Abstract Purpose of Review We compare natural effects and separable effects under nonparametric structural equation models with independent errors, highlighting their similarities and differences. By examining their required properties and sufficient conditions for identification, we aim to provide deeper insights into mediation analysis. Recent Findings If certain assumptions about confounding, positivity, and consistency are met, we can identify natural direct and indirect effects under nonparametric structural equation models with independent errors. However, these effects have been criticized because they rely on a specific cross-world quantity, and the so-called cross-world independence assumption cannot be empirically verified. Furthermore, interventions on the mediator may sometimes be challenging to even conceive. As an alternative approach, separable effects have recently been proposed and applied in mediation analysis, often under finest fully randomized causally interpretable structured tree graph models. These effects are defined without relying on any cross-world quantities and are claimed to be identifiable under assumptions that are testable in principle, thereby addressing some of the challenges associated with natural direct and indirect effects. Summary To conduct meaningful mediation analysis, it is crucial to clearly define the research question of interest, and the choice of methods should align with the nature of the question and the assumptions researchers are willing to make. Examining the underlying philosophical perspectives on causation and manipulation can provide valuable insights.

PurposeElectronic health record (EHR) data have become essential and commonplace in epidemiological and clinical research. In this narrative review on the use of EHR data in epidemiology, I discuss appropriate research questions, common biases, and potential sensitivity analyses focusing on recent work that has been done to improve the internal and external validity of EHR-based studies.Recent FindingsAn appropriate research question addresses issues of EHR-data availability and accessibility, while patient selection forces into healthcare may result in a sample that lacks representativeness. Natural language processing tools are becoming widespread and tailored to EHR use for operationalizing unstructured data. Common biases identified in the literature include misclassification and measurement error, informed presence bias, selection bias and sampling error, and residual confounding. SummaryEHR data are unlike other observational data sources and carry assumptions about patient selection and clinical documentation that can impact the validity of the analyses. Potential sensitivity analyses including quantitative bias analysis can help to understand the impact of one or more of these biases on the study findings.

This review summarizes recent developments in causal decomposition analysis (CDA), a modeling framework for reducing disparities. Rather than describing the current or past drivers of a disparity, CDA estimates the effect of an intervention to change the distribution of a variable or set of variables that are distributed differently or have different effects between groups. Furthermore, CDA clarifies how, through covariate adjustment, ethics and justice are implicit in any definition of disparity and may be incorporated into an intervention. CDA has been applied to disparities in health, sociology, education, and computer science. The CDA framework consists of four steps: formulating a meaningful estimand, articulating identification assumptions to link an appropriate dataset with the estimand, choosing an appropriate estimator, and conducting statistical inference. Estimators have been developed for various types of data and to address particular statistical challenges. However, some estimators adjust for all available covariates in all parts of the model, without discussing ethical implications. Meanwhile, the literature has covered some but not all potential violations of standard CDA modeling assumptions. CDA builds on previous methods for studying disparities by articulating causal estimands that transparently reflect implicit value judgements about health disparities. This review outlines the broad framework of CDA methodology, selected implementations, practical considerations, and current limitations and alternatives.

Purpose of ReviewDespite cervical cancer (CC) being a cancer that can be eliminated, CC disparities persist such that minoritized populations shoulder a disproportionate mortality burden. This may reflect upstream, fundamental drivers of health that impede equitable access to prevention, screening, early detection, and treatment among some groups. This systematic review summarizes evidence on the relationships between structural racism and CC care across the continuum.Recent FindingsFollowing PRISMA guidelines, we conducted a comprehensive search for peer-reviewed, English-language studies relevant to our research question that were published from 2012–2022 using PubMed, CINAHL, Web of Science, and Embase. Of 8,924 articles identified, 4,383 duplicates were removed, and 4,541 underwent screening, with 206 articles meeting eligibility criteria for inclusion in our data synthesis. Among reviewed studies, 60.2% (n = 124) compared CC outcomes by race and ethnicity, often as proxies for upstream racism. Key findings included evidence of lower CC screening rates among Asian American and Pacific Islander women and higher rates among Black and Hispanic/Latinx women. Barriers to healthcare access and socioeconomic status (SES) factors contributed to delayed follow-up, later-stage CC diagnoses, and poorer outcomes, particularly for Black and Hispanic/Latinx women and those residing in low-SES neighborhoods.SummaryThis review underscores associations between race, ethnicity, SES, and outcomes across the CC continuum. Most studies examined racial and ethnic disparities in the outcomes of interest rather than directly evaluating measures of structural racism. Future research should refine measures of structural racism to deepen our understanding of its impact on CC across the care continuum.

Purpose of ReviewRecent advances in Artificial Intelligence (AI) present new and not widely recognized opportunities to advance the rigor, scope, efficiency, and impact of epidemiologic research aiming to make causal inferences or causal decisions. We describe recent developments, challenges, and examples for integrating varied AI tools into the steps of Petersen and van der Laan’s causal inference roadmap and causal decision-making tasks.Recent FindingsAI tools relevant to causal research in epidemiology include predictive models, unsupervised learning, causal structure learning, causal estimation, and generative models. Opportunities exist to integrate AI at each stage of the causal roadmap. This includes the use of generative models to synthesize scientific literature and identify knowledge gaps; causal structure learning to discover or hypothesize causal structures from data; unsupervised learning from unstructured text to generate quantitative variables for analysis; predictive models to drive clinical or policy interventions; generative or causal models to assess or establish identifiability; causal models for estimating statistical parameters; and generative models to create text, tables, and figures to interpret and disseminate findings. Researchers must be mindful of potential pitfalls of AI tools such as insufficient training data, poor accuracy, biases, and ethical and legal concerns.SummaryDiverse AI tools are available to support causal research in epidemiology. Steps of the causal inference roadmap cannot yet be fully automated, but thoughtful “collaboration” between investigators and AI tools may accelerate or deepen the research at each step.