How Can Quantitative Analysis Be Used to Improve Occupational Health without Reinforcing Social Inequalities? An Examination of Statistical Methods

Abstract

Taking account of sex and gender in occupational health studies poses statistical challenges. Other sociodemographic variables, such as racialization, class, and age, also affect the relations between workplace exposures and health and interact with sex and gender. Our objective was to perform a critical review of conventional and emerging statistical tools, examining whether each analysis takes account of sociodemographic variables (1) in a way that contributes to identification of critical occupational determinants of health (2) while taking account of relevant population characteristics to reflect intersectional approaches to health and (3) using sample sizes and population characteristics available to researchers. A two-step search was conducted: (1) a scientific watch concerning the statistical tools most commonly used in occupational health over the past 20 years; (2) a screening of the 1980–2022 literature with a focus on emerging tools. Our examination shows that regressions with adjustment for confounders and stratification fail to reveal the sociodemographic mechanisms that interact with occupational health problems, endangering the identification of occupational risks. Multilevel (notably MAIHDA) analyses, decision tree, cluster, and latent analyses are useful methods to consider when seeking to orientate prevention. Researchers should consider methods that adequately reveal the mechanisms connecting sociodemographic variables and occupational health outcomes.