Survival-related Selection Bias in Studies of Racial Health Disparities: The Importance of the Target Population and Study Design

Optimising the use of electronic medical records for large scale research in psychiatry

The explosion and abundance of digital data could facilitate large-scale research for psychiatry and mental health. Research using so-called "real world data"-such as electronic medical/health records-can be resource-efficient, facilitate rapid hypothesis generation and testing, complement existing evidence (e.g. from trials and evidence-synthesis) and may enable a route to translate evidence into clinically effective, outcomes-driven care for patient populations that may be under-represented. However, the interpretation and processing of real-world data sources is complex because the clinically important 'signal' is often contained in both structured and unstructured (narrative or "free-text") data. Techniques for extracting meaningful information (signal) from unstructured text exist and have advanced the re-use of routinely collected clinical data, but these techniques require cautious evaluation. In this paper, we survey the opportunities, risks and progress made in the use of electronic medical record (real-world) data for psychiatric research.

Selection Bias in Health Research: Quantifying, Eliminating, or Exacerbating Health Disparities?

Purpose of review

To summarize recent literature on selection bias in disparities research addressing either descriptive or causal questions, with examples from dementia research.

Recent findings

Defining a clear estimand, including the target population, is essential to assess whether generalizability bias or collider-stratification bias are threats to inferences. Selection bias in disparities research can result from sampling strategies, differential inclusion pipelines, loss to follow-up, and competing events. If competing events occur, several potentially relevant estimands can be estimated under different assumptions, with different interpretations. The apparent magnitude of a disparity can differ substantially based on the chosen estimand. Both randomized and observational studies may misrepresent health disparities or heterogeneity in treatment effects if they are not based on a known sampling scheme.

Conclusion

Researchers have recently made substantial progress in conceptualization and methods related to selection bias. This progress will improve the relevance of both descriptive and causal health disparities research.

Alcohol use and the longitudinal HIV care continuum for people with HIV who enrolled in care between 2011 and 2019

PURPOSE

We described the impact of alcohol use on longitudinal engagement in HIV care including loss to follow-up, durability of viral suppression, and death.

METHODS

We followed a cohort of 1781 people with HIV from enrolled in care at one of seven US clinics, 2011-2019 through 102 months. We used a multistate, time-varying Markov process and restricted mean time to summarize engagement in HIV care over follow-up according to baseline self-reported alcohol use (none, moderate, or unhealthy).

RESULTS

Our sample (86% male, 54% White) had median age of 35 years. Over 102 months, people with no, moderate, and unhealthy alcohol use averaged 62.3, 61.1, and 59.5 months virally suppressed, respectively. People who reported unhealthy or moderate alcohol use spent 5.1 (95% confidence intervals (CI): 0.8, 9.3) and 7.6 (95%CI: 3.1, 11.7) more months lost to care than nondrinkers. Compared to no use, unhealthy alcohol use was associated with 3.4 (95%CI: -5.6, -1.6) fewer months in care, not virally suppressed. There were no statistically significant differences after adjustment for demographic and clinical characteristics.

CONCLUSIONS

Moderate or unhealthy drinking at enrollment in HIV care was associated with poor retention in care. Alcohol use was not associated with time spent virally suppressed.

Considering Questions Before Methods in Dementia Research With Competing Events and Causal Goals

Studying causal exposure effects on dementia is challenging when death is a competing event. Researchers often interpret death as a potential source of bias, although bias cannot be defined or assessed if the causal question is not explicitly specified. Here we discuss 2 possible notions of a causal effect on dementia risk: the "controlled direct effect" and the "total effect." We provide definitions and discuss the "censoring" assumptions needed for identification in either case and their link to familiar statistical methods. We illustrate concepts in a hypothetical randomized trial on smoking cessation in late midlife, and emulate such a trial using observational data from the Rotterdam Study, the Netherlands, 1990-2015. We estimated a total effect of smoking cessation (compared with continued smoking) on 20-year dementia risk of 2.1 (95% confidence interval: -0.1, 4.2) percentage points and a controlled direct effect of smoking cessation on 20-year dementia risk had death been prevented of -2.7 (95% confidence interval: -6.1, 0.8) percentage points. Our study highlights how analyses corresponding to different causal questions can have different results, here with point estimates on opposite sides of the null. Having a clear causal question in view of the competing event and transparent and explicit assumptions are essential to interpreting results and potential bias.

Adverse childhood experiences and adult outcomes using a causal framework perspective: Challenges and opportunities

BACKGROUND

Research on the effect of adverse childhood experiences (ACEs) on adult outcomes has typically relied on retrospective assessment of ACEs and cumulative scores. However, this approach raises methodological challenges that can limit the validity of findings.

OBJECTIVE

The aims of this paper are 1) to present the value of directed acyclic graphs (DAGs) to identify and mitigate potential problems related to confounding and selection bias, and 2) to question the meaning of a cumulative ACE score.

RESULTS

Adjusting for variables that post-date childhood could block mediated pathways that are part of the total causal effect while conditioning on adult variables, which often serve as proxies for childhood variables, can create collider stratification bias. Because exposure to ACEs can affect the likelihood of reaching adulthood or study entry, selection bias could be introduced via restricting selection on a variable affected by ACEs in the presence of unmeasured confounding. In addition to challenges regarding causal structure, using a cumulative score of ACEs assumes that each type of adversity will have the same effect on a given outcome, which is unlikely considering differing risk across adverse experiences.

CONCLUSIONS

DAGs provide a transparent approach of the researchers' assumed causal relationships and can be used to overcome issues related to confounding and selection bias. Researchers should be explicit about their operationalization of ACEs and how it is to be interpreted in the context of the research question they are trying to answer.

Placental epigenetic gestational aging in relation to maternal sociodemographic factors and smoking among infants born extremely preterm: a descriptive study

Social determinants of health (SDoH) are defined as the conditions in which people are born, grow, live, work, and age. The distribution of these conditions is influenced by underlying structural factors and may be linked to adverse pregnancy outcomes through epigenetic modifications of gestational tissues. A promising modification is epigenetic gestational age (eGA), which captures 'biological' age at birth. Measuring eGA in placenta, an organ critical for foetal development, may provide information about how SDoH 'get under the skin' during pregnancy to influence birth outcomes and ethnic/racial disparities. We examined relationships of placental eGA with sociodemographic factors, smoking, and two key clinical outcomes: Apgar scores and NICU length of stay. Using the Robust Placental Clock, we estimated eGA for placental samples from the Extremely Low Gestational Age Newborns cohort (N = 408). Regression modelling revealed smoking during pregnancy was associated with placental eGA acceleration (i.e., eGA higher than chronologic gestational age). This association differed by maternal race: among infants born to mothers racialized as Black, we observed greater eGA acceleration (+0.89 week, 95% CI: 0.38, 1.40) as compared to those racialized as white (+0.27 week, 95% CI: -0.06, 0.59). Placental eGA acceleration was also correlated with shorter NICU lengths of stay, but only among infants born to mothers racialized as Black (-0.08 d/week-eGA, 95% CI: -0.12, -0.05). Together, these observed associations suggest that interpretations of epigenetic gestational aging may be tissue-specific.

Recommendations for Using Causal Diagrams to Study Racial Health Disparities

There have been calls for race to be denounced as a biological variable and for a greater focus on racism, instead of solely race, when studying racial health disparities in the United States. These calls are grounded in extensive scholarship and the rationale that race is not a biological variable, but instead socially constructed, and that structural/institutional racism is a root cause of race-related health disparities. However, there remains a lack of clear guidance for how best to incorporate these assertions about race and racism into tools, such as causal diagrams, that are commonly used by epidemiologists to study population health. We provide clear recommendations for using causal diagrams to study racial health disparities that were informed by these calls. These recommendations consider a health disparity to be a difference in a health outcome that is related to social, environmental, or economic disadvantage. We present simplified causal diagrams to illustrate how to implement our recommendations. These diagrams can be modified based on the health outcome and hypotheses, or for other group-based differences in health also rooted in disadvantage (e.g., gender). Implementing our recommendations may lead to the publication of more rigorous and informative studies of racial health disparities.

Toward a Clearer Definition of Selection Bias When Estimating Causal Effects

Selection bias remains a subject of controversy. Existing definitions of selection bias are ambiguous. To improve communication and the conduct of epidemiologic research focused on estimating causal effects, we propose to unify the various existing definitions of selection bias in the literature by considering any bias away from the true causal effect in the referent population (the population before the selection process), due to selecting the sample from the referent population, as selection bias. Given this unified definition, selection bias can be further categorized into two broad types: type 1 selection bias owing to restricting to one or more level(s) of a collider (or a descendant of a collider) and type 2 selection bias owing to restricting to one or more level(s) of an effect measure modifier. To aid in explaining these two types-which can co-occur-we start by reviewing the concepts of the target population, the study sample, and the analytic sample. Then, we illustrate both types of selection bias using causal diagrams. In addition, we explore the differences between these two types of selection bias, and describe methods to minimize selection bias. Finally, we use an example of "M-bias" to demonstrate the advantage of classifying selection bias into these two types.

Allostatic Load and Mortality: A Systematic Review and Meta-Analysis

INTRODUCTION

Allostatic load, a measure of stress-related physiologic dysregulation, is associated with numerous mortality risk factors. This systematic review and meta-analysis examines the relationship between high allostatic load (i.e., increased dysregulation versus low dysregulation) and mortality (cardiovascular disease and all-cause mortality).

METHODS

Systematic searches of 2 databases conducted in May 2021 yielded 336 unique records; 17 eligible studies (2001-2020) were included.

RESULTS

High allostatic load was associated with increased risk of all-cause mortality across all the 17 individual studies (hazard ratio=1.08-2.75) and in 6 of 8 studies examining cardiovascular disease mortality (hazard ratio=1.19-3.06). Meta-analyses indicated that high allostatic load was associated with increased risk of all-cause mortality, overall (hazard ratio=1.22, 95% CI=1.14, 1.30, n=10) and across subgroups (hazard ratio=1.11-1.41), and similarly for cardiovascular disease mortality (hazard ratio=1.31, 95% CI=1.10, 1.57, n=6). Although studies were generally of good quality (n=13), heterogeneity was high in most pooled estimates (I²>90%).

DISCUSSION

In this review of relatively good-quality studies, high allostatic load was associated with an increased mortality risk of 22% for all-cause mortality and 31% for cardiovascular disease mortality. Thus, allostatic load is an emerging and potent modifiable risk factor for all-cause and cardiovascular disease mortality that shows promise as a prognostic indicator for mortality. The heterogeneity in allostatic load assessment across studies highlights the need for standardized measurement. The findings underscore the importance of allostatic load's dynamic nature, which may be especially relevant for mitigating mortality risk in younger adults. Because older adults are oversampled, future allostatic load research should prioritize younger adults and longitudinal monitoring and specific cardiovascular disease mortality risk associations and individualize behavioral and lifestyle targets for reducing allostatic load.

Clinical profile and outcome of recurrent infective endocarditis

AIMS

Purpose of this study is to compare the clinical course and outcome of patients with recurrent versus first-episode infective endocarditis (IE).

METHODS

Patients with recurrent and first-episode IE enrolled in the EUROpean ENDOcarditis (EURO-ENDO) registry including 156 centres were identified and compared using propensity score matching. Recurrent IE was classified as relapse when IE occurred ≤6 months after a previous episode or reinfection when IE occurred >6 months after the prior episode.

RESULTS

3106 patients were enrolled: 2839 (91.4%) patients with first-episode IE (mean age 59.4 (±18.1); 68.3% male) and 267 (8.6%) patients with recurrent IE (mean age 58.1 (±17.7); 74.9% male). Among patients with recurrent IE, 13.2% were intravenous drug users (IVDUs), 66.4% had a repaired or replaced valve with the tricuspid valve being more frequently involved compared with patients with first-episode IE (20.3% vs 14.1%; p=0.012). In patients with a first episode of IE, the aortic valve was more frequently involved (45.6% vs 39.5%; p=0.061). Recurrent relapse and reinfection were 20.6% and 79.4%, respectively. Staphylococcus aureus was the microorganism most frequently observed in both groups (p=0.207). There were no differences in in-hospital and post-hospitalisation mortality between recurrent and first-episode IE. In patients with recurrent IE, in-hospital mortality was higher in IVDU patients. Independent predictors of poorer in-hospital and 1-year outcome, including the occurrence of cardiogenic and septic shock, valvular disease severity and failure to undertake surgery when indicated, were similar for recurrent and first-episode IE.

CONCLUSIONS

In-hospital and 1-year mortality was similar in patients with recurrent and first-episode IE who shared similar predictors of poor outcome.

Invited Commentary: Counterfactuals in Social Epidemiology-Thinking Outside of "the Box"

There are tensions inherent between many of the social exposures examined within social epidemiology and the assumptions embedded in quantitative potential-outcomes-based causal inference framework. The potential-outcomes framework characteristically requires a well-defined hypothetical intervention. As noted by Galea and Hernán (Am J Epidemiol. 2020;189(3):167-170), for many social exposures, such well-defined hypothetical exposures do not exist or there is no consensus on what they might be. Nevertheless, the quantitative potential-outcomes framework can still be useful for the study of some of these social exposures by creative adaptations that 1) redefine the exposure, 2) separate the exposure from the hypothetical intervention, or 3) allow for a distribution of hypothetical interventions. These various approaches and adaptations are reviewed and discussed. However, even these approaches have their limits. For certain important historical and social determinants of health such as social movements or wars, the quantitative potential-outcomes framework with well-defined hypothetical interventions is the wrong tool. Other modes of inquiry are needed.

Invited Commentary: Making Causal Inference More Social and (Social) Epidemiology More Causal

A society's social structure and the interactions of its members determine when key drivers of health occur, for how long they last, and how they operate. Yet, it has been unclear whether causal inference methods can help us find meaningful interventions on these fundamental social drivers of health. Galea and Hernán propose we place hypothetical interventions on a spectrum and estimate their effects by emulating trials, either through individual-level data analysis or systems science modeling (Am J Epidemiol. 2020;189(3):167-170). In this commentary, by way of example in health disparities research, we probe this "closer engagement of social epidemiology with formal causal inference approaches." The formidable, but not insurmountable, tensions call for causal reasoning and effect estimation in social epidemiology that should always be enveloped by a thorough understanding of how systems and the social exposome shape risk factor and health distributions. We argue that one way toward progress is a true partnership of social epidemiology and causal inference with bilateral feedback aimed at integrating social epidemiologic theory, causal identification and modeling methods, systems thinking, and improved study design and data. To produce consequential work, we must make social epidemiology more causal and causal inference more social.

Commentary: Causal Inference for Social Exposures

Social epidemiology seeks to describe and quantify the causal effects of social institutions, interactions, and structures on human health. To accomplish this task, we define exposures as treatments and posit populations exposed or unexposed to these well-defined regimens. This inferential structure allows us to unambiguously estimate and interpret quantitative causal parameters and to investigate how these may be affected by biases such as confounding. This paradigm has been challenged recently by some critics who favor broadening the exposures that may be studied beyond treatments to also consider states. Defining the exposure protocol of an observational study is a continuum of specificity, and one may choose to loosen this definition, incurring the cost of causal parameters that become commensurately more vague. The advantages and disadvantages of broader versus narrower definitions of exposure are matters of continuing debate in social epidemiology as in other branches of epidemiology.