Targeted Estimation of the Relationship Between Childhood Adversity and Fluid Intelligence in a US Population Sample of Adolescents

Application of targeted maximum likelihood estimation in public health and epidemiological studies: a systematic review

PURPOSE

The targeted maximum likelihood estimation (TMLE) statistical data analysis framework integrates machine learning, statistical theory, and statistical inference to provide a least biased, efficient, and robust strategy for estimation and inference of a variety of statistical and causal parameters. We describe and evaluate the epidemiological applications that have benefited from recent methodological developments.

METHODS

We conducted a systematic literature review in PubMed for articles that applied any form of TMLE in observational studies. We summarized the epidemiological discipline, geographical location, expertize of the authors, and TMLE methods over time. We used the Roadmap of Targeted Learning and Causal Inference to extract key methodological aspects of the publications. We showcase the contributions to the literature of these TMLE results.

RESULTS

Of the 89 publications included, 33% originated from the University of California at Berkeley, where the framework was first developed by Professor Mark van der Laan. By 2022, 59% of the publications originated from outside the United States and explored up to seven different epidemiological disciplines in 2021-2022. Double-robustness, bias reduction, and model misspecification were the main motivations that drew researchers toward the TMLE framework. Through time, a wide variety of methodological, tutorial, and software-specific articles were cited, owing to the constant growth of methodological developments around TMLE.

CONCLUSIONS

There is a clear dissemination trend of the TMLE framework to various epidemiological disciplines and to increasing numbers of geographical areas. The availability of R packages, publication of tutorial papers, and involvement of methodological experts in applied publications have contributed to an exponential increase in the number of studies that understood the benefits and adoption of TMLE.

Intimate Partner Violence and Child and Adolescent Cognitive Development: A Systematic Review

Intimate partner violence (IPV) is a public health and human rights issue, with millions of children affected worldwide. While several reviews have explored the emotional-behavioural functioning of children exposed to IPV, this review aimed to examine the relationship between children's exposure to IPV and their cognitive development, and to identify associated factors such as aspects of parenting. The databases MEDLINE, PsycInfo, EMBASE, Family and Society Studies Worldwide, CINAHL, and ERIC were searched using key words related to IPV, such as domestic, family, partner, interparental, spousal, marital, violence, abuse, aggression, assault, combined with key words related to cognitive functioning, such as neuropsychological, executive, intelligence, learning, memory, and key words related to children and adolescents. A total of 38 studies met the criteria for review which included reporting an estimate of the relationship between IPV and cognition using direct assessments of cognitive functioning. Approximately 70% of studies found a relationship between IPV and poorer cognitive functioning, with general IQ the most frequently assessed domain of functioning, followed by verbal abilities and academic skills. Most studies assessed skills during early childhood, with fewer studies assessing children during middle childhood and adolescence. Results were consistent across cognitive domains and developmental stages. In terms of factors associated with IPV and cognition, a range of demographic, individual, and family factors were included, with several studies exploring mediating and moderating mechanisms. The findings suggest that IPV in childhood is associated with poorer cognitive skills across a range of domains. Implications for policy, practice and research are discussed.

Threat and deprivation are associated with distinct aspects of cognition, emotional processing, and psychopathology in children and adolescents

Exposure to childhood adversity has been consistently associated with poor developmental outcomes, but it is unclear whether these associations vary across different forms of adversity. We examined cross-sectional and longitudinal associations between threat and deprivation with cognition, emotional processing, and psychopathology in a middle-income country. The sample consisted of 2511 children and adolescents (6-17 years old) from the Brazilian High-Risk Cohort for Mental Conditions. Parent reports on childhood adversity were used to construct adversity latent constructs. Psychopathology was measured by the Child Behavior Checklist (CBCL) to generate a measure of general psychopathology (the "p" factor). Executive function (EF) and attention orienting toward angry faces were assessed using cognitive tasks. All measures were acquired at two time-points 3 years apart and associations were tested using general linear models. Higher levels of psychopathology were predicted by higher levels of threat cross-sectionally and longitudinally, and by deprivation longitudinally. For EF, worse performance was associated only with deprivation at baseline and follow-up. Finally, threat was associated with attention orienting towards angry faces cross-sectionally, but neither form of adversity was associated with changes over time in attention bias. Our results suggest that threat and deprivation have differential associations with cognitive development and psychopathology. Exposure to adversity during childhood is a complex phenomenon with meaningful influences on child development. Because adversity can take many forms, dimensional models might help to disentangle the specific developmental correlates of different types of early experience. A video abstract of this article can be viewed at https://www.youtube.com/watch?v=uEU0L8exyTM.

Mapping of machine learning approaches for description, prediction, and causal inference in the social and health sciences

Machine learning (ML) methodology used in the social and health sciences needs to fit the intended research purposes of description, prediction, or causal inference. This paper provides a comprehensive, systematic meta-mapping of research questions in the social and health sciences to appropriate ML approaches by incorporating the necessary requirements to statistical analysis in these disciplines. We map the established classification into description, prediction, counterfactual prediction, and causal structural learning to common research goals, such as estimating prevalence of adverse social or health outcomes, predicting the risk of an event, and identifying risk factors or causes of adverse outcomes, and explain common ML performance metrics. Such mapping may help to fully exploit the benefits of ML while considering domain-specific aspects relevant to the social and health sciences and hopefully contribute to the acceleration of the uptake of ML applications to advance both basic and applied social and health sciences research.

Evaluating the robustness of targeted maximum likelihood estimators via realistic simulations in nutrition intervention trials

Several recently developed methods have the potential to harness machine learning in the pursuit of target quantities inspired by causal inference, including inverse weighting, doubly robust estimating equations and substitution estimators like targeted maximum likelihood estimation. There are even more recent augmentations of these procedures that can increase robustness, by adding a layer of cross-validation (cross-validated targeted maximum likelihood estimation and double machine learning, as applied to substitution and estimating equation approaches, respectively). While these methods have been evaluated individually on simulated and experimental data sets, a comprehensive analysis of their performance across real data based simulations have yet to be conducted. In this work, we benchmark multiple widely used methods for estimation of the average treatment effect using ten different nutrition intervention studies data. A nonparametric regression method, undersmoothed highly adaptive lasso, is used to generate the simulated distribution which preserves important features from the observed data and reproduces a set of true target parameters. For each simulated data, we apply the methods above to estimate the average treatment effects as well as their standard errors and resulting confidence intervals. Based on the analytic results, a general recommendation is put forth for use of the cross-validated variants of both substitution and estimating equation estimators. We conclude that the additional layer of cross-validation helps in avoiding unintentional over-fitting of nuisance parameter functionals and leads to more robust inferences.

Cumulative risk, factor analysis, and latent class analysis of childhood adversity data in a nationally representative sample

BACKGROUND

Childhood adversity is a multifaceted construct that is in need of comprehensive operationalisation.

OBJECTIVE

The aim of this study was to explore the optimal method to operationalise a scale of adverse childhood experiences (ACEs).

PARTICIPANTS AND SETTING

Data were from Wave 1 of the Personality and Total Health (PATH) Through Life Project (N = 7485, 51% women). Participants from three age groups (20-25, 40-45, 60-65) retrospectively reported their childhood experiences of domestic adversity on a 17-item scale (e.g., physical abuse, verbal abuse, neglect, poverty).

METHODS

We compared three approaches to operationalising the 17-item scale: a cumulative risk approach, factor analysis, and latent class analysis (LCA). The cumulative risk and dimensional models were represented by a unidimensional and two-dimensional model respectively using confirmatory factor analysis (CFA).

RESULTS

The cumulative risk approach and LCA were viable approaches to operationalising ACE data in PATH. CFA of the dimensional model produced latent factors of threat and deprivation that were highly correlated, potentially leading to problems with multicollinearity when estimating associations. LCA revealed six classes of ACEs: high adversity, low adversity, low affection, authoritarian upbringing, high parental dysfunction, and moderate parental dysfunction.

CONCLUSION

Our study found multiple latent classes within a 17-item questionnaire assessing domestic adversity. Using both the cumulative method and latent class approach may be a more informative approach when examining the relationship between ACEs and later health outcomes. Future ACE studies may benefit by considering multi-dimensional approaches to operationalising adversity.

Intimate partner violence during infancy and cognitive outcomes in middle childhood: Results from an Australian community-based mother and child cohort study

The cognitive functioning of children who experience intimate partner violence (IPV) has received less attention than their emotional-behavioral outcomes. Drawing upon data from 615 (48.4% female) 10-year-old Australian-born children and their mothers (9.6% of mothers born in non-English speaking countries) participating in a community-based longitudinal study between 2004 and 2016, this study examined the associations between IPV in infancy and cognition in middle childhood (at age 10). Results showed that IPV in the first 12 months of life was associated with lower general cognitive ability and poorer executive attention but not working memory skills. IPV in middle childhood (in the 10th year postpartum) was not associated with cognition. This study provides evidence for the long-term impact of early life exposure to IPV on children's cognition, and points to the importance of early intervention to optimize development.

Extracurricular Activities, Screen Media Activity, and Sleep May Be Modifiable Factors Related to Children's Cognitive Functioning: Evidence From the ABCD Study®

This study used a machine learning framework in conjunction with a large battery of measures from 9,718 school-age children (ages 9-11) from the Adolescent Brain Cognitive DevelopmentSM (ABCD) Study to identify factors associated with fluid cognitive functioning (FCF), or the capacity to learn, solve problems, and adapt to novel situations. The identified algorithm explained 14.74% of the variance in FCF, replicating previously reported socioeconomic and mental health contributors to FCF, and adding novel and potentially modifiable contributors, including extracurricular involvement, screen media activity, and sleep duration. Pragmatic interventions targeting these contributors may enhance cognitive performance and protect against their negative impact on FCF in children.

A scoping review on the use of machine learning in research on social determinants of health: Trends and research prospects

Background

Machine learning (ML) has spread rapidly from computer science to several disciplines. Given the predictive capacity of ML, it offers new opportunities for health, behavioral, and social scientists. However, it remains unclear how and to what extent ML is being used in studies of social determinants of health (SDH).

Methods

Using four search engines, we conducted a scoping review of studies that used ML to study SDH (published before May 1, 2020). Two independent reviewers analyzed the relevant studies. For each study, we identified the research questions, Results, data, and algorithms. We synthesized our findings in a narrative report.

Results

Of the initial 8097 hits, we identified 82 relevant studies. The number of publications has risen during the past decade. More than half of the studies (n = 46) used US data. About 80% (n = 66) utilized surveys, and 70% (n = 57) employed ML for common prediction tasks. Although the number of studies in ML and SDH is growing rapidly, only a few studies used ML to improve causal inference, curate data, or identify social bias in predictions (i.e., algorithmic fairness).

Conclusions

While ML equips researchers with new ways to measure health outcomes and their determinants from non-conventional sources such as text, audio, and image data, most studies still rely on traditional surveys. Although there are no guarantees that ML will lead to better social epidemiological research, the potential for innovation in SDH research is evident as a result of harnessing the predictive power of ML for causality, data curation, or algorithmic fairness.

Comparison of three models of adverse childhood experiences: Associations with child and adolescent internalizing and externalizing symptoms

Exposure to adverse childhood experiences (ACEs) is prevalent and confers risk for psychopathology later in life. Approaches to understanding the impact of ACEs on development include the independent risk approach, the Dimensional Model of Adversity and Psychopathology (DMAP) distinguishing between threat and deprivation events, and the cumulative risk approach. The present research provides an empirical confirmation of DMAP and a comparison of these three approaches in predicting internalizing and externalizing symptoms in youth. In Study 1, mental health professionals (N = 57) rated ACEs as threat or deprivation events. These ratings were used to create composites to represent the DMAP approach in Study 2. With cross-sectional and longitudinal data from children and adolescents in state custody (N = 23,850), hierarchical linear regression analyses examined independent risk, DMAP, and cumulative risk models in predicting internalizing symptoms, disinhibited externalizing symptoms, and antagonistic externalizing symptoms. All three approaches produced significant models and revealed associations between exposure to ACEs and symptoms. Individual risk accounted for significantly more variance in symptoms than cumulative risk and DMAP. Cumulative risk masked differential associations between ACEs and psychological symptoms found in the individual risk and DMAP approaches. (PsycInfo Database Record (c) 2021 APA, all rights reserved).

Rethinking Concepts and Categories for Understanding the Neurodevelopmental Effects of Childhood Adversity

Discovering the processes through which early adverse experiences affect children's nervous-system development, health, and behavior is critically important for developing effective interventions. However, advances in our understanding of these processes have been constrained by conceptualizations that rely on categories of adversity that are overlapping, have vague boundaries, and lack consistent biological evidence. Here, we discuss central problems in understanding the link between early-life adversity and children's brain development. We conclude by suggesting alternative formulations that hold promise for advancing knowledge about the neurobiological mechanisms through which adversity affects human development.

Early life stress and development: potential mechanisms for adverse outcomes

BACKGROUND

Chronic and/or extreme stress in early life, often referred to as early adversity, childhood trauma, or early life stress, has been associated with a wide range of adverse effects on development. However, while early life stress has been linked to negative effects on a number of neural systems, the specific mechanisms through which early life stress influences development and individual differences in children's outcomes are still not well understood.

MAIN TEXT

The current paper reviews the existing literature on the neurobiological effects of early life stress and their ties to children's psychological and behavioral development.

CONCLUSIONS

Early life stress has persistent and pervasive effects on prefrontal-hypothalamic-amygdala and dopaminergic circuits that are at least partially mediated by alterations in hypothalamic-pituitary-adrenal axis function. However, to date, this research has primarily utilized methods of assessment that focus solely on children's event exposures. Incorporating assessment of factors that influence children's interpretation of stressors, along with stressful events, has the potential to provide further insight into the mechanisms contributing to individual differences in neurodevelopmental effects of early life stress. This can aid in further elucidating specific mechanisms through which these neurobiological changes influence development and contribute to risk for psychopathology and health disorders.

Earlier age at menarche as a transdiagnostic mechanism linking childhood trauma with multiple forms of psychopathology in adolescent girls

BACKGROUND

Although early life adversity (ELA) increases risk for psychopathology, mechanisms linking ELA with the onset of psychopathology remain poorly understood. Conceptual models have argued that ELA accelerates development. It is unknown whether all forms of ELA are associated with accelerated development or whether early maturation is a potential mechanism linking ELA with psychopathology. We examine whether two distinct dimensions of ELA - threat and deprivation - have differential associations with pubertal timing in girls, and evaluate whether accelerated pubertal timing is a mechanism linking ELA with the onset of adolescent psychopathology.

METHODS

Data were drawn from a large, nationally representative sample of 4937 adolescent girls. Multiple forms of ELA characterized by threat and deprivation were assessed along with age at menarche (AAM) and the onset of DSM-IV fear, distress, externalizing, and eating disorders.

RESULTS

Greater exposure to threat was associated with earlier AAM (B = -0.1, p = 0.001). Each 1-year increase in AAM was associated with reduced odds of fear, distress, and externalizing disorders post-menarche (ORs = 0.74-0.85). Earlier AAM significantly mediated the association between exposure to threat and post-menarche onset of distress (proportion mediated = 6.2%), fear (proportion mediated = 16.3%), and externalizing disorders (proportion mediated = 2.9%).

CONCLUSIONS

Accelerated pubertal development in girls may be one transdiagnostic pathway through which threat-related experiences confer risk for the adolescent onset of mental disorders. Early pubertal maturation is a marker that could be used in both medical and mental health settings to identify trauma-exposed youth that are at risk for developing a mental disorder during adolescence in order to better target early interventions.

The Flynn effect for fluid IQ may not generalize to all ages or ability levels: a population-based study of 10,000 US adolescents

Generational changes in IQ (the Flynn Effect) have been extensively researched and debated. Within the US, gains of 3 points per decade have been accepted as consistent across age and ability level, suggesting that tests with outdated norms yield spuriously high IQs. However, findings are generally based on small samples, have not been validated across ability levels, and conflict with reverse effects recently identified in Scandinavia and other countries. Using a well-validated measure of fluid intelligence, we investigated the Flynn Effect by comparing scores normed in 1989 and 2003, among a representative sample of American adolescents ages 13-18 (n=10,073). Additionally, we examined Flynn Effect variation by age, sex, ability level, parental age, and SES. Adjusted mean IQ differences per decade were calculated using generalized linear models. Overall the Flynn Effect was not significant; however, effects varied substantially by age and ability level. IQs increased 2.3 points at age 13 (95% CI=2.0, 2.7), but decreased 1.6 points at age 18 (95% CI=-2.1, -1.2). IQs decreased 4.9 points for those with IQ<70 (95% CI=-4.9, -4.8), but increased 3.5 points among those with IQ>130 (95% CI=3.4, 3.6). The Flynn Effect was not meaningfully related to other background variables. Using the largest sample of US adolescent IQs to date, we demonstrate significant heterogeneity in fluid IQ changes over time. Reverse Flynn Effects at age 18 are consistent with previous data, and those with lower ability levels are exhibiting worsening IQ over time. Findings by age and ability level challenge generalizing IQ trends throughout the general population.

Associations between cortical thickness and reasoning differ by socioeconomic status in development

Although lower socioeconomic status (SES) is generally negatively associated with performance on cognitive assessments, some children from lower-SES backgrounds perform as well as their peers from higher-SES backgrounds. Yet little research has examined whether the neural correlates of individual differences in cognition vary by SES. The current study explored whether relationships between cortical structure and fluid reasoning differ by SES in development. Fluid reasoning, a non-verbal component of IQ, is supported by a distributed frontoparietal network, with evidence for a specific role of rostrolateral prefrontal cortex (RLPFC). In a sample of 115 4-7-year old children, bilateral thickness of RLPFC differentially related to reasoning by SES: thicker bilateral RLPFC positively correlated with reasoning ability in children from lower-SES backgrounds, but not in children from higher-SES backgrounds. Similar results were found in an independent sample of 59 12-16-year old adolescents. Furthermore, young children from lower-SES backgrounds with strong reasoning skills were the only group to show a positive relationship between RLPFC thickness and age. In sum, we found that relationships between cortical thickness and cognition differ by SES during development.