An integrative Bayesian approach to matrix-based analysis in neuroimaging

Understanding the correlation structure associated with brain regions is a central goal in neuroscience, as it informs about interregional relationships and network organization. Correlation structure can be conveniently captured in a matrix that indicates the relationships among brain regions, which could involve electroencephalogram sensors, electrophysiology recordings, calcium imaging data, or functional magnetic resonance imaging (FMRI) data-We call this type of analysis matrix-based analysis, or MBA. Although different methods have been developed to summarize such matrices across subjects, including univariate general linear models (GLMs), the available modeling strategies tend to disregard the interrelationships among the regions, leading to "inefficient" statistical inference. Here, we develop a Bayesian multilevel (BML) modeling framework that simultaneously integrates the analyses of all regions, region pairs (RPs), and subjects. In this approach, the intricate relationships across regions as well as across RPs are quantitatively characterized. The adoption of the Bayesian framework allows us to achieve three goals: (a) dissolve the multiple testing issue typically associated with seeking evidence for the effect of each RP under the conventional univariate GLM; (b) make inferences on effects that would be treated as "random" under the conventional linear mixed-effects framework; and (c) estimate the effect of each brain region in a manner that indexes their relative "importance". We demonstrate the BML methodology with an FMRI dataset involving a cognitive-emotional task and compare it to the conventional GLM approach in terms of model efficiency, performance, and inferences. The associated program MBA is available as part of the AFNI suite for general use.

A tutorial on individual participant data meta-analysis using Bayesian multilevel modeling to estimate alcohol intervention effects across heterogeneous studies

This paper provides a tutorial companion for the methodological approach implemented in Huh et al. (2015) that overcame two major challenges for individual participant data (IPD) meta-analysis. Specifically, we show how to validly combine data from heterogeneous studies with varying numbers of treatment arms, and how to analyze highly-skewed count outcomes with many zeroes (e.g., alcohol and substance use outcomes) to estimate overall effect sizes. These issues have important implications for the feasibility, applicability, and interpretation of IPD meta-analysis but have received little attention thus far in the applied research literature. We present a Bayesian multilevel modeling approach for combining multi-arm trials (i.e., those with two or more treatment groups) in a distribution-appropriate IPD analysis. Illustrative data come from Project INTEGRATE, an IPD meta-analysis study of brief motivational interventions to reduce excessive alcohol use and related harm among college students. Our approach preserves the original random allocation within studies, combines within-study estimates across all studies, overcomes between-study heterogeneity in trial design (i.e., number of treatment arms) and/or study-level missing data, and derives two related treatment outcomes in a multivariate IPD meta-analysis. This methodological approach is a favorable alternative to collapsing or excluding intervention groups within multi-arm trials, making it possible to directly compare multiple treatment arms in a one-step IPD meta-analysis. To facilitate application of the method, we provide annotated computer code in R along with the example data used in this tutorial.

Longitudinal Changes of Functional Capacities Among Adolescent Female Basketball Players

Background: The interpretation of young athletes' performance during pubertal years is important to support coaches' decisions, as performance may be erroneously interpreted due to the misalignment between chronological age (CA), biological age (BA) and sport age (SA). Aim: Using a Bayesian multilevel approach, the variation in longitudinal changes in performance was examined considering the influence of CA, BA (age at menarche), SA, body size, and exposure to training among female basketball players. Method: The study had a mixed-longitudinal design. Thirty eight female basketball players (aged 13.38 ± 1.25 years at baseline) were measured three times per season. CA, BA and SA were obtained. Anthropometric and functional measures: countermovement jump, Line drill (LD), Yo-Yo (Yo-Yo IR1). Based on the sum of the z-scores, an index of overall performance was estimated. The effects of training on longitudinal changes in performance were modeled. Results: A decrease in the rate of improvements was apparent at about 14 years of age. When aligned for BA, the slowing of the rate of improvements is apparent about 2 years after menarche for LD. For countermovement jump longitudinal changes, when performance was aligned for BA improvements became linear. For Yo-Yo IR1 and performance index, both indicators showed a linear trend of improvement when aligned for CA and BA, separately. Older players showed higher rates of improvement for Yo-Yo IR1 and performance index from pre-season to end-season. When considering performance changes aligned for BA it was apparent an improvement of performance as players became biologically mature. Conclusions and Implications: The alignment of CA with BA and SA provides important information for coaches. Human growth follows a genetically determined pattern, despite variation in both tempo and timing. When the effects of maturation reach their end, all the girls went through the same process. Hence, there is no need to artificially manipulate youth competitions in order to accelerate gains that sooner or later reach their peak and tend to flat their improvement curve.

Accounting for retest effects in cognitive testing with the Bayesian double exponential model via intensive measurement burst designs

Monitoring early changes in cognitive performance is useful for studying cognitive aging as well as for detecting early markers of neurodegenerative diseases. Repeated evaluation of cognition via a measurement burst design can accomplish this goal. In such design participants complete brief evaluations of cognition, multiple times per day for several days, and ideally, repeat the process once or twice a year. However, long-term cognitive change in such repeated assessments can be masked by short-term within-person variability and retest learning (practice) effects. In this paper, we show how a Bayesian double exponential model can account for retest gains across measurement bursts, as well as warm-up effects within a burst, while quantifying change across bursts in peak performance. We also highlight how this approach allows for the inclusion of person-level predictors and draw intuitive inferences on cognitive change with Bayesian posterior probabilities. We use older adults' performance on cognitive tasks of processing speed and spatial working memory to demonstrate how individual differences in peak performance and change can be related to predictors of aging such as biological age and mild cognitive impairment status.

Identifying and assessing the impact of key neighborhood-level determinants on geographic variation in stroke: a machine learning and multilevel modeling approach

Background

Stroke is a chronic cardiovascular disease that puts major stresses on U.S. health and economy.

The prevalence of stroke exhibits a strong geographical pattern at the state-level, where a cluster of southern states with a substantially higher prevalence of stroke has been called the stroke belt of the nation. Despite this recognition, the extent to which key neighborhood characteristics affect stroke prevalence remains to be further clarified.

Methods

We generated a new neighborhood health data set at the census tract level on nearly 27,000 tracts by pooling information from multiple data sources including the CDC’s 500 Cities Project 2017 data release. We employed a two-stage modeling approach to understand how key neighborhood-level risk factors affect the neighborhood-level stroke prevalence in each state of the US. The first stage used a state-of-the-art Bayesian machine learning algorithm to identify key neighborhood-level determinants. The second stage applied a Bayesian multilevel modeling approach to describe how these key determinants explain the variability in stroke prevalence in each state.

Results

Neighborhoods with a larger proportion of older adults and non-Hispanic blacks were associated with neighborhoods with a higher prevalence of stroke. Higher median household income was linked to lower stroke prevalence. Ozone was found to be positively associated with stroke prevalence in 10 states, while negatively associated with stroke in five states. There was substantial variation in both the direction and magnitude of the associations between these four key factors with stroke prevalence across the states.

Conclusions

When used in a principled variable selection framework, high-performance machine learning can identify key factors of neighborhood-level prevalence of stroke from wide-ranging information in a data-driven way. The Bayesian multilevel modeling approach provides a detailed view of the impact of key factors across the states. The identified major factors and their effect mechanisms can potentially aid policy makers in developing area-based stroke prevention strategies.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12889-020-09766-3.

Multilevel moderated mediation model with ordinal outcome

Although increasingly complex models have been proposed in mediation literature, there is no model nor software that incorporates the multiple possible generalizations of the simple mediation model jointly. We propose a flexible moderated mediation model allowing for (1) a hierarchical structure of clustered data, (2) more and possibly correlated mediators, and (3) an ordinal outcome. The motivating data set is obtained from a European study in nursing research. Patients' willingness to recommend their treating hospital was recorded in an ordinal way. The research question is whether such recommendation directly depends on system-level features in the organization of nursing care, or whether these associations are mediated by 2 measurements of nursing care left undone and possibly moderated by nurse education. We have developed a Bayesian approach and accompanying program that takes all the above generalizations into account.

Provider Contributions to Disparities in Mental Health Care

OBJECTIVE

Disparities in diagnosis of mental health problems and in access to treatment among racial-ethnic groups are apparent across different behavioral conditions, particularly in the quality of treatment for depression. This study aimed to determine how much disparities differ across providers.

METHODS

Bayesian mixed-effects models were used to estimate whether disparities in patient adherence to antidepressant medication (N=331,776) or psychotherapy (N=275,095) were associated with specific providers. Models also tested whether providers who achieved greater adherence to treatment, on average, among non-Hispanic white patients than among patients from racial-ethnic minority groups attained lower disparities and whether the percentage of patients from racial-ethnic minority groups in a provider caseload was associated with disparities.

RESULTS

Disparities in adherence to both antidepressant medication and psychotherapy were associated with the provider. Provider performance with non-Hispanic white patients was negatively correlated with provider-specific disparities in adherence to psychotherapy but not to antidepressants. A higher proportion of patients from racial-ethnic minority groups in a provider's caseload was associated with lower adherence among non-Hispanic white patients, lower disparities in adherence to psychotherapy, and greater disparities in adherence to antidepressant medication.

CONCLUSIONS

Adherence to depression treatment among a provider's patients from racial-ethnic minority groups was related to adherence among that provider's non-Hispanic white patients, but evidence also suggested provider-specific disparities. Efforts among providers to decrease disparities might focus on improving the general skill of providers who treat more patients from racial-ethnic minority groups as well as offering culturally based training to providers with notable disparities.

Growth and weight status of Brazilian children with autism spectrum disorders: A mixed longitudinal study

OBJECTIVE

This study examined the growth status and physical development of Brazilian children with autism spectrum disorders from 4 to 15 years of age. Furthermore, it was examined whether variation in growth patterns and weight status was influenced by the use of psychotropic medications.

METHODS

One-hundred and twenty children aged 3.6-12.1 years at baseline (average=7.2 years, SD=2.3 years) diagnosed with autism spectrum disorders were measured on three repeated occasions across a 4-year period. Stature, body mass, and body mass index were considered. Bayesian multilevel modeling was used to describe the individual growth patterns.

RESULTS

Growth in stature was comparable to the age-specific 50th percentile for Centers for Disease Control and Prevention reference data until approximately 8 years, but a substantial decrease in growth rate was observed thereafter, reaching the age-specific 5th percentile at 15 years of age. Both body mass and body mass index values were, on average, higher than both the Brazilian and Centers for Disease Control and Prevention age-specific 95th percentile reference until 8 years, but below the 50th specific-age percentile at the age of 15 years.

CONCLUSIONS

Brazilian boys with autism spectrum disorders between 4 and 15 years appear to have impaired growth in stature after 8-9 years of age, likely impacting pubertal growth. A high prevalence of overweight and obesity was observed in early childhood, although a trend of substantial decrease in body mass and body mass index was apparent when children with autism spectrum disorders entered the years of pubertal development.

A hierarchical bayesian approach to multi-trait clinical quantitative trait locus modeling

Recent advances in high-throughput genotyping and transcript profiling technologies have enabled the inexpensive production of genome-wide dense marker maps in tandem with huge amounts of expression profiles. These large-scale data encompass valuable information about the genetic architecture of important phenotypic traits. Comprehensive models that combine molecular markers and gene transcript levels are increasingly advocated as an effective approach to dissecting the genetic architecture of complex phenotypic traits. The simultaneous utilization of marker and gene expression data to explain the variation in clinical quantitative trait, known as clinical quantitative trait locus (cQTL) mapping, poses challenges that are both conceptual and computational. Nonetheless, the hierarchical Bayesian (HB) modeling approach, in combination with modern computational tools such as Markov chain Monte Carlo (MCMC) simulation techniques, provides much versatility for cQTL analysis. Sillanpää and Noykova (2008) developed a HB model for single-trait cQTL analysis in inbred line cross-data using molecular markers, gene expressions, and marker-gene expression pairs. However, clinical traits generally relate to one another through environmental correlations and/or pleiotropy. A multi-trait approach can improve on the power to detect genetic effects and on their estimation precision. A multi-trait model also provides a framework for examining a number of biologically interesting hypotheses. In this paper we extend the HB cQTL model for inbred line crosses proposed by Sillanpää and Noykova to a multi-trait setting. We illustrate the implementation of our new model with simulated data, and evaluate the multi-trait model performance with regard to its single-trait counterpart. The data simulation process was based on the multi-trait cQTL model, assuming three traits with uncorrelated and correlated cQTL residuals, with the simulated data under uncorrelated cQTL residuals serving as our test set for comparing the performances of the multi-trait and single-trait models. The simulated data under correlated cQTL residuals were essentially used to assess how well our new model can estimate the cQTL residual covariance structure. The model fitting to the data was carried out by MCMC simulation through OpenBUGS. The multi-trait model outperformed its single-trait counterpart in identifying cQTLs, with a consistently lower false discovery rate. Moreover, the covariance matrix of cQTL residuals was typically estimated to an appreciable degree of precision under the multi-trait cQTL model, making our new model a promising approach to addressing a wide range of issues facing the analysis of correlated clinical traits.

The cardiac correlates of feeling safe in everyday life: A Bayesian replication study

BACKGROUND

Feeling safe and secure has been proposed to dampen autonomic arousal and buffer threat responses. In a previous study, we could show that momentary ratings of subjective safety were associated with elevated heart rate variability (specifically, root mean square of successive differences; RMSSD) and lower heart rate in everyday life, thus suggesting a health-protective role of feeling safe.

METHODS

This study aimed to replicate this effect in a sample of N = 79 adults, applying Bayesian statistics with prior effects of the original study.

RESULTS

Using an ecological momentary assessment (EMA) across three days we could replicate the effect of lower heart rate and higher RMSSD in moments when participants felt more safe. In accordance with the original study, we could also show that the effect on heart rate were independent of RMSSD, thus suggesting a contribution of sympathetic activity to this effect.

CONCLUSION

The findings confirm the connection between momentary feelings of safety and cardiac regulation, thus substantiating research on the health-protective role of psychological safety.

Associations of prior concussion severity with brain microstructure using mean apparent propagator magnetic resonance imaging

Magnetic resonance imaging (MRI) diffusion studies have shown chronic microstructural tissue abnormalities in athletes with history of concussion, but with inconsistent findings. Concussions with post-traumatic amnesia (PTA) and/or loss of consciousness (LOC) have been connected to greater physiological injury. The novel mean apparent propagator (MAP) MRI is expected to be more sensitive to such tissue injury than the conventional diffusion tensor imaging. This study examined effects of prior concussion severity on microstructure with MAP-MRI. Collegiate-aged athletes (N = 111, 38 females; ≥6 months since most recent concussion, if present) completed semistructured interviews to determine the presence of prior concussion and associated injury characteristics, including PTA and LOC. MAP-MRI metrics (mean non-Gaussian diffusion [NG Mean], return-to-origin probability [RTOP], and mean square displacement [MSD]) were calculated from multi-shell diffusion data, then evaluated for associations with concussion severity through group comparisons in a primary model (athletes with/without prior concussion) and two secondary models (athletes with/without prior concussion with PTA and/or LOC, and athletes with/without prior concussion with LOC only). Bayesian multilevel modeling estimated models in regions of interest (ROI) in white matter and subcortical gray matter, separately. In gray matter, the primary model showed decreased NG Mean and RTOP in the bilateral pallidum and decreased NG Mean in the left putamen with prior concussion. In white matter, lower NG Mean with prior concussion was present in all ROI across all models and was further decreased with LOC. However, only prior concussion with LOC was associated with decreased RTOP and increased MSD across ROI. Exploratory analyses conducted separately in male and female athletes indicate associations in the primary model may differ by sex. Results suggest microstructural measures in gray matter are associated with a general history of concussion, while a severity-dependent association of prior concussion may exist in white matter.

Learning from learning loss: Bayesian updating in academic universal screening during learning disruptions

We used Bayesian ordinal regression methods to examine reading and math screening predictive strength and accuracy before and after learning disruptions related to the Covid-19 pandemic. Using a Bayesian updating procedure in which model estimates from previous years were used as Bayesian priors in following years, we found that reading and math screening was similarly predictive before and after Covid-19 prolonged unplanned school closures (PUSCs) and subsequent learning disruptions (odds ratios range across years: 15-25). We additionally found that predictive strength and accuracy varied across grade levels, but this grade variation was insensitive to learning disruptions. These findings demonstrate the practical applicability of Bayesian updating to universal screening prediction, particularly in the context of PUSCs or other learning disruptions that may impact student academic needs. Limitations and future directions for Bayesian methods in screening are discussed.

Multilevel Modelling of the Individual and Regional Level Variability in Predictors of Incomplete Antenatal Care Visit among Women of Reproductive Age in Ethiopia: Classical and Bayesian Approaches

Background: Antenatal care is an operational public health intervention to minimize maternal and child morbidity and mortality. However, for varied reasons, many women fail to complete the recommended number of visits. The objective of this study was to assess antenatal care utilization and identify the factors associated with the incomplete antenatal care visit among reproductive age women in Ethiopia. Methods: The 2019 Ethiopian Mini Demographic and Health Survey data were used for this study. Multilevel logistic regression analysis and two level binary logistic regression models were utilized. Results: Around 56.8% of women in Ethiopia did not complete the recommended number of antenatal care visits. Women from rural areas were about 1.622 times more likely to have incomplete antenatal care compared to women from urban areas. Women who had no pregnancy complication signs were about 2.967 times more likely to have incomplete antenatal care compared to women who had pregnancy complication signs. Women who had a slight problem and a big problem with the distance from a health center were about 1.776 and 2.973 times more likely, respectively, to have incomplete antenatal care compared to women whose distance from a health center was not a problem. Furthermore, women who had ever terminated pregnancy were about 10.6% less likely to have incomplete antenatal care compared to women who had never terminated pregnancy. Conclusions: The design and strengthening of existing interventions (e.g., small clinics) should consider identified factors aimed at facilitating antenatal care visits to promote maternal and child health related outcomes. Issues related to urban–rural disparities and noted hotspot areas for incomplete antenatal care visits should be given special attention.

The role of hyper-palatable foods in energy intake measured using mobile food photography methodology

OBJECTIVE

Hyper-palatable foods (HPF) contain nutrient combinations that are hypothesized to maximize their rewarding effects during consumption. Due to their strong reinforcing properties, HPF are hypothesized to lead to greater energy intake within a meal. However, this premise has not been tested in free-living conditions. The current study examined the association between within-meal HPF intake and 1) measured energy intake and 2) self-reported overeating, assessed within eating occasions using smartphone-based food photography methodology.

METHODS

A total of 29 participants reported food intake and eating experiences (N=345 total eating occasions) in real-time for 4 days using smartphone-based food photography methodology. HPF were identified using a standardized definition. Bayesian multilevel modeling was conducted to investigate the within-person effects of proportional calorie intake from HPF (%kcal from HPF) on total energy intake and subjective overeating. Pre-meal hunger and proportional energy intake from high energy dense (HED) foods were included as covariates.

RESULTS

Results revealed that when participants consumed more %kcal from HPF than their average, they consumed greater total energy during eating occasions, even when controlling for pre-meal hunger and %kcal from HED foods (median β = 0.09, 95% HDI [0.02, 0.16], pd. = 99.56%). Additionally, consuming more %kcal from HPF than average was associated with greater eating despite feeling full, when controlling covariates (median β = 0.15, 95% HDI [-0.02, 0.34], pd = 96.45%).

CONCLUSIONS

The findings supported the premise that HPF themselves may yield greater energy intake and eating despite satiation, measured in real-time and free-living conditions.