A Bayesian approach to comparing common models of life-course epidemiology.
Ontology highlight
ABSTRACT: Background
Life-course epidemiology studies people's health over long periods, treating repeated measures of their experiences (usually risk factors) as predictors or causes of subsequent morbidity and mortality. Three hypotheses or models often guide the analyst in assessing these sequential risks: the accumulation model (all measurement occasions are equally important for predicting the outcome), the critical period model (only one occasion is important) and the sensitive periods model (a catch-all model for any other pattern of temporal dependence).Methods
We propose a Bayesian omnibus test of these three composite models, as well as post hoc decompositions that identify their best respective sub-models. We test the approach via simulations, before presenting an empirical example that relates five sequential measurements of body weight to an RNAseq measure of colorectal-cancer disposition.Results
The approach correctly identifies the life-course model under which the data were simulated. Our empirical cohort study indicated with >90% probability that colorectal-cancer disposition reflected a sensitive process, with current weight being most important but prior body weight also playing a role.Conclusions
The Bayesian methods we present allow precise inferences about the probability of life-course models given the data and are applicable in realistic scenarios involving causal analysis and missing data.
SUBMITTER: Chumbley J
PROVIDER: S-EPMC8580273 | biostudies-literature |
REPOSITORIES: biostudies-literature
Similar Datasets
Project description:BackgroundLife course epidemiology examines associations between repeated measures of risk and health outcomes across different phases of life. Empirical research, however, is often based on discrete-time models that assume that sporadic measurement occasions fully capture underlying long-term continuous processes of risk.MethodsWe propose (i) the functional relevant life course model (fRLM), which treats repeated, discrete measures of risk as unobserved continuous processes, and (ii) a testing procedure to assign probabilities that the data correspond to conceptual models of life course epidemiology (critical period, sensitive period and accumulation models). The performance of the fRLM is evaluated with simulations, and the approach is illustrated with empirical applications relating body mass index (BMI) to mRNA-seq signatures of chronic kidney disease, inflammation and breast cancer.ResultsSimulations reveal that fRLM identifies the correct life course model with three to five repeated assessments of risk and 400 subjects. The empirical examples reveal that chronic kidney disease reflects a critical period process and inflammation and breast cancer likely reflect sensitive period mechanisms.ConclusionsThe proposed fRLM treats repeated measures of risk as continuous processes and, under realistic data scenarios, the method provides accurate probabilities that the data correspond to commonly studied models of life course epidemiology. fRLM is implemented with publicly-available software.
| S-EPMC10859158 | biostudies-literature
Project description:BackgroundDifferent hypotheses have been proposed in life course epidemiology on how a time-varying exposure can affect health or disease later in life. Researchers are often interested in investigating the probability of these hypotheses based on observed life course data. However, current techniques based on model/variable selection do not provide a direct estimate of this probability. We propose an alternative technique for a continuous exposure, using a Bayesian approach that has specific advantages, to investigate which life course hypotheses are supported by the observed data.MethodsWe demonstrate the technique, the relevant life course exposure model (RLM), using simulations. We also analyse data from a case-control study on risk factors of oral cancer, with repeated measurements of betel quid chewing across life. We investigate the relative importance of chewing one quid of betel per day, at three life periods: ?20?years, 21-40?years and above 40?years of age, on the risk of developing oral cancer.ResultsRLM was able to correctly identify the life course hypothesis under which the data were simulated. Results from the case-control study showed that there was 74.3% probability that betel quid exposure earlier in life, compared with later, results in higher odds of developing oral cancer later in life.ConclusionsRLM is a useful option to identify the life course hypothesis supported by the observed data prior to the estimation of a causal effect.
| S-EPMC6208282 | biostudies-literature
Project description:The application of the life course approach to social epidemiology has helped epidemiologists theoretically examine social gradients in population health. Longitudinal data with rich contextual information collected repeatedly and advanced statistical approaches have made this challenging task easier. This review paper provides an overview of the life course approach in epidemiology, its research application, and future challenges. In summary, a systematic approach to methods, including theoretically guided measurement of socioeconomic position, would assist researchers in gathering evidence for reducing social gradients in health, and collaboration across individual disciplines will make this task achievable.
| S-EPMC4150004 | biostudies-literature
Project description:BackgroundUnderstanding the processes across childhood and adolescence that affect later life inequalities depends on many variables for a large number of individuals measured over substantial time periods. Linkable administrative data were used to generate birth cohorts and to study pathways of inequity in childhood and early adolescence leading to differences in educational attainment. Advantages and disadvantages of using large administrative data bases for such research were highlighted.MethodsChildren born in Manitoba, Canada between 1982 and 1995 were followed until age 19 (N = 89,763), with many time-invariant measures serving as controls. Five time-varying predictors of high school graduation-three social and two health-were modelled using logistic regression and a framework for examining predictors across the life course. For each time-varying predictor, six temporal patterns were tested: full, accumulation of risk, sensitive period, and three critical period models.ResultsPredictors measured in early adolescence generated the highest odds ratios, suggesting the importance of adolescence. Full models provided the best fit for the three time-varying social measures. Residence in a low-income neighborhood was a particularly influential predictor of not graduating from high school. The transmission of risk across developmental periods was also highlighted; exposure in one period had significant implications for subsequent life stages.ConclusionThis study advances life course epidemiology, using administrative data to clarify the relationships among several measures of social behavior, cognitive development, and health. Analyses of temporal patterns can be useful in studying such other outcomes as educational achievement, teen pregnancy, and workforce participation.
| S-EPMC5744927 | biostudies-literature
Project description:BackgroundA relationship between current socio-economic position and subjective quality of life has been demonstrated, using wellbeing, life and needs satisfaction approaches. Less is known regarding the influence of different life course socio-economic trajectories on later quality of life. Several conceptual models have been proposed to help explain potential life course effects on health, including accumulation, latent, pathway and social mobility models. This systematic review aimed to assess whether evidence supported an overall relationship between life course socio-economic position and quality of life during adulthood and if so, whether there was support for one or more life course models.MethodsA review protocol was developed detailing explicit inclusion and exclusion criteria, search terms, data extraction items and quality appraisal procedures. Literature searches were performed in 12 electronic databases during January 2012 and the references and citations of included articles were checked for additional relevant articles. Narrative synthesis was used to analyze extracted data and studies were categorized based on the life course model analyzed.ResultsTwelve studies met the eligibility criteria and used data from 10 datasets and five countries. Study quality varied and heterogeneity between studies was high. Seven studies assessed social mobility models, five assessed the latent model, two assessed the pathway model and three tested the accumulation model. Evidence indicated an overall relationship, but mixed results were found for each life course model. Some evidence was found to support the latent model among women, but not men. Social mobility models were supported in some studies, but overall evidence suggested little to no effect. Few studies addressed accumulation and pathway effects and study heterogeneity limited synthesis.ConclusionsTo improve potential for synthesis in this area, future research should aim to increase study comparability. Recommendations include testing all life course models within individual studies and the use of multiple measures of socio-economic position and quality of life. Comparable cross-national data would be beneficial to enable investigation of between-country differences.
| S-EPMC3490823 | biostudies-literature
Project description:Linear-bilinear models, especially the additive main effects and multiplicative interaction (AMMI) model, are widely applicable to genotype-by-environment interaction (GEI) studies in plant breeding programs. These models allow a parsimonious modeling of GE interactions, retaining a small number of principal components in the analysis. However, one aspect of the AMMI model that is still debated is the selection criteria for determining the number of multiplicative terms required to describe the GE interaction pattern. Shrinkage estimators have been proposed as selection criteria for the GE interaction components. In this study, a Bayesian approach was combined with the AMMI model with shrinkage estimators for the principal components. A total of 55 maize genotypes were evaluated in nine different environments using a complete blocks design with three replicates. The results show that the traditional Bayesian AMMI model produces low shrinkage of singular values but avoids the usual pitfalls in determining the credible intervals in the biplot. On the other hand, Bayesian shrinkage AMMI models have difficulty with the credible interval for model parameters, but produce stronger shrinkage of the principal components, converging to GE matrices that have more shrinkage than those obtained using mixed models. This characteristic allowed more parsimonious models to be chosen, and resulted in models being selected that were similar to those obtained by the Cornelius F-test (? = 0.05) in traditional AMMI models and cross validation based on leave-one-out. This characteristic allowed more parsimonious models to be chosen and more GEI pattern retained on the first two components. The resulting model chosen by posterior distribution of singular value was also similar to those produced by the cross-validation approach in traditional AMMI models. Our method enables the estimation of credible interval for AMMI biplot plus the choice of AMMI model based on direct posterior distribution retaining more GEI pattern in the first components and discarding noise without Gaussian assumption as requested in F-based tests or deal with parametric problems as observed in traditional AMMI shrinkage method.
| S-EPMC4497624 | biostudies-literature
Project description:In data modelling using the composite Pareto distribution, any observations above a particular threshold value are assumed to follow Pareto type distribution, whereas the rest of the observations are assumed to follow a different distribution. This paper proposes on the use of Bayesian approach to the composite Pareto models involving specification of the prior distribution on the proportion of data coming from the Pareto distribution, instead of assuming the prior distribution on the threshold, as often done in the literature. Based on a simulation study, it is found that the parameter estimates determined when using uniform prior on the proportion is less biased as compared to the point estimates determined when using uniform prior on the threshold. Applications on income data and finance are included for illustrative examples.
| S-EPMC8460011 | biostudies-literature
Project description:Data obtained from genitourinary medicine clinics through a comprehensive surveillance system were used in a Bayesian mixed-effects Poisson regression model to explore socio-demographic individual and ecological risk factors for gonorrhoea in London, as well as its spatial clustering. The spatial analysis was performed at the Middle-layer Super Output Area level (median population size 7200). A total of 12452 individuals were diagnosed during the 2-year study period (2009-2010). The study confirmed the presence of 'core areas' of high incidence, and identified 'core' high-risk groups, in particular young adults (16-29 years), males, black Caribbeans and more deprived areas. The individual (age, sex, ethnicity) and area-level (deprivation, teenage pregnancies, students) model covariates accounted for 48% of the variance. Most of the remaining variance was explained by the spatial effect, thus capturing other spatially distributed factors associated with gonorrhoea, such as local sexual networks. These findings will be useful in identifying areas for targeted interventions, such as STI testing and health promotion.
| S-EPMC9161308 | biostudies-literature
Project description:Mobility is the most studied and most relevant physical ability affecting quality of life with strong prognostic value for disability and survival. Natural selection has built the "engine" of mobility with great robustness, redundancy, and functional reserve. Efficient patterns of mobility can be acquired during development even by children affected by severe impairments. Analogously, age-associated impairments in mobility-related physiological systems are compensated and overt limitations of mobility only occur when the severity can no longer be compensated. Mobility loss in older persons usually results from multiple impairments in the central nervous system, muscles, joints, and energetic and sensory physiological systems. Early preclinical changes in these physiological systems that precede mobility loss have been poorly studied. Peak performance, rate of decline, compensatory behaviors, or subclinical deterioration of physiological resources may cumulatively influence both timing of mobility loss and chances of recovery, but their role as risk factors has not been adequately characterized. Understanding the natural history of these early changes and intervening on them would likely be the most effective strategy to reduce the burden of disability in the population. For example, young women with low bone peak mass could be counseled to start strength resistance exercise to reduce their high risk of developing osteoporosis and fracture later in life. Expanding this approach to other physiological domains requires collecting and interpreting data from life course epidemiological studies, establishing normative measures of mobility, physical function, and physical activity, and connecting them with life course trajectories of the mobility-relevant physiological domains.
| S-EPMC4978365 | biostudies-literature
Project description:Gaussian graphical models are commonly used to characterize conditional (in)dependence structures (i.e., partial correlation networks) of psychological constructs. Recently attention has shifted from estimating single networks to those from various subpopulations. The focus is primarily to detect differences or demonstrate replicability. We introduce two novel Bayesian methods for comparing networks that explicitly address these aims. The first is based on the posterior predictive distribution, with a symmetric version of Kullback-Leibler divergence as the discrepancy measure, that tests differences between two (or more) multivariate normal distributions. The second approach makes use of Bayesian model comparison, with the Bayes factor, and allows for gaining evidence for invariant network structures. This overcomes limitations of current approaches in the literature that use classical hypothesis testing, where it is only possible to determine whether groups are significantly different from each other. With simulation we show the posterior predictive method is approximately calibrated under the null hypothesis (α = .05) and has more power to detect differences than alternative approaches. We then examine the necessary sample sizes for detecting invariant network structures with Bayesian hypothesis testing, in addition to how this is influenced by the choice of prior distribution. The methods are applied to posttraumatic stress disorder symptoms that were measured in 4 groups. We end by summarizing our major contribution, that is proposing 2 novel methods for comparing Gaussian graphical models (GGMs), which extends beyond the social-behavioral sciences. The methods have been implemented in the R package BGGM. (PsycInfo Database Record (c) 2020 APA, all rights reserved).
| S-EPMC8572134 | biostudies-literature