Addressing complex structures of measurement error arising in the exposure assessment in occupational epidemiology using a Bayesian hierarchical approach

Raphael Rehms, Nicole Ellenbach Veronika DeffnerSabine Hoffmann · 2025

Read Addressing complex structures of measurement error arising in the exposure assessment in occupational epidemiology using a Bayesian hierarchical approach by Raphael Rehms, Nicole Ellenbach Veronika DeffnerSabine Hoffmann (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-17161.txt.

Introduction

Raphael Rehms and co-authors build a Bayesian model that untangles messy measurement error in workplace exposure studies. Their 2025 paper starts from a hard fact: exposures are often rebuilt years later, for groups of workers, by shifting methods. Past work often assumed simple error shapes.

The layered model allows mixed error types, error forms that change by period or site, and errors linked across time and workers. A test case uses German uranium miners to check how radon exposure links to lung cancer deaths. Simulations show the method corrects bias from very tangled error patterns. The authors also probe how the fix holds up when parts of the model are misspecified.

Catalog record

CREATORRaphael Rehms, Nicole Ellenbach Veronika DeffnerSabine Hoffmann
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE72 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2503-17161

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Addressing complex structures of measurement error arising in the exposure assessment in occupational epidemiology using a Bayesian hierarchical approach Raphael Rehms, Nicole Ellenbach Veronika DeffnerSabine Hoffmann Abstract Abstract Exposure assessment in occupational epidemiology may involve multiple unknown quantities that are measured or reconstructed simultaneously for groups of workers and over several years. Additionally, exposures may be collected using different assessment strategies, depending on the period of exposure. As a consequence, researchers who are analyzing occupational cohort studies are commonly faced with challenging structures of exposure measurement error, involving complex dependence structures and multiple measurement error models, depending on the period of exposure. However, previous work has often made many simplifying assumptions concerning these errors. In this work, we propose a Bayesian hierarchical approach to account for a broad range of error structures arising in occupational epidemiology. The considered error structures may involve several unknown quantities that can be subject to mixtures of Berkson and classical measurement error. It is possible to account for different error structures, depending on the exposure period and the location of a worker. Moreover, errors can present complex dependence structures over time and between workers. We illustrate the proposed hierarchical approach on a subgroup of the German cohort of uranium miners to account for potential exposure uncertainties in the association between radon exposure and lung cancer mortality. The performance of the proposed approach and its sensitivity to model misspecification are evaluated in a simulation study. The results show that biases in estimates arising from very complex measurement errors can be corrected through the proposed Bayesian hierarchical approach. 1 Introduction In occupational epidemiology, researchers are often interested in the association between the cumulative exposure to a specific chemical or physical agent and the time until an event occurs, such as a diagnosis of or death from a certain disease. In this situation, exposure is time-dependent and ongoing, and the exposure history of workers may be collected using different assessment strategies depending on the period of exposure. During the estimation process, measurement errors can arise, leading to complex patterns of exposure uncertainty, where the structure and magnitude of measurement error can vary over time. In many occupational cohort studies, there are no prospective exposure measurements. As it is often infeasible or too costly to measure or to estimate exposure values for individual workers, many occupational cohort studies rely on job-exposure matrices (JEMs). In general, JEMs provide information about exposure levels for certain job categories or titles [ 16 ] . The assigned exposure value may also vary with respect to location and year. Although it is widely acknowledged that exposure measurement error can have deleterious consequences on the validity of statistical inference and may lead to erroneous conclusions, researchers in the field of occupational epidemiology do not often account for these errors [ 10 ] . Many of the publications account for measurement error use Bayesian approaches [ 6 , 49 , 11 , 5 , 26 , 34 ] , while frequentist methods are also employed [ 16 , 35 , 34 , 3 ] . It is often assumed that errors follow a simple structure where deviations of individual exposures for workers from the assigned exposure level in a JEM is described as unshared Berkson error [ 4 , 7 , 30 , 51 ] . However, the estimation of the exposure values in a JEM often involves the estimation of multiple uncertain quantities that may be subject to complex structures of measurement error. Exposure values in a JEM are often reconstructed retrospectively by experts [ 44 ] , leading to measurement errors that may affect several groups of workers and several exposure years at the same time. As a consequence, group-level estimates and individual job conditions may give rise to mixtures of Berkson and classical measurement error with complex dependency structures. While it is difficult to account for complex error structures using available measurement error correction methods, they pose serious threats to the validity of statistical inference in occupational epidemiology. Moreover, it is unlikely that the bias introduced by these complex measurement error structures can be adequately corrected for with methods that assume simple measurement error structures, such as unshared Berkson error. In previous work, we found that uncertainty components shared within workers cause more bias in risk estimation than components of unshared exposure uncertainty and that this can lead to an attenuation of the exposure-response curve for high exposure values [ 25 ] , a phenomenon that is frequently observed in occupational cohort studies [ 24 , 50 , 51 ] . The aim of this paper is to demonstrate that even highly complex measurement error structures, which may commonly arise in occupational cohort studies, can be effectively addressed. We illustrate this by proposing a flexible hierarchical Bayesian framework capable of accounting for these complexities, applied to the German cohort of uranium miners. Section 2 describes measurement error structures that typically arise through the prospective and retrospective exposure assessment in occupational epidemiology and the basic methodology to account for them using a Bayesian hierarchical approach. In section 3, we demonstrate the flexibility of the approach as a proof of concept using the German cohort of uranium miners as illustrative example. We account for highly complex structures of measurement errors that arises when studying the association between radon exposure and lung cancer mortality, presenting preliminary results for a selective subgroup of the Wismut cohort [ 28 , 31 ] . Section 4 presents a simulation study in which we evaluate the performance of the proposed approach and assess its sensitivity to model misspecification. In section 5 we discuss our results and give an outlook for future work. 2 Accounting for measurement error in occupational epidemiology 2.1 Measurement error characteristics in occupational epidemiology When describing measurement error, one commonly distinguishes classical and Berkson error. A classical measurement error model describes the error prone observed value Z as a function of the true value X and of an error term U that is independent of X . For additive and multiplicative error, classical measurement error can be written as Z=X+U and Z=X\cdot U with X\perp\!\!\!\perp U , respectively. Conversely, for Berkson error, the error term U is independent of the observed value Z . Again, we can write X=Z+U and X=Z\cdot U with Z\perp\!\!\!\perp U for additive and multiplicative Berkson error, respectively. A classical measurement error model is often employed to describe the measurement arising from a measurement device or through the estimation of experts whereas a Berkson model describes the situation where the true and unknown value of a quantity of interest deviates from a fixed and observed value. In an occupational cohort study, it is in general plausible to assume that exposure measurement error is non-differential, i.e. that errors are independent of the outcome since it is unlikely that errors arising in the exposure estimation in an occupational cohort depend on the (future) disease status of individual workers. If a JEM is used, the same exposure level is typically assigned to all workers in a job category (potentially also as a function of year and location). As a consequence, measurement errors that arise from the estimation of this common exposure level will affect all workers

Read the complete text