Causal Questions Research Articles

Measurement error and information bias in causal diagrams: mapping epidemiological concepts and graphical structures.

Measurement error and information bias are ubiquitous in epidemiology, yet directed acyclic graphs (DAGs) are infrequently used to represent them, in contrast with confounding and selection bias. This represents a missed opportunity to leverage the full utility of DAGs to depict associations between the variables we actually analyse in practice: empirically measured variables, which are necessarily measured with error. In this article, we focus on applying causal diagrams to depict the data-generating mechanisms that give rise to the data we analyse, including measurement error. We begin by considering empirical data considerations using a general example, and then build up to a specific worked example from the clinical epidemiology of hearing health. Throughout, our goal is to highlight both the challenges and the benefits of using DAGs to depict measurement error. In addition to the application of DAGs to conceptual causal questions (which pertain to unmeasured constructs free from measurement error), which is common, we highlight the advantages associated with applying DAGs to also include empirically measured variables and-potentially-information bias. We also highlight the implications implied by this use of DAGs, particularly regarding the unblocked backdoor path causal structure. Ultimately, we seek to help increase the clarity with which epidemiologists can map traditional epidemiological concepts (such as information bias and confounding) onto causal graphical structures.

Sexual Harassment, Sexual Violence, and Mental Health Outcomes: Causal Inference with Ambiguous Exposures.

Social exposures and their impact on mental health has proven hard to capture, partly owing to the complex and multifaceted nature of social reality. Sexual harassment and sexual violence (SHV) are no exceptions. SHV can be conceptualized as a continuum of negative sexual experiences whose severity vary depending on multiple determinants. Further, SHV can be conceptualized as either discrete events or as a generally hostile sexual environment represented by latent variables. With any of these conceptualizations, SHV constitutes a broad construct containing many kinds of negative experiences. This ambiguity poses challenges for determining the mental health consequences, as different forms of SHV may vary in terms of their mental health impact. We discuss different conceptualizations of SHV in relation to mental health outcomes through the lens of the potential outcomes framework, with a focus on the consistency condition. The multiple versions of treatment theory is presented to show how to provide formal interpretations of causal estimates under ambiguous exposures. Lastly, we provide suggestions on how the increase the clarity and interpretability of the effects of SHV on mental health, by increasing the precision of the causal questions and the use of more specific definitions of SHV.

Methods in Causal Inference Part 2: Interaction, Mediation, and Time-Varying Treatments

Abstract The analysis of ‘moderation’, ‘interaction’, ‘mediation’, and ‘longitudinal growth’ is widespread in the human sciences, yet subject to confusion. To clarify these concepts, it is essential to state causal estimands, which requires specifying counterfactual contrasts for a target population on an appropriate scale. Once causal estimands are defined, we must consider their identification. I employ causal directed acyclic graphs (causal DAGs) and Single World Intervention Graphs to elucidate identification workflows. I show that when multiple treatments exist, common methods for statistical inference, such as multi-level regressions and statistical structural equation models, cannot typically recover the causal quantities we seek. By properly framing and addressing causal questions of interaction, mediation, and time-varying treatments, we can expose the limitations of popular methods and guide researchers to a clearer understanding of the phenomena that animate our interests.

Burnout and moral injuries after foreign deployment among medical personnel of the German armed forces: a pre-post study.

Given a high amount of workplace stressors, burnout syndrome, as a depression-related syndrome, is highly relevant for medical service soldiers. This study aims to examine their effects with regard to moral injuries and personal values following foreign deployment. This longitudinal study included 91 soldiers of the German Armed Forces Medical Service. Participants completed the Maslach Burnout Inventory (MBI) and the Portrait-Value-Questionnaire (PVQ) before and after a foreign deployment as well as the Moral Injury Scale (SMBE) after deployment. Analysis has been conducted using t-tests to assess potential changes in MBI and PVQ scales between pre-test - t1 (2-4 weeks before deployment) and post-test - t2 (up to 6 months after deployment). In addition, correlations were examined between moral injuries (MI) after deployment and MBI scores at t1 and t2 as well as between personal values (PVQ t1) and MBI scores at t1 and t2. The MBI subscales showed mild to moderate burnout symptoms at both pre- and post-tests, with a slight deterioration during the study period, albeit not significant. There were no significant mean differences in PVQ between measurement points. Nevertheless, PVQ self-direction and tradition at t1 correlated negatively with MBI INV at t2 (PVQ SD r = -.21, p = .043) and MBI PA at t2 (PVQ TR r = -.23, p = .027). Furthermore, the subscale PVQ power at t1 correlated positively with MBI PA at t2 (PVQ PO r = .28, p = .006), meanwhile PVQ universalism at t1 correlated positively with MBI INV at t1 (PVQ UN r = .25, p = .018). Furthermore, positive correlations were found between moral injuries at t2 (SMBE total score, SMBE_Sub1, SMBE_Sub2) and MBI subscales Emotional Exhaustion (EE; r = -.54, p = .001), Depersonalization (DP; r = .38, p = .001), and Involvement (INV; r = .30, p = .004) before and after the deployment period. No correlation was found between MI and MBI subscale Personal Accomplishment (PA). The results indicate that medical service soldiers exhibit mild to moderate burnout symptoms even before deployment. Significant associations between moral injuries and burnout were found in 3 out of 4 MBI subscales (EE, DP, INV). There was a significant association with a stronger moral injury and higher burnout levels, persisting both before and after the study period. Furthermore, our results suggest that personal value orientations might be meaningful predictors of burnout. Hence, causal questions regarding general work stress among medical service soldiers should be further explored in more detailed studies. Further research could lay the foundation for future approaches in psychotherapy as well as primary and secondary prevention in this field.

Causal analyses in longitudinal observational studies in oral health: A scoping review.

Causal analysis including causal inference and causal mediation is pivotal to inform effective interventions. In modern epidemilogy, causal analysis involves four key steps: formulating causal questions, employing directed acyclic graphs (DAGs), conducting data management and selecting statistical strategies. Our objective was to conduct a scoping review to assess how longitudinal observational studies (LOSs) in dental field have integrated these four steps to contribute leverage evidence that inform oral public health interventions. LOSs focusing on determinants of dental caries published from 2012 to 2024 were systematically retrieved from five major databases. The Joanna Briggs Institute-scoping review guidance and the Covidence application were employed to identify eligible LOSs for being reviewed. Out of the 85 eligible LOSs, none formulated causal hypothesis by applying 'what if' question or investigated mediation across three levels of the determinants of oral health. A minority (18 studies, ~21.2%) employed DAGs to visualise relationships among study variables, while only one third (33 studies, ~39%) clearly defined confounders. The majority (64 studies, ~75%) incorporated a time-varying feature of their data, yet only a few (11 studies) fully leveraged this advanced aspect. Among these studies that fully utilised time-varying data, more than half encountered challenges in employing robust statistics to address confounders arising from such data dynamics. Dental LOSs have, to date, mostly focused on investigating associations over causality, often neglecting the four-step causal analysis and not fully utilising time-varying data. Researchers necessitate to shift their focus to causal inference and prioritise building capacity in causal analysis with a consistent four-step approach to advance the field. Studies exploring mechanisms linking determinants of dental caries across levels and leveraging time-varying data are strongly encouraged.

Gaps in the usage and reporting of multiple imputation for incomplete data: findings from a scoping review of observational studies addressing causal questions

BackgroundMissing data are common in observational studies and often occur in several of the variables required when estimating a causal effect, i.e. the exposure, outcome and/or variables used to control for confounding. Analyses involving multiple incomplete variables are not as straightforward as analyses with a single incomplete variable. For example, in the context of multivariable missingness, the standard missing data assumptions (“missing completely at random”, “missing at random” [MAR], “missing not at random”) are difficult to interpret and assess. It is not clear how the complexities that arise due to multivariable missingness are being addressed in practice. The aim of this study was to review how missing data are managed and reported in observational studies that use multiple imputation (MI) for causal effect estimation, with a particular focus on missing data summaries, missing data assumptions, primary and sensitivity analyses, and MI implementation.MethodsWe searched five top general epidemiology journals for observational studies that aimed to answer a causal research question and used MI, published between January 2019 and December 2021. Article screening and data extraction were performed systematically.ResultsOf the 130 studies included in this review, 108 (83%) derived an analysis sample by excluding individuals with missing data in specific variables (e.g., outcome) and 114 (88%) had multivariable missingness within the analysis sample. Forty-four (34%) studies provided a statement about missing data assumptions, 35 of which stated the MAR assumption, but only 11/44 (25%) studies provided a justification for these assumptions. The number of imputations, MI method and MI software were generally well-reported (71%, 75% and 88% of studies, respectively), while aspects of the imputation model specification were not clear for more than half of the studies. A secondary analysis that used a different approach to handle the missing data was conducted in 69/130 (53%) studies. Of these 69 studies, 68 (99%) lacked a clear justification for the secondary analysis.ConclusionEffort is needed to clarify the rationale for and improve the reporting of MI for estimation of causal effects from observational data. We encourage greater transparency in making and reporting analytical decisions related to missing data.

The Causal Roadmap and Simulations to Improve the Rigor and Reproducibility of Real-data Applications.

The Causal Roadmap outlines a systematic approach to asking and answering questions of cause and effect: define the quantity of interest, evaluate needed assumptions, conduct statistical estimation, and carefully interpret results. To protect research integrity, it is essential that the algorithm for statistical estimation and inference be prespecified prior to conducting any effectiveness analyses. However, it is often unclear which algorithm will perform optimally for the real-data application. Instead, there is a temptation to simply implement one's favorite algorithm, recycling prior code or relying on the default settings of a computing package. Here, we call for the use of simulations that realistically reflect the application, including key characteristics such as strong confounding and dependent or missing outcomes, to objectively compare candidate estimators and facilitate full specification of the statistical analysis plan. Such simulations are informed by the Causal Roadmap and conducted after data collection but prior to effect estimation. We illustrate with two worked examples. First, in an observational longitudinal study, we use outcome-blind simulations to inform nuisance parameter estimation and variance estimation for longitudinal targeted minimum loss-based estimation. Second, in a cluster randomized trial with missing outcomes, we use treatment-blind simulations to examine type-I error control in two-stage targeted minimum loss-based estimation. In both examples, realistic simulations empower us to prespecify an estimation approach that is expected to have strong finite sample performance and also yield quality-controlled computing code for the actual analysis. Together, this process helps to improve the rigor and reproducibility of our research.

Identification and estimation of causal effects in the presence of confounded principal strata.

Principal stratification has become a popular tool to address a broad class of causal inference questions, particularly in dealing with non-compliance and truncation by death problems. The causal effects within principal strata, which are determined by joint potential values of the intermediate variable, also known as the principal causal effects, are often of interest in these studies. The analysis of principal causal effects from observational studies mostly relies on the ignorability assumption of treatment assignment, which requires practitioners to accurately measure as many covariates as possible so that all potential sources of confounders are captured. However, in practice, collecting all potential confounding factors can be challenging and costly, rendering the ignorability assumption questionable. In this paper, we consider the identification and estimation of causal effects when treatment and principal stratification are confounded by unmeasured confounding. Specifically, we establish the nonparametric identification of principal causal effects using a pair of negative controls to mitigate unmeasured confounding, requiring they have no direct effect on the outcome variable. We also provide an estimation method for principal causal effects. Extensive simulations and a leukemia study are employed for illustration.

Bidirectional causal associations between aging and major mental disorders: A population-based study using the two-sample mendelian randomization method from the UK biobank (AM-SRNMA 002)

AimsWhile observational studies have suggested associations linking aging and mental disorders, the question of causality has remained unclear. This study aimed to explore the causal relationship between aging level and major mental disorders. MethodsWe utilized Two-Sample Mendelian randomization (2SMR) with mental disorders data and aging indicators information from an extensive genome-wide association study (GWAS) database. The GWAS database is a comprehensive resource that compiles genetic association data, encompassing a sample size of over 450,000 individuals. We employed five methods for 2SMR and single nucleotide polymorphisms were chosen as instrumental variables. ResultsOur analyses consistently supported a bidirectional causal association between the Frailty Index (FI) and Major Depressive Disorder (MDD). Furthermore, our findings indicated potential influences, such as Attention-Deficit/Hyperactivity Disorder (ADHD) and Bipolar Disorder (BD) affecting GrimAge, and Anxiety Disorder (AD) impacting Left Hand Grip Strength (LHGS). In contrast, we observed no significant correlations for other mental disorders on FI, Telomere Length (TL), GrimAge, Appendicular Lean Mass (ALM), and LHGS. In the reverse direction, FI showed a significant impact on the risk of MDD, AD, and ADHD, while LHGS affected the risk of MDD. Importantly, no significant associations were found between other factors and the risk of MDD, BD, AD, Schizophrenia (SZ), and ADHD. ConclusionsThis 2SMR analysis has presented evidence for a bidirectional causal relationship between FI and MDD, while the relationship between ADHD, BD, and GrimAge should be more considered. Our study provides genetic evidence supporting a causal link between aging indicators and several mental illnesses.

Ratios in regression analyses with causal questions.

Ratios in regression analyses with causal questions.

Comparison of caffeine consumption behavior with plasma caffeine levels as exposure measures in drug-target Mendelian randomization.

Mendelian randomization is an epidemiological technique that can explore the potential effect of perturbing a pharmacological target. Plasma caffeine levels can be used as a biomarker to measure the pharmacological effects of caffeine. Alternatively, this can be assessed using a behavioral proxy, such as average number of caffeinated drinks consumed per day. Either variable can be used as the exposure in a Mendelian randomization investigation, and to select which genetic variants to use as instrumental variables. Another possibility is to choose variants in gene regions with known biological relevance to caffeine level regulation. These choices affect the causal question that is being addressed by the analysis, and the validity of the analysis assumptions. Further, even when using the same genetic variants, the sign of Mendelian randomization estimates (positive or negative) can change depending on the choice of exposure. Some genetic variants that decrease caffeine metabolism associate with higher levels of plasma caffeine, but lower levels of caffeine consumption, as individuals with these variants require less caffeine consumption for the same physiological effect. We explore Mendelian randomization estimates for the effect of caffeine on body mass index, and discuss implications for variant and exposure choice in drug target Mendelian randomization investigations.

A step-by-step guide to causal study design using real-world data

AbstractDue to the need for generalizable and rapidly delivered evidence to inform healthcare decision-making, real-world data have grown increasingly important to answer causal questions. However, causal inference using observational data poses numerous challenges, and relevant methodological literature is vast. We endeavored to identify underlying unifying themes of causal inference using real-world healthcare data and connect them into a single schema to aid in observational study design, and to demonstrate this schema using a previously published research example. A multidisciplinary team (epidemiology, biostatistics, health economics) reviewed the literature related to causal inference and observational data to identify key concepts. A visual guide to causal study design was developed to concisely and clearly illustrate how the concepts are conceptually related to one another. A case study was selected to demonstrate an application of the guide. An eight-step guide to causal study design was created, integrating essential concepts from the literature, anchored into conceptual groupings according to natural steps in the study design process. The steps include defining the causal research question and the estimand; creating a directed acyclic graph; identifying biases and design and analytic techniques to mitigate their effect, and techniques to examine the robustness of findings. The cardiovascular case study demonstrates the applicability of the steps to developing a research plan. This paper used an existing study to demonstrate the relevance of the guide. We encourage researchers to incorporate this guide at the study design stage in order to elevate the quality of future real-world evidence.

Bruxism, temporomandibular disorders, and headache: a narrative review of correlations and causalities.

The co-occurrence of bruxism, temporomandibular disorders (TMDs), and headache is common in patients. However, there is conflicting evidence regarding whether this association is simply a result of their high prevalence or whether there are indeed causal relationships. This review provides an overview of the current state of research while taking into account the controversies surrounding research methods, particularly in definitions and diagnostic standards. Bruxism-defined as repetitive jaw muscle activity during sleep or wakefulness-is not a painful disorder but may-particularly in co-occurrence with TMD-worsen pre-existing headache. It seems important to differentiate between sleep and awake bruxism because of different impact on pathophysiological processes in different primary headache disorders such as migraine and tension-type headache. Temporomandibular disorder is a heterogenous entity with both myofascial and arthrogenous types of pain in addition to nonpainful disorders. Research suggests a correlation between TMD pain and migraine, as well as between awake bruxism and tension-type headache. However, psychosocial factors may act as confounders in these relationships. Determining causality is challenging because of the limited number of experimental and clinical studies conducted on this topic. The main finding is an apparent lack of consensus on the definition and assessment criteria for bruxism. Treatment wise, it is important to differentiate all 3 conditions because treatment of one condition may have an effect on the other 2 without proving causality. For future research, it is crucial to establish greater consistency and applicability in diagnostic procedures and definitions. In addition, more experimental and clinical studies investigating the question of causality are needed.

Estimating the causal effects of income on health: how researchers’ definitions of “income” matter

BackgroundThere is a well-established cross-sectional association between income and health, but estimates of the causal effects of income vary substantially. Different definitions of income may lead to substantially different empirical results, yet research is often framed as investigating “the effect of income” as if it were a single, easily definable construct.Methods/ResultsThe aim of this paper is to introduce a taxonomy for definitional and conceptual issues in studying individual- or household-level income for health research. We focus on (1) the definition of the income measure (earned and unearned; net, gross, and disposable; real and nominal; individual and household; relative and absolute income) and (2) the definition of the causal contrast (amount, functional form assumptions/transformations, direction, duration of change, and timing of exposure and follow-up). We illustrate the application of the taxonomy to four examples from the published literature.ConclusionsQuantified estimates of causal effects of income on health and wellbeing have crucial relevance for policymakers to anticipate the consequences of policies targeting the social determinants of health. However, much prior evidence has been limited by lack of clarity in distinguishing between different causal questions. The present framework can help researchers explicitly and precisely articulate income-related exposures and causal questions.

#271 The nexus between TMA/TTP, MIS-C, and SLE

Abstract Background and Aims The long-term complications of COVID-19 infection & Multisystem inflammatory syndrome in children (MIS-C) on immune system are still unknown, however there're reports on development of autoimmune disease, highlighting their immune-related nature. MIS-C can induce Thrombotic Microangiopathy (TMA), by inducing endothelial dysfunction, platelet activation, & microthrombi formation, whereas SLE induces TMA/Thrombotic thrombocytopenic purpura (TTP), with ADAMTS13 activity < 10%, or 2ry TMA with ADAMTS13 activity ≥ 10%, where SLE/TTP carries a bad prognosis. Method Our case is one of its kind, where a 12-year- old girl, who was previously infected with SARS-CoV-2, presented with new-onset SLE, associated MIS-C & 2ry TMA/TTP. She presented with convulsions, oliguria, heart failure, impaired kidney function, microangiopathic hemolytic anemia, pancytopenia, highly elevated LDH, & was diagnosed as TMA. Results COVID serology was +ve for IgG, & -ve for IgM, & -ve SARS-CoV-2 PCR, a diagnosis of MIS-C was postulated. Further investigations revealed low C3, C4, +ve ANA, anti-ds DNA & anti-phospholipid Antibodies. Brain imaging showed left cerebellar calcification. Echocardiography showed biventricular dysfunction with EF 30%, where the diagnosis of SLE according to Euler/ACR criteria was confirmed, with 2ry TMA. Emergency hemodialysis, PRBCs transfusion, pulse steroid with total plasma exchange (TPE), were initiated without marked improvement. Rituximab was started on 4th day, but she had generalized convulsion, arrythmia & sudden death occurred. Results of ADAMTS13 activity was < 1%, with +ve autoantibodies, so diagnosis of TMA/TTP was confirmed. Conclusion TMA are a diverse group of disorders that causes microvasculature thrombosis, which is inherited, or acquired 2ry to SLE or COVID-19 infection, where high index of suspicions should be raised for early diagnosis and prompt management, meanwhile COVID-19 infection can induce the development of autoimmune diseases. Herein we raise the question of cause of TMA/TTP, was it 2ry only SLE, or to the previous COVID-19 infection & MIS-C, and if this SLE itself was 2ry also to the previous COVID infection, or all it was by chance?

Causal Event Graph-Guided Language-based Spatiotemporal Question Answering

Large Language Models have excelled at encoding and leveraging language patterns in large text-based corpora for various tasks, including spatiotemporal event-based question answering (QA). However, due to encoding a text-based projection of the world, they have also been shown to lack a full bodied understanding of such events, e.g., a sense of intuitive physics, and cause-and-effect relationships among events. In this work, we propose using causal event graphs (CEGs) to enhance language understanding of spatiotemporal events in language models, using a novel approach that also provides proofs for the model’s capture of the CEGs. A CEG consists of events denoted by nodes, and edges that denote cause and effect relationships among the events. We perform experimentation and evaluation of our approach for benchmark spatiotemporal QA tasks and show effective performance, both quantitative and qualitative, over state-of-the-art baseline methods.

Generalizing toward Nonrespondents: Effect Estimates in Survey Experiments Are Broadly Similar for Eager and Reluctant Participants

Abstract Survey experiments on probability samples are a popular method for investigating population-level causal questions due to their strong internal validity. However, lower survey response rates and an increased reliance on online convenience samples raise questions about the generalizability of survey experiments. We examine this concern using data from a collection of 50 survey experiments which represent a wide range of social science studies. Recruitment for these studies employed a unique double sampling strategy that first obtains a sample of “eager” respondents and then employs much more aggressive recruitment methods with the goal of adding “reluctant” respondents to the sample in a second sampling wave. This approach substantially increases the number of reluctant respondents who participate and also allows for straightforward categorization of eager and reluctant survey respondents within each sample. We find no evidence that treatment effects for eager and reluctant respondents differ substantially. Within demographic categories often used for weighting surveys, there is also little evidence of response heterogeneity between eager and reluctant respondents. Our results suggest that social science findings based on survey experiments, even in the modern era of very low response rates, provide reasonable estimates of population average treatment effects among a deeper pool of survey respondents in a wide range of settings.

Causal Inference About the Effects of Interventions From Observational Studies in Medical Journals.

Many medical journals, including JAMA, restrict the use of causal language to the reporting of randomized clinical trials. Although well-conducted randomized clinical trials remain the preferred approach for answering causal questions, methods for observational studies have advanced such that causal interpretations of the results of well-conducted observational studies may be possible when strong assumptions hold. Furthermore, observational studies may be the only practical source of information for answering some questions about the causal effects of medical or policy interventions, can support the study of interventions in populations and settings that reflect practice, and can help identify interventions for further experimental investigation. Identifying opportunities for the appropriate use of causal language when describing observational studies is important for communication in medical journals. A structured approach to whether and how causal language may be used when describing observational studies would enhance the communication of research goals, support the assessment of assumptions and design and analytic choices, and allow for more clear and accurate interpretation of results. Building on the extensive literature on causal inference across diverse disciplines, we suggest a framework for observational studies that aim to provide evidence about the causal effects of interventions based on 6 core questions: what is the causal question; what quantity would, if known, answer the causal question; what is the study design; what causal assumptions are being made; how can the observed data be used to answer the causal question in principle and in practice; and is a causal interpretation of the analyses tenable? Adoption of the proposed framework to identify when causal interpretation is appropriate in observational studies promises to facilitate better communication between authors, reviewers, editors, and readers. Practical implementation will require cooperation between editors, authors, and reviewers to operationalize the framework and evaluate its effect on the reporting of empirical research.

Harmful compared to what? The problem of gaming and ambiguous causal questions.

There has been much concern regarding potential harmful effects of video game-play in the past 40 years, but limited progress in understanding its causal role. This paper discusses the basic requirements for identifying causal effects of video game-play and argues that most research to date has focused upon ambiguous causal questions. Video games and mental health are discussed from the perspective of causal inference with compound exposures; that is, exposures with multiple relevant variants that affect outcomes in different ways. Not only does exposure to video games encompass multiple different factors, but also not playing video games is equally ambiguous. Estimating causal effects of a compound exposure introduces the additional challenge of exposure-version confounding. Without a comparison of well-defined interventions, research investigating the effects of video game-play will be difficult to translate into actionable health interventions. Interventions that target games should be compared with other interventions aimed at improving the same outcomes.

Reliable Estimation of Causal Effects Using Predictive Models

In recent years, machine learning algorithms have been widely adopted across many fields due to their efficiency and versatility. However, the complexity of predictive models has led to a lack of interpretability in automatic decision-making. Recent works have improved general interpretability by estimating the contributions of input features to the predictions of a pre-trained model. Drawing on these improvements, practitioners seek to gain causal insights into the underlying data-generating mechanisms. To this end, works have attempted to integrate causal knowledge into interpretability, as non-causal techniques can lead to paradoxical explanations. In this paper, we argue that each question about a causal effect requires its own reasoning and that relying on an initial predictive model trained on an arbitrary set of variables may result in quantification problems when estimating all possible effects. As an alternative, we advocate for a query-driven methodology that addresses each causal question separately. Assuming that the causal structure relating the variables is known, we propose to employ the tools of causal inference to quantify a particular effect as a formula involving observable probabilities. We then derive conditions on the selection of variables to train a predictive model that is tailored for the causal question of interest. Finally, we identify suitable eXplainable AI (XAI) techniques to estimate causal effects from the model predictions. Furthermore, we introduce a novel method for estimating direct effects through intervention on causal mechanisms.