Tuesday, Aug 6: 8:30 AM - 10:20 AM
5091
Contributed Papers
Oregon Convention Center
Room: CC-E147
This session presents the latest advancements in Bayesian methods and their applications to a wide range of epidemiological challenges. Researchers will share advancements that enhance robustness and scalability in effect estimation and exposure measurement, incorporate complex interaction effects, expand the toolkit for survival analysis, and provide deeper insights into chemical exposure risk assessment. Furthermore, discussions will include novel ideas for improving infectious disease forecasting, as well as addressing missing covariate challenges in analyzing racial and ethnic health disparities, particularly informed by the spatio-temporal spread of COVID-19.
Main Sponsor
Section on Statistics in Epidemiology
Presentations
Accounting for exposure measurement errors has been recognized as a crucial problem in environmental epidemiology. Bayesian hierarchical models offer a coherent probabilistic framework for evaluating associations between environmental exposures and health effects, which take into account exposure measurement errors introduced by uncertainty in exposure estimates as well as spatial misalignment. While 2-stage Bayesian analyses are often regarded as a good alternative to fully Bayesian analyses when joint estimation is not feasible, there has been minimal research on how to properly propagate uncertainty from the exposure model to the health model in the case of a large number of participant locations along with spatially correlated exposures. We propose a scalable 2-stage Bayesian approach, called a sparse MVN prior approach, based on Vecchia approximation. We compare its performance with existing approaches via simulation, demonstrating results comparable to the fully Bayesian approach. We investigate the association between source-specific and pollutant-specific exposures and birth outcomes for 2012 in Harris County, Texas, using several approaches, including the proposed method.
Keywords
Environmental health
Spatial exposure measurement error
Two-stage Bayesian model
Uncertainty propagation
Vecchia approximation
Combining propensity and prognostic scores enhances the efficiency of matching methods in estimating average treatment effect in observational studies. This paper aims to provide a Bayesian approach of double score estimation as well as a theoretical support of the consistency of the Bayesian estimator. Specifically, we explore the performance of a semiparametric Bayesian model, utilizing Gaussian process priors and addressing potential model mis-specification. We derive asymptotic results to validate the consistency of Bayesian estimators as the sample size increases. Particularly noteworthy is the demonstrated superiority of double-score Bayesian estimators in estimating both the population and conditional average treatment effects. In the simulation study, we analyze the performance of these models under various scenarios with a finite sample size. The results generated by the MCMC algorithm indicate doubly robust estimation under specific conditions. We also apply our proposed single/double score model to a real-world dataset, yielding results that align with existing studies utilizing matching methods.
Keywords
Bayesian Semiparametric
Gaussian Process
Propensity and Prognostic Scores Matching
Bayesian Casual Inference
Markov chain Monte Carlo
Heterogeneous Treatment Effect Estimation
We propose a Bayesian generalized Weibull regression method and develop Accelerated Time to Failure models using Bayesian methods. The parameter estimation procedure is carried out using Hamiltonian Monte Carlo algorithm with No-U-Turn Sampler and compares the results of generalized Weibull regression with exponentiated Weibull regression, Weibull regression, and log-normal distribution across simulated and clinical data sets. We examine the effectiveness of generalized Weibull distribution as a survival model and compare it to more studied probability distributions. In addition to monotone and bathtub hazard shapes, the additional shape parameter in the generalized Weibull distribution provides flexibility to model a broader class of monotone hazard rates.
Keywords
Weibull Regression
Bayesian Inference
Hamiltonian Monte Carlo
No-U-Turn Sampler
Accelerated Failure Time
Epidemiological evidence supports an association between maternal exposure to air pollution and birth and child health outcomes. Typically, such associations are estimated by regressing a scalar outcome on daily or weekly measures of exposure during pregnancy using a distributed lag model. However, these associations may be modified by area- or individual-level factors. We propose a Bayesian distributed lag interaction model that allows for a continuous index, a weighted average of multiple modifiers, to modify the association between repeated measures of exposure and an outcome. We estimate our model with a spline cross-basis in a Bayesian hierarchical model. Our model framework allows for simultaneous estimation of index weights and the exposure-time-response function. The index parameterization regularizes the model when modifiers are correlated. Through simulations, we showed that our model out-performs competing methods when there are multiple modifiers of unknown importance. We applied our proposed method to a Colorado birth cohort and estimated the association between birth weight and air pollution modified by a continuous index comprising area- and individual-level factors.
Keywords
distributed lag models
bayesian hierarchical models
splines
effect modification
environmental epidemiology
fine particulate matter
Analyzing health effects associated with exposure to environmental chemical mixtures is a challenging problem in epidemiology, toxicology, and exposure science. In particular, when there are a large number of chemicals under consideration it is difficult to estimate the interactive effects without incorporating reasonable prior information. Based on substantive considerations, researchers believe that true interactions between chemicals need to incorporate their corresponding main effects. In this paper, we use this prior knowledge through a shrinkage prior that a priori assumes an interaction term can only occur when the corresponding main effects exist. Our initial development is for logistic regression with linear chemical effects. We extend this formulation to include non-linear exposure effects and to account for exposure subject to detection limit. We develop an MCMC algorithm using a shrinkage prior that shrinks the interaction terms closer to zero as the main effects get closer to zero. We examine the performance of our methodology through simulation studies and illustrate an analysis of chemical interactions in a case-control study in cancer.
Keywords
Chemical mixture
Interaction
Shrinkage
Collapsed Gibbs
Accurate estimation of region-specific infectious disease reproductive numbers is critical for disaster planning and resource allocation. However, estimates of the reproductive number can substantially vary by region, especially in the absence of quality data. We propose a "Bayesian spatiotemporal model (via Bayesian-INLA)" to improve estimation of the reproductive number by borrowing information from similar regions. We show that this method can lead to improved infectious disease forecasts over standard approaches. We employ an extended SEIR-type compartmental model for forecasting the COVID-19 disease at the county level in South Carolina. We employ the model in forecasting a wave of COVID-19 using the reproductive number estimate incorporating the information from the current wave and previous wave through the standard techniques and the proposed method, then assess the model performance for all the estimation methods using the percentage agreement metric. We also compare the results based on the percentage agreement of county ranking for identifying the most impacted areas. The proposed method is effective in small area infectious disease forecasting.
Keywords
Infectious Disease Epidemiology
Modeling
COVID-19
Reproductive Number
Bayesian Methodology
While panel data of disease incidence are more plentiful than ever, utilizing these data to promote equitable decision-making requires careful modelling of subpopulation-level disparities. In these kinds of tasks, e.g., comparing the trajectory of the COVID-19 pandemic between different racial/ethnic groups, a prevailing challenge is the presence of nonignorable missingness in the demographic covariates of interest. Unfortunately, most spatio-temporal models used in epidemiology inaccurately assume unobserved covariates are missing-at-random (MAR), and most missingness-process models are not spatio-temporal in nature. We respond to this issue with a Bayesian methodology for spatio-temporal modelling the joint distribution of disease counts and discrete-covariate missingness. We also demonstrate the advantage of our model using a simulation study. Finally, we apply the model to COVID-19 incidence data collected in Michigan to describe racial/ethnic disparities in the progression of the COVID-19 pandemic.
Keywords
Count Data Modelling
MNAR Covariates
Bayesian Inference
Spatio-Temporal Disease Modelling
Infectious-Disease