Causal Inference Studies Research Articles

A synthesis of pathways linking diet, metabolic risk and cardiovascular disease: a framework to guide further research and approaches to evidence-based practice

Cardiovascular disease (CVD) is the most common non-communicable disease occurring globally. Although previous literature has provided useful insights into the important role that diet plays in CVD prevention and treatment, understanding the causal role of diets is a difficult task considering inherent and introduced weaknesses of observational (e.g. not properly addressing confounders and mediators) and experimental research designs (e.g. not appropriate or well designed). In this narrative review, we organised current evidence linking diet, as well as conventional and emerging physiological risk factors, with CVD risk, incidence and mortality in a series of diagrams. The diagrams presented can aid causal inference studies as they provide a visual representation of the types of studies underlying the associations between potential risk markers/factors for CVD. This may facilitate the selection of variables to be considered and the creation of analytical models. Evidence depicted in the diagrams was systematically collected from studies included in the British Nutrition Task Force report on diet and CVD and database searches, including Medline and Embase. Although several markers and disorders linked to conventional and emerging risk factors for CVD were identified, the causal link between many remains unknown. There is a need to address the multifactorial nature of CVD and the complex interplay between conventional and emerging risk factors with natural and built environments, while bringing the life course into the spotlight.

Causal inference methods and its application in the study of health effects of air pollution

The adverse health effects of air pollution remains a daunting public health problem globally. The research of the health effects of air pollution provides important evidence for ambient air quality standard establishments and air pollution interventions. In recent years, causal inference has been gradually introduced into the observational study of environmental epidemiology, which provides more statistical method options for the study of causal relationships between air pollution and population health effects. Controlling confounders in observational studies is a major challenge for causal inference. This study introduces the causal inference methods for the identification and control of confounding factors currently used in the study of air pollution and population health effects, in order to provide methodological reference and basis for the causal inference study of air pollution and population health effects in China.

Revisiting the g-null Paradox.

The (noniterative conditional expectation) parametric g-formula is an approach to estimating causal effects of sustained treatment strategies from observational data. An often-cited limitation of the parametric g-formula is the g-null paradox: a phenomenon in which model misspecification in the parametric g-formula is guaranteed in some settings consistent with the conditions that motivate its use (i.e., when identifiability conditions hold and measured time-varying confounders are affected by past treatment). Many users of the parametric g-formula acknowledge the g-null paradox as a limitation when reporting results but still require clarity on its meaning and implications. Here, we revisit the g-null paradox to clarify its role in causal inference studies. In doing so, we present analytic examples and a simulation-based illustration of the bias of parametric g-formula estimates under the conditions associated with this paradox. Our results highlight the importance of avoiding overly parsimonious models for the components of the g-formula when using this method.

Cannabinoid exposure as a major driver of pediatric acute lymphoid Leukaemia rates across the USA: combined geospatial, multiple imputation and causal inference study

BackgroundAcute lymphoid leukaemia (ALL) is the commonest childhood cancer whose incidence is rising in many nations. In the USA, between 1975 and 2016, ALL rates (ALLRs) rose 93.51% from 1.91 to 3.70/100,000 < 20 years. ALL is more common in Caucasian-Americans than amongst minorities. The cause of both the rise and the ethnic differential is unclear, however, prenatal cannabis exposure was previously linked with elevated childhood leukaemia rates. We investigated epidemiologically if cannabis use impacted nationally on ALLRs, its ethnic effects, and if the relationship was causal.MethodsState data on overall, and ethnic ALLR from the Surveillance Epidemiology and End Results databank of the Centre for Disease Control (CDC) and National Cancer Institute (NCI) were combined with drug (cigarettes, alcoholism, cannabis, analgesics, cocaine) use data from the National Survey of Drug Use and Health; 74.1% response rate. Income and ethnicity data was from the US Census bureau. Cannabinoid concentration was from the Drug Enforcement Agency Data. Data was analyzed in R by robust and spatiotemporal regression.ResultsIn bivariate analyses a dose-response relationship was demonstrated between ALLR and Alcohol Use Disorder (AUD), cocaine and cannabis exposure, with the effect of cannabis being strongest (β-estimate = 3.33(95%C.I. 1.97, 4.68), P = 1.92 × 10− 6). A strong effect of cannabis use quintile on ALLR was noted (Chi.Sq. = 613.79, P = 3.04 × 10− 70). In inverse probability weighted robust regression adjusted for other substances, income and ethnicity, cannabis was independently significant (β-estimate = 4.75(0.48, 9.02), P = 0.0389). In a spatiotemporal model adjusted for all drugs, income, and ethnicity, cannabigerol exposure was significant (β-estimate = 0.26(0.01, 0.52), P = 0.0444), an effect increased by spatial lagging (THC: β-estimate = 0.47(0.12, 0.82), P = 0.0083). After missing data imputation ethnic cannabis exposure was significant (β-estimate = 0.64(0.55, 0.72), P = 3.1 × 10− 40). 33/35 minimum e-Values ranged from 1.25 to 3.94 × 1036 indicative of a causal relationship. Relaxation of cannabis legal paradigms had higher ALLR (Chi.Squ.Trend = 775.12, P = 2.14 × 10− 112). Cannabis legal states had higher ALLR (2.395 ± 0.039 v. 2.127 ± 0.008 / 100,000, P = 5.05 × 10− 10).ConclusionsData show that ALLR is associated with cannabis consumption across space-time, is associated with the cannabinoids, THC, cannabigerol, cannabinol, cannabichromene, and cannabidiol, contributes to ethnic differentials, demonstrates prominent quintile effects, satisfies criteria for causality and is exacerbated by cannabis legalization.

Advancing Causal Inference: Differences-in-Differences vs. Bayesian Structural Time Series Models

As competition becomes fierce and competitive environments change more dynamically than ever, understanding the implications of any strategic decision has become important for firms to maximize the effect of their resource allocation. Thus, evaluating the causal inference of strategy implementation is critical for strategists, yet few studies address practitioners’ needs in this manner. While a few studies have prioritized the need for causal inference by employing the differences-in-differences (DD) method in management research, the inherit limitations of this method pose problems for the suitability and reliability of its results. Thus, this study aims to introduce a new approach for causal inference in management research, the Bayesian structural time series (BSTS) model. The BSTS analysis enables us to investigate causal inference from existing whether or not to when and how long. Using customer data from a Korean e-commerce platform, we show why BSTS can be superior to DD thereby providing a basis for subsequent use of causal inference studies delivering practical management implications.

Characteristics and outcomes of 627 044 COVID-19 patients living with and without obesity in the United States, Spain, and the United Kingdom

BackgroundA detailed characterization of patients with COVID-19 living with obesity has not yet been undertaken. We aimed to describe and compare the demographics, medical conditions, and outcomes of COVID-19 patients living with obesity (PLWO) to those of patients living without obesity.MethodsWe conducted a cohort study based on outpatient/inpatient care and claims data from January to June 2020 from Spain, the UK, and the US. We used six databases standardized to the OMOP common data model. We defined two non-mutually exclusive cohorts of patients diagnosed and/or hospitalized with COVID-19; patients were followed from index date to 30 days or death. We report the frequency of demographics, prior medical conditions, and 30-days outcomes (hospitalization, events, and death) by obesity status.ResultsWe included 627 044 (Spain: 122 058, UK: 2336, and US: 502 650) diagnosed and 160 013 (Spain: 18 197, US: 141 816) hospitalized patients with COVID-19. The prevalence of obesity was higher among patients hospitalized (39.9%, 95%CI: 39.8−40.0) than among those diagnosed with COVID-19 (33.1%; 95%CI: 33.0−33.2). In both cohorts, PLWO were more often female. Hospitalized PLWO were younger than patients without obesity. Overall, COVID-19 PLWO were more likely to have prior medical conditions, present with cardiovascular and respiratory events during hospitalization, or require intensive services compared to COVID-19 patients without obesity.ConclusionWe show that PLWO differ from patients without obesity in a wide range of medical conditions and present with more severe forms of COVID-19, with higher hospitalization rates and intensive services requirements. These findings can help guiding preventive strategies of COVID-19 infection and complications and generating hypotheses for causal inference studies.

Estimaciones de Causalidad con Imagen Médica en Oncología

This work defines a research on data strategy focused on medical imaging and derived image biomarkers to critically assess the concept of causal inference and uncertainties. Computational observational studies will be valued to generate casual inference from real world data. Our main goal is to propose a scientific methodology that allows to estimate causalities from observational studies through quality control of large databases, definition of plausible hypotheses, using computational estimated models and artificial intelligence tools. The computational approach of radiology to precision medicine by using epidemiological strategies is based on causal inference studies relies on real-world data observational, longitudinal, case-control analysis designed (being case the presence, and control the absence of the event to be estimated). In this new research setting, we consider disease in classical epidemiology as phenotyping, response to treatment and final prognosis; and exposure equals to the presence of a radiomic, dynamic image biomarker or AI modeling solution. Research with data on which causality is to be inferred must control for recruitment of closed cases, in which the researcher does not intervene in the patient’s clinical history but works on databases, collecting data to be secondary used in generating consistent causalities.

Commentary: Using potential outcomes causal methods to assess whether reductions in PM2.5 result in decreased mortality

Causal inference regarding exposures to ambient fine particulate matter (PM2.5) and mortality estimated from observational studies is limited by confounding, among other factors. In light of a variety of causal inference frameworks and methods that have been developed over the past century to specifically quantify causal effects, three research teams were selected in 2016 to evaluate the causality of PM2.5-mortality association among Medicare beneficiaries, using their own selections of causal inference methods and study designs but the same data sources. With a particular focus on controlling for unmeasured confounding, two research teams adopted an instrumental variables approach under a quasi-experiment or natural experiment study design, whereas one team adopted a structural nested mean model under the traditional cohort study design. All three research teams reported results supporting an estimated counterfactual causal relationship between ambient PM2.5 and all-cause mortality, and their estimated causal relationships are largely of similar magnitudes to recent epidemiological studies based on regression analyses with omitted potential confounders. The causal methods used by all three research teams were built upon the potential outcomes framework. This framework has marked conceptual advantages over regression-based methods in addressing confounding and yielding unbiased estimates of average treatment effect in observational epidemiological studies. However, potential violations of the unverifiable assumptions underlying each causal method leave the results from all three studies subject to biases. We also note that the studies are not immune to some other common sources of bias, including exposure measurement errors, ecological study design, model uncertainty and specification errors, and irrelevant exposure windows, that can undermine the validity of causal inferences in observational studies. As a result, despite some apparent consistency of study results from the three research teams with the wider epidemiological literature on PM2.5-mortality statistical associations, caution seems warranted in drawing causal conclusions from the results. A possible way forward is to improve study design and reduce dependence of conclusions on untested assumptions by complementing potential outcomes methods with structural causal modeling and information-theoretic methods that emphasize empirically tested and validated relationships.

Comparing Bayesian models for simultaneity judgement with different causal assumptions

In recent years, Bayesian causal inference has emerged as a computational principle for multisensory integration; it considers the causal relationships of multisensory stimuli and combines stimuli only when they are causally related. Many empirical studies have confirmed the predictions of Bayesian causal inference models and have begun to reveal the neural mechanisms of their computations. However, some studies have suggested that Bayesian causal inference does not always hold, and it is still unclear if Bayesian causal inference is a general computational principle or whether it simply represents a limited range of multisensory integration. Most theoretical and empirical Bayesian causal inference studies have focused on the binary judgement of pairs of multisensory stimuli or sequences of stimuli. One empirical study presented a more complex case with one auditory stimulus and two visual stimuli. In this study, I propose three Bayesian causal models with different causal structures regarding the same scenario. The first model assumes that an auditory stimulus is integrated with a visual stimulus only when they share the same source. During model derivation, I present a hierarchical formulation of causal structural priors in complex situations with multiple number of audio-visual stimuli. The second model assumes that an auditory stimulus is forcibly integrated with one of two visual stimuli. The third model assumes that all three stimuli are forcibly integrated. First, with minimum assumptions, I demonstrated that the full causal model explains existing experimental results better than the other two models. Meanwhile, by introducing another modelling parameter for simultaneity judgement criterion in different experimental conditions, the extended versions of the other two models also described the experimental results almost as accurately as the full causal model. It was confirmed that all three models explain the empirical results of simultaneity judgements of audio-visual stimuli, as well as the enhancement of sensitivity, to some degree. Although these models are difficult to distinguish based on existing data, they provide different interpretations of the underlying mechanisms of empirical findings. Additionally, I demonstrated that these models make qualitatively different predictions, and presented experimental manipulations that can distinguish the models based on the underlying causal structures of multiple multisensory stimuli.

Association between Malnutrition Inflammation Score and Latent Tuberculosis among Chronic Hemodialysis Patients

Objective: To investigate the association of malnutrition with latent tuberculosis (TB) among chronic kidney disease (CKD) patients on hemodialysis (HD). Methods: This was a cross-sectional study conducted at the Hemodiaylisis Unit of Dr. Hasan Sadikin General Hospital, Bandung, Indonesia. Subjects were patients aged &gt;18 years who had undergone HD twice a week for at least three months. Patients suspected of active tuberculosis (TB), malignancy, or immunocompromised were excluded. Latent TB was diagnosed using the interferon-gamma release assays (IGRA). Malnutrition was defined by a malnutrition inflammation score (MIS) of less than 5. All data including age, sex, CKD etiologies, and laboratory findings were obtained and recorded in a case report form. Results: A total of 120 subjects were involved in this study. Subjects with positive, negative, and indeterminate IGRA results were 39.2%, 56.7%, and 4.2%, respectively. There was no significant differences in subjects characteristics between positive and negative IGRA subjects. The MIS&gt;5 was shown to have no statistically significant association with positive IGRA subjects (OR=3.47, 95%CI 0.93–12.93). Conclusion: Malnutrition based on an MIS score of less than 5 is not statically associated, but clinically associated, with latent TB. Further causal inference study to investigate these associations is needed.

The Causal Relationship Between Portal Usage and Self-Efficacious Health Information-Seeking Behaviors: Secondary Analysis of the Health Information National Trends Survey Data.

BackgroundPatient portals have drawn much attention, as they are considered an important tool for health providers in facilitating patient engagement. However, little is known about whether the intensive use of patient portals contributes to improved management of patients’ health in terms of their confidence in acquiring health information and exercising self-care. There is a lack of randomized trials with these outcomes measured both pre- and postadoption of patient portals.ObjectiveThe aim of this study was to examine the causal relationship between the usage of patient portals and patients’ self-efficacy toward obtaining health information and performing self-care.MethodsThis study was a secondary data analysis that used data from a US national survey, the National Cancer Institute’s Health Information National Trends Survey 5 Cycle 1. Patient portal usage frequency was used to define the treatment. Survey items measuring self-efficacy on a Likert-type scale were selected as the main outcomes, including patients’ confidence in obtaining health information and performing self-care. To establish causality using survey data, we adopted the instrumental variables method. To determine the direction of the causal relationship in the presence of high-dimensional confounders, we further proposed a novel testing framework that employs conditional independence tests in a directed acyclic graph. The average causal effect was measured using the two-stage least squares regression method.ResultsWe showed that frequently using patient portals improves patients’ confidence in obtaining health information. The estimand of the weighted average causal effect was 0.14 (95% CI 0.06-0.23; P<.001). This means that when increasing the portal usage intensity, for instance, from 1-2 times to 3-5 times per year, the expected average increase in confidence level measured on a Likert-type scale would be 0.14. However, we could not conclusively determine the causal effect between patient portal usage and patients’ confidence in exercising self-care.ConclusionsThe results support the use of patient portals and encourage better support and education to patients. The proposed statistical method can be used to exploit the potential of national survey data for causal inference studies.

Interpretation of Guidelines for controlling confounding factors and reporting results in causal inference studies

Causal inference research is a causal test designed to assess the impact of exposures on outcomes.Both experimental and observational studies can be used to examine causal associations between exposure factors and outcomes. Experimental studies are sometimes limited by factors such as ethics or experimental conditions. Observational studies account for a large proportion in clinical studies, but the effectiveness and research value of observational studies will be affected if the design of observational studies is not rigorous and the confounding factors are not well controlled.The Guidelines for controlling confounding factors and reporting results in causal inference studie formulated by a special group of 47 editors from 35 journals from all over the world provide good guidance to researchers. This article interprets the guidelines and hopes to provide help for clinical researchers.

The cost of illiteracy: A causal inference study on how illiteracy affects physical and mental health

Objective: The aim of this study was to investigate the impact of illiteracy on physical health and mental health. Design: Matching methods (nearest neighbour matching, Mahalanobis metric matching, and propensity score matching). Setting: Elderly people at least 65 years old in 22 provinces of China. Methods: The analysis used data from Chinese Longitudinal Healthy Longevity Survey (CLHLS). The independent variable was a dummy variable, which was coded as 1 for illiterate or semiliterate, and 0 for literate. Dependent variables were indicators of physical and mental health derived from the survey results. Matching methodologies controlled for confounding variables including age, sex, living sites, access to tap water and financial support. Results: Illiteracy was found to have a significant impact on physical health, exercise habits, anxiety, loneliness and happiness. On average, illiteracy decreased physical health by 19.9%, decreased exercise habits by 7%, increased anxiety by 11.56%, increased loneliness by 17.6% and decreased happiness by 11.3%. Conclusion: Findings confirm the past literature in which illiteracy has been found to be adversely associated with physical and mental health. The analysis uniquely found that illiteracy had a higher cost on mental health as compared to physical health for elderly people in China.

Regulatory Stress Tests and Bank Responses

In this paper, we investigate how the regulatory stress test framework in the European Union affects banks’ investment decisions and portfolio choices. Using the causal inference and event study methodology, we document a substantial impact of EU-wide stress tests in 2011, 2014 and 2016 on the banks’ portfolio strategies. The banks subject to regulatory stress tests tend to structure their portfolios with lower risk assets that is reflected in a decline in risk-weighted assets as compared to the control group. At the same time, the dynamic of realized risk that is measured by the proportion of non-performing exposure in portfolios remains unaffected. The estimates based on two alternative subsamples indicate that the magnitude of such effect rise with the increase in the size of the bank´s assets.

Monkeys and humans implement causal inference to simultaneously localize auditory and visual stimuli

The environment is sampled by multiple senses, which are woven together to produce a unified perceptual state. However, optimally unifying such signals requires assigning particular signals to the same or different underlying objects or events. Many prior studies (especially in animals) have assumed fusion of cross-modal information, whereas recent work in humans has begun to probe the appropriateness of this assumption. Here we present results from a novel behavioral task in which both monkeys (Macaca mulatta) and humans localized visual and auditory stimuli and reported their perceived sources through saccadic eye movements. When the locations of visual and auditory stimuli were widely separated, subjects made two saccades, while when the two stimuli were presented at the same location they made only a single saccade. Intermediate levels of separation produced mixed response patterns: a single saccade to an intermediate position on some trials or separate saccades to both locations on others. The distribution of responses was well described by a hierarchical causal inference model that accurately predicted both the explicit "same vs. different" source judgments as well as biases in localization of the source(s) under each of these conditions. The results from this task are broadly consistent with prior work in humans across a wide variety of analogous tasks, extending the study of multisensory causal inference to nonhuman primates and to a natural behavioral task with both a categorical assay of the number of perceived sources and a continuous report of the perceived position of the stimuli.NEW & NOTEWORTHY We developed a novel behavioral paradigm for the study of multisensory causal inference in both humans and monkeys and found that both species make causal judgments in the same Bayes-optimal fashion. To our knowledge, this is the first demonstration of behavioral causal inference in animals, and this cross-species comparison lays the groundwork for future experiments using neuronal recording techniques that are impractical or impossible in human subjects.

The effect of intervention in nickel concentrations on benthic macroinvertebrates: A case study of statistical causal inference in ecotoxicology

Field survey-based ecological risk assessments for trace metals are conducted to examine the necessity and/or effectiveness of management intervention, such as setting of environmental quality standards. Observational datasets often involve confounders that may bias estimation of the effects of intervention (e.g., reduction of trace-metal concentrations through regulation). The field of ecotoxicology lags behind some other research fields in understanding proper analytical procedures for causal inference from observational datasets; there are only a few field survey-based ecotoxicological studies that have explicitly controlled for confounders in their statistical analyses. In the present study, we estimated the effect of intervention in nickel concentrations on Ephemeroptera, Plecoptera, and Trichoptera richness in rivers in Japan. We also provide detailed explanations for the backgrounds of spurious associations derived from confounders and on proper analytical procedures for obtaining an unbiased estimate of the targeted intervention effect by using regression analysis. We constructed a multiple regression model based on a causal diagram for aquatic insects and environmental factors, and on “the backdoor criterion,” that enabled us to determine the set of covariates required to obtain an unbiased estimate of the targeted intervention effect from regression coefficients. We found that management intervention in nickel concentrations may be ineffective compared to intervention in organic pollution, and that analysis ignoring the confounders overestimated the effect of intervention in nickel concentrations. Our results highlight the fact that confounders can lead to misjudging the necessity for management of anthropogenic chemical substances. Confounders should be explicitly specified and statistically controlled to achieve a comprehensive assessment of ecological risks for various substances.

Interleukin-1 receptor antagonist, interleukin-2 receptor alpha subunit and amyotrophic lateral sclerosis.

To clarify the causal associations of interleukin-1 receptor antagonist (IL-1ra) and interleukin-2 receptor alpha subunit (IL-2rα) with the risk of amyotrophic lateral sclerosis (ALS). A two-sample Mendelian randomization study design was employed. Single-nucleotide polymorphisms associated with IL-1ra (n=2) and IL-2rα (n=1) at the genome-wide significance level were used as unbiased instrumental variables. Summary-level data for ALS were obtained from Project MinE, an international collaboration consortium with 12577 ALS cases and 23475 controls of European descent. Genetic predisposition to higher levels of IL-1ra was significantly associated with lower odds of ALS. For a 1-SD increase of circulating IL-1ra levels, the odds ratio of ALS was 0.64 (95% confidence intervals, 0.46-0.88; P=0.005). There was a borderline inverse association between IL-2rα levels and ALS (odds ratio, 0.91; 95% confidence intervals, 0.83-1.00; P=0.058). Interleukin-1 receptor antagonist levels were inversely associated with ALS, suggesting that interleukin-1 inhibitors may lower the risk of this always fatal disease. The role of IL-2rα levels in ALS needs further verification in causal inference studies with larger sample sizes.

Heterogeneous Treatment and Spillover Effects Under Clustered Network Interference

The bulk of causal inference studies rules out the presence of interference between units. However, in many real-world settings units are interconnected by social, physical or virtual ties and the effect of a treatment can spill from one unit to other connected individuals in the network. In these settings, interference should be taken into account to avoid biased estimates of the treatment effect, but it can also be leveraged to save resources and provide the intervention to a lower percentage of the population where the treatment is more effective and where the effect can spill over to other susceptible individuals. In fact, different people might respond differently not only to the treatment received but also to the treatment received by their network contacts. Understanding the heterogeneity of treatment and spillover effects can help policy-makers in the scale-up phase of the intervention, it can guide the design of targeting strategies with the ultimate goal of making the interventions more cost-effective, and it might even allow generalizing the level of treatment spillover effects in other populations. In this paper, we develop a machine learning method that makes use of tree-based algorithms and an Horvitz-Thompson estimator to assess the heterogeneity of treatment and spillover effects with respect to individual, neighborhood and network characteristics in the context of clustered network interference. We illustrate how the proposed binary tree methodology performs in a Monte Carlo simulation study. Additionally, we provide an application on a randomized experiment aimed at assessing the heterogeneous effects of information sessions on the uptake of a new weather insurance policy in rural China.

Respiratory parameters and acute kidney injury in acute respiratory distress syndrome: a causal inference study.

Assess the respiratory-related parameters associated with subsequent severe acute kidney injury in mechanically ventilated patients with acute respiratory distress syndrome (ARDS). Retrospective cohort, analyzing a large public database-Multiparameter Intelligent Monitoring in Intensive Care-III. Adult patients with at least 48 h of mechanical ventilation (MV), under volume controlled ventilation and an oxygenation index less than 300 mmHg were included. A total of 1,142 patients had complete data and were included in the final analyses. According to a causal directed acyclic graph (DAG) that included respiratory system compliance (Crs), tidal volume (Vt), driving pressure (ΔP), plateau pressure (PPlat), PEEP, PaO2 and PaCO2 as possible exposures related to severe AKI, only Crs and PEEP levels had significant causal association with severe acute kidney injury (AKI) (OR 0.90, 95% CI: 0.84-0.94 for each 5-mL/cmH2O reduction in Crs; OR, 1.05 95% CI: 1.03-1.10 for each 1-cmH2O increase of PEEP). Using mediation analysis, we examined whether any mechanical ventilation, blood gas or hemodynamic parameters could explain the effects of Csr on AKI. Only PEEP mediated the significant but small effect (less than 5%) of Csr on severe AKI. The effects of PEEP, in turn, were not mediated by any other evaluated parameter. Several sensitivity analyses with (I) need of renal replacement therapy (RRT) as an alternative outcome and (II) only patients with Vt <8 mL/kg, confirmed our main findings. In trying to validate our DAG assumptions, we confirmed that only ΔP was associated with mortality but not with severe AKI. Crs and PEEP are the only respiratory-related variables with a direct causal association in severe AKI. No mechanical ventilator or blood gas parameter mediated the effects of Crs. Approaches reducing Vt and/or ΔP in ARDS can have limited effect on renal protection.

Transparent Reporting on Research Using Unstructured Electronic Health Record Data to Generate 'Real World' Evidence of Comparative Effectiveness and Safety.

Research that makes secondary use of administrative and clinical healthcare databases is increasingly influential for regulatory, reimbursement, and other healthcare decision-making. Consequently, there are numerous guidance documents on reporting for studies that use 'real-world' data captured in administrative claims and electronic health record (EHR) databases. These guidance documents are intended to improve transparency, reproducibility, and the ability to evaluate validity and relevance of design and analysis decisions. However, existing guidance does not differentiate between structured and unstructured information contained in EHRs, registries, or other healthcare data sources. While unstructured text is convenient and readily interpretable in clinical practice, it can be difficult to use for investigation of causal questions, e.g., comparative effectiveness and safety, until data have been cleaned and algorithms applied to extract relevant information to structured fields for analysis. The goal of this paper is to increase transparency for healthcare decision makers and causal inference researchers by providing general recommendations for reporting on steps taken to make unstructured text-based data usable for comparative effectiveness and safety research. These recommendations are designed to be used as an adjunct for existing reporting guidance. They are intended to provide sufficient context and supporting information for causal inference studies involving use of natural language processing- or machine learning-derived data fields, so that researchers, reviewers, and decision makers can be confident in their ability to evaluate the validity and relevance of derived measures for exposures, inclusion/exclusion criteria, covariates, and outcomes for the causal question of interest.