adolescents. Research should consider the co-occurrence of bullying and discrimination.

HIV co-infection increases the risk of post-tuberculosis mortality among persons who initiated treatment for drug-resistant tuberculosis.

Little is known regarding the relationship between common comorbidities in persons with tuberculosis (TB) (including human immunodeficiency virus [HIV], diabetes, and hepatitis C virus [HCV]) with post-TB mortality. We conducted a retrospective cohort study among persons who initiated treatment for rifampicin-resistant and multi/extensively drug-resistant (RR and M/XDR) TB reported to the country of Georgia's TB surveillance during 2009-2017. Exposures included HIV serologic status, diabetes, and HCV status. Our outcome was all-cause post-TB mortality determined by cross-validating vital status with Georgia's death registry through November 2019. We estimated adjusted hazard rate ratios (aHR) and 95% confidence intervals (CI) of post-TB mortality among participants with and without comorbidities using cause-specific hazard regressions. Among 1032 eligible participants, 34 (3.3%) died during treatment and 87 (8.7%) died post-TB treatment. Among those who died post-TB treatment, the median time to death was 21 months (interquartile range 7-39) post-TB treatment. After adjusting for confounders, the hazard rates of post-TB mortality were higher among participants with HIV co-infection (aHR=3.74, 95%CI 1.77-7.91) compared to those without HIV co-infection. In our cohort, post-TB mortality occurred most commonly in the first three years post-TB treatment. Linkage to care for common TB comorbidities post-treatment may reduce post-TB mortality rates.

HIV co-infection increases the risk of post-tuberculosis mortality among persons who initiated treatment for drug-resistant tuberculosis.

Little is known regarding the relationship between common comorbidities in persons with tuberculosis (TB) (including human immunodeficiency virus [HIV], diabetes, and hepatitis C virus [HCV]) and post-TB mortality. We conducted a retrospective cohort study among persons who initiated treatment for rifampicin-resistant or multi/extensively drug-resistant (RR or M/XDR) TB reported to the country of Georgia's TB surveillance during 2009-2017. Exposures included HIV serologic status, diabetes, and HCV status. Our outcome was all-cause post-TB mortality determined by cross-validating vital status with Georgia's death registry through November 2019. We estimated adjusted hazard rate ratios (aHR) and 95% confidence intervals (CI) of post-TB mortality among participants with and without comorbidities using cause-specific hazard regressions. Among 1032 eligible participants, 34 (3.3%) died during treatment and 87 (8.7%) died post-TB treatment. The median time to post-TB death was 21 months (interquartile range 7-39) after TB treatment. After adjusting for confounders, the hazard rates of post-TB mortality were higher among participants with HIV co-infection (aHR = 3.74, 95%CI 1.77-7.91) compared to those without HIV co-infection. In our cohort, post-TB mortality occurred most commonly in the first 3 years post-TB treatment. Linkage to care for common TB comorbidities post-treatment may reduce post-TB mortality rates.