Causal and Associational Language in Observational Health Research: A Systematic Evaluation.

We estimated the degree to which language used in the high-profile medical/public health/epidemiology literature implied causality using language linking exposures to outcomes and action recommendations; examined disconnects between language and recommendations; identified the most common linking phrases; and estimated how strongly linking phrases imply causality. We searched for and screened 1,170 articles from 18 high-profile journals (65 per journal) published from 2010-2019. Based on written framing and systematic guidance, 3 reviewers rated the degree of causality implied in abstracts and full text for exposure/outcome linking language and action recommendations. Reviewers rated the causal implication of exposure/outcome linking language as none (no causal implication) in 13.8%, weak in 34.2%, moderate in 33.2%, and strong in 18.7% of abstracts. The implied causality of action recommendations was higher than the implied causality of linking sentences for 44.5% or commensurate for 40.3% of articles. The most common linking word in abstracts was "associate" (45.7%). Reviewers' ratings of linking word roots were highly heterogeneous; over half of reviewers rated "association" as having at least some causal implication. This research undercuts the assumption that avoiding "causal" words leads to clarity of interpretation in medical research.

Avoiding overadjustment bias in social epidemiology through appropriate covariate selection: a primer.

Obtaining accurate estimates of the causal effects of socioeconomic position (SEP) on health is important for public health interventions. To do this, researchers must identify and adjust for all potential confounding variables, while avoiding inappropriate adjustment for mediator variables on a causal pathway between the exposure and outcome. Unfortunately, 'overadjustment bias' remains a common and under-recognized problem in social epidemiology. This paper offers an introduction on selecting appropriate variables for adjustment when examining effects of SEP on health, with a focus on overadjustment bias. We discuss the challenges of estimating different causal effects including overadjustment bias, provide guidance on overcoming them, and consider specific issues including the timing of variables across the life-course, mutual adjustment for socioeconomic indicators, and conducting systematic reviews. We recommend three key steps to select the most appropriate variables for adjustment. First, researchers should be clear about their research question and causal effect of interest. Second, using expert knowledge and theory, researchers should draw causal diagrams representing their assumptions about the interrelationships between their variables of interest. Third, based on their causal diagram(s) and causal effect(s) of interest, researchers should select the most appropriate set of variables, which maximizes adjustment for confounding while minimizing adjustment for mediators.

Estimating the effects of lockdown timing on COVID-19 cases and deaths in England: A counterfactual modelling study.

BACKGROUND: During the first wave of the COVID-19 pandemic, the United Kingdom experienced one of the highest per-capita death tolls worldwide. It is debated whether this may partly be explained by the relatively late initiation of voluntary social distancing and mandatory lockdown measures. In this study, we used simulations to estimate the number of cases and deaths that would have occurred in England by 1 June 2020 if these interventions had been implemented one or two weeks earlier, and the impact on the required duration of lockdown. METHODS: Using official reported data on the number of Pillar 1 lab-confirmed cases of COVID-19 and associated deaths occurring in England from 3 March to 1 June, we modelled: the natural (i.e. observed) growth of cases, and the counterfactual (i.e. hypothetical) growth of cases that would have occurred had measures been implemented one or two weeks earlier. Under each counterfactual condition, we estimated the expected number of deaths and the time required to reach the incidence observed under natural growth on 1 June. RESULTS: Introducing measures one week earlier would have reduced by 74% the number of confirmed COVID-19 cases in England by 1 June, resulting in approximately 21,000 fewer hospital deaths and 34,000 fewer total deaths; the required time spent in full lockdown could also have been halved, from 69 to 35 days. Acting two weeks earlier would have reduced cases by 93%, resulting in between 26,000 and 43,000 fewer deaths. CONCLUSIONS: Our modelling supports the claim that the relatively late introduction of social distancing and lockdown measures likely increased the scale, severity, and duration of the first wave of COVID-19 in England. Our results highlight the importance of acting swiftly to minimise the spread of an infectious disease when case numbers are increasing exponentially.

Analyses of 'change scores' do not estimate causal effects in observational data.

BACKGROUND: In longitudinal data, it is common to create 'change scores' by subtracting measurements taken at baseline from those taken at follow-up, and then to analyse the resulting 'change' as the outcome variable. In observational data, this approach can produce misleading causal-effect estimates. The present article uses directed acyclic graphs (DAGs) and simple simulations to provide an accessible explanation for why change scores do not estimate causal effects in observational data. METHODS: Data were simulated to match three general scenarios in which the outcome variable at baseline was a (i) 'competing exposure' (i.e. a cause of the outcome that is neither caused by nor causes the exposure), (ii) confounder or (iii) mediator for the total causal effect of the exposure variable at baseline on the outcome variable at follow-up. Regression coefficients were compared between change-score analyses and the appropriate estimator(s) for the total and/or direct causal effect(s). RESULTS: Change-score analyses do not provide meaningful causal-effect estimates unless the baseline outcome variable is a 'competing exposure' for the effect of the exposure on the outcome at follow-up. Where the baseline outcome is a confounder or mediator, change-score analyses evaluate obscure estimands, which may diverge substantially in magnitude and direction from the total and direct causal effects. CONCLUSION: Future observational studies that seek causal-effect estimates should avoid analysing change scores and adopt alternative analytical strategies.

Adjustment for energy intake in nutritional research: a causal inference perspective.

BACKGROUND: Four models are commonly used to adjust for energy intake when estimating the causal effect of a dietary component on an outcome: 1) the "standard model" adjusts for total energy intake, 2) the "energy partition model" adjusts for remaining energy intake, 3) the "nutrient density model" rescales the exposure as a proportion of total energy, and 4) the "residual model" indirectly adjusts for total energy by using a residual. It remains underappreciated that each approach evaluates a different estimand and only partially accounts for confounding by common dietary causes. OBJECTIVES: We aimed to clarify the implied causal estimand and interpretation of each model and evaluate their performance in reducing dietary confounding. METHODS: Semiparametric directed acyclic graphs and Monte Carlo simulations were used to identify the estimands and interpretations implied by each model and explore their performance in the absence or presence of dietary confounding. RESULTS: The "standard model" and the mathematically identical "residual model" estimate the average relative causal effect (i.e., a "substitution" effect) but provide biased estimates even in the absence of confounding. The "energy partition model" estimates the total causal effect but only provides unbiased estimates in the absence of confounding or when all other nutrients have equal effects on the outcome. The "nutrient density model" has an obscure interpretation but attempts to estimate the average relative causal effect rescaled as a proportion of total energy. Accurate estimates of both the total and average relative causal effects may instead be derived by simultaneously adjusting for all dietary components, an approach we term the "all-components model." CONCLUSIONS: Lack of awareness of the estimand differences and accuracy of the 4 modeling approaches may explain some of the apparent heterogeneity among existing nutritional studies. This raises serious questions regarding the validity of meta-analyses where different estimands have been inappropriately pooled.

Choice of Outcome in COVID-19 Studies and Implications for Policy: Mortality and Fatality.

In this brief communication, we discuss the confusion of mortality with fatality in the interpretation of evidence in the coronavirus disease 2019 (COVID-19) pandemic, and how this confusion affects the translation of science into policy and practice. We discuss how this confusion has influenced COVID-19 policy in France, Sweden, and the United Kingdom and discuss the implications for decision-making about COVID-19 vaccine distribution. We also discuss how this confusion is an example of a more general statistical fallacy we term the "Missing Link Fallacy."

Reflection on modern methods: generalized linear models for prognosis and intervention-theory, practice and implications for machine learning.

Prediction and causal explanation are fundamentally distinct tasks of data analysis. In health applications, this difference can be understood in terms of the difference between prognosis (prediction) and prevention/treatment (causal explanation). Nevertheless, these two concepts are often conflated in practice. We use the framework of generalized linear models (GLMs) to illustrate that predictive and causal queries require distinct processes for their application and subsequent interpretation of results. In particular, we identify five primary ways in which GLMs for prediction differ from GLMs for causal inference: (i) the covariates that should be considered for inclusion in (and possibly exclusion from) the model; (ii) how a suitable set of covariates to include in the model is determined; (iii) which covariates are ultimately selected and what functional form (i.e. parameterization) they take; (iv) how the model is evaluated; and (v) how the model is interpreted. We outline some of the potential consequences of failing to acknowledge and respect these differences, and additionally consider the implications for machine learning (ML) methods. We then conclude with three recommendations that we hope will help ensure that both prediction and causal modelling are used appropriately and to greatest effect in health research.

Use of directed acyclic graphs (DAGs) to identify confounders in applied health research: review and recommendations.

BACKGROUND: Directed acyclic graphs (DAGs) are an increasingly popular approach for identifying confounding variables that require conditioning when estimating causal effects. This review examined the use of DAGs in applied health research to inform recommendations for improving their transparency and utility in future research. METHODS: Original health research articles published during 1999-2017 mentioning 'directed acyclic graphs' (or similar) or citing DAGitty were identified from Scopus, Web of Science, Medline and Embase. Data were extracted on the reporting of: estimands, DAGs and adjustment sets, alongside the characteristics of each article's largest DAG. RESULTS: A total of 234 articles were identified that reported using DAGs. A fifth (n = 48, 21%) reported their target estimand(s) and half (n = 115, 48%) reported the adjustment set(s) implied by their DAG(s). Two-thirds of the articles (n = 144, 62%) made at least one DAG available. DAGs varied in size but averaged 12 nodes [interquartile range (IQR): 9-16, range: 3-28] and 29 arcs (IQR: 19-42, range: 3-99). The median saturation (i.e. percentage of total possible arcs) was 46% (IQR: 31-67, range: 12-100). 37% (n = 53) of the DAGs included unobserved variables, 17% (n = 25) included 'super-nodes' (i.e. nodes containing more than one variable) and 34% (n = 49) were visually arranged so that the constituent arcs flowed in the same direction (e.g. top-to-bottom). CONCLUSION: There is substantial variation in the use and reporting of DAGs in applied health research. Although this partly reflects their flexibility, it also highlights some potential areas for improvement. This review hence offers several recommendations to improve the reporting and use of DAGs in future research.

Time to reality check the promises of machine learning-powered precision medicine.

Machine learning methods, combined with large electronic health databases, could enable a personalised approach to medicine through improved diagnosis and prediction of individual responses to therapies. If successful, this strategy would represent a revolution in clinical research and practice. However, although the vision of individually tailored medicine is alluring, there is a need to distinguish genuine potential from hype. We argue that the goal of personalised medical care faces serious challenges, many of which cannot be addressed through algorithmic complexity, and call for collaboration between traditional methodologists and experts in medical machine learning to avoid extensive research waste.

A causal inference perspective on the analysis of compositional data.

BACKGROUND: Compositional data comprise the parts of some whole, for which all parts sum to that whole. They are prevalent in many epidemiological contexts. Although many of the challenges associated with analysing compositional data have been discussed previously, we do so within a formal causal framework by utilizing directed acyclic graphs (DAGs). METHODS: We depict compositional data using DAGs and identify two distinct effect estimands in the generic case: (i) the total effect, and (ii) the relative effect. We consider each in the context of three specific example scenarios involving compositional data: (1) the relationship between the economically active population and area-level gross domestic product; (2) the relationship between fat consumption and body weight; and (3) the relationship between time spent sedentary and body weight. For each, we consider the distinct interpretation of each effect, and the resulting implications for related analyses. RESULTS: For scenarios (1) and (2), both the total and relative effects may be identifiable and causally meaningful, depending upon the specific question of interest. For scenario (3), only the relative effect is identifiable. In all scenarios, the relative effect represents a joint effect, and thus requires careful interpretation. CONCLUSIONS: DAGs are useful for considering causal effects for compositional data. In all analyses involving compositional data, researchers should explicitly consider and declare which causal effect is sought and how it should be interpreted.

Analysing trajectories of a longitudinal exposure: A causal perspective on common methods in lifecourse research.

Longitudinal data is commonly analysed to inform prevention policies for diseases that may develop throughout life. Commonly methods interpret the longitudinal data as a series of discrete measurements or as continuous patterns. Some of the latter methods condition on the outcome, aiming to capture 'average' patterns within outcome groups, while others capture individual-level pattern features before relating these to the outcome. Conditioning on the outcome may prevent meaningful interpretation. Repeated measurements of a longitudinal exposure (weight) and later outcome (glycated haemoglobin levels) were simulated to match three scenarios: one with no causal relationship between growth rate and glycated haemoglobin; two with a positive causal effect of growth rate on glycated haemoglobin. Two methods that condition on the outcome and one that did not were applied to the data in 1000 simulations. The interpretation of the two-step method matched the simulation in all causal scenarios, but that of the methods conditioning on the outcome did not. Methods that condition on the outcome do not accurately represent a causal relationship between a longitudinal pattern and outcome. Researchers considering longitudinal data should carefully determine if they wish to analyse longitudinal data as a series of discrete time points or by extracting pattern features.

The Association Between Childhood Leukemia and Population Mixing: An Artifact of Focusing on Clusters?

BACKGROUND: Studies investigating the population-mixing hypothesis in childhood leukemia principally use two analytical approaches: (1) nonrandom selection of areas according to specific characteristics, followed by comparisons of their incidence of childhood leukemia with that expected based on the national average; and (2) regression analyses of region-wide data to identify characteristics associated with the incidence of childhood leukemia. These approaches have generated contradictory results. We compare these approaches using observed and simulated data. METHODS: We generated 10,000 simulated regions using the correlation structure and distributions from a United Kingdom dataset. We simulated cases using a Poisson distribution with the incidence rate set to the national average assuming the null hypothesis that only population size drives the number of cases. Selection of areas within each simulated region was based on characteristics considered responsible for elevated infection rates (population density and inward migration) and/or elevated leukemia rates. We calculated effect estimates for 10,000 simulations and compared results to corresponding observed data analyses. RESULTS: When the selection of areas for analysis is based on apparent clusters of childhood leukemia, biased assessments occur; the estimated 5-year incidence of childhood leukemia ranged between zero and eight per 10,000 children in contrast to the simulated two cases per 10,000 children, similar to the observed data. Performing analyses on region-wide data avoids these biases. CONCLUSIONS: Studies using nonrandom selection to investigate the association between childhood leukemia and population mixing are likely to have generated biased findings. Future studies can avoid such bias using a region-wide analytical strategy. See video abstract at, http://links.lww.com/EDE/B431.

Survival, by Birth Weight and Gestational Age, in Individuals With Congenital Heart Disease: A Population-Based Study.

BACKGROUND: Congenital heart disease (CHD) survival estimates are important to understand prognosis and evaluate health and social care needs. Few studies have reported CHD survival estimates according to maternal and fetal characteristics. This study aimed to identify predictors of CHD survival and report conditional survival estimates. METHODS AND RESULTS: Cases of CHD (n=5070) born during 1985-2003 and notified to the Northern Congenital Abnormality Survey (NorCAS) were matched to national mortality information in 2008. Royston-Parmar regression was performed to identify predictors of survival. Five-year survival estimates conditional on gestational age at delivery, birth weight, and year of birth were produced for isolated CHD (ie, CHD without extracardiac anomalies). Year of birth, gestational age, birth weight, and extracardiac anomalies were independently associated with mortality (all P≤0.001). Five-year survival for children born at term (37-41 weeks) in 2003 with average birth weight (within 1 SD of the mean) was 96.3% (95% CI, 95.6-97.0). Survival was most optimistic for high-birth-weight children (>1 SD from the mean) born post-term (≥42 weeks; 97.9%; 95% CI, 96.8-99.1%) and least optimistic for very preterm (<32 weeks) low-birth-weight (<1 SD from mean) children (78.8%; 95% CI, 72.8-99.1). CONCLUSIONS: Five-year CHD survival is highly influenced by gestational age and birth weight. For prenatal counseling, conditional survival estimates provide best- and worst-case scenarios, depending on final gestational age and birth weight. For postnatal diagnoses, they can provide parents with more-accurate predictions based on their baby's birth weight and gestational age.

Risk estimates of recurrent congenital anomalies in the UK: a population-based register study.

BACKGROUND: Recurrence risks for familial congenital anomalies in successive pregnancies are known, but this information for major structural anomalies is lacking. We estimated the absolute and relative risks of recurrent congenital anomaly in the second pregnancy for women with a history of a congenital anomaly in the first pregnancy, for all major anomaly groups and subtypes. METHODS: Population-based register data on 18,605 singleton pregnancies affected by major congenital anomaly occurring in 872,493 singleton stillbirths, live births and terminations of pregnancy for fetal anomaly were obtained from the Northern Congenital Abnormality Survey, North of England, UK, for 1985-2010. Absolute risks (ARs) and relative risks (RRs) for recurrent congenital anomaly (overall, from a similar group, from a dissimilar group) in the second pregnancy were estimated by history of congenital anomaly (overall, by group, by subtype) in the first pregnancy. RESULTS: The estimated prevalences of congenital anomaly in first and second pregnancies were 275 (95% CI 270-281) and 163 (95% CI 159-168) per 10,000 respectively. For women whose first pregnancy was affected by congenital anomaly, the AR of recurrent congenital anomaly in the second pregnancy was 408 (95% CI 365-456) per 10,000, 2.5 (95% CI 2.3-2.8, P < 0.0001) times higher than for those with unaffected first pregnancies. For similar anomalies, the recurrence risk was considerably elevated (RR = 23.8, 95% CI 19.6-27.9, P < 0.0001), while for dissimilar anomalies the increase was more modest (RR = 1.4, 95% CI 1.2-1.6, P = 0.001), although the ARs for both were 2%. CONCLUSIONS: Absolute recurrence risks varied between 1 in 20 and 1 in 30 for most major anomaly groups. At pre-conception and antenatal counselling, women whose first pregnancy was affected by a congenital anomaly and who are planning a further pregnancy may find it reassuring that, despite high relative risks, the absolute recurrence risk is relatively low.

Variation in health care for children and young people with cerebral palsies: a retrospective multicentre audit study.

AIM: To explore the provision and variations in care for children and young people with cerebral palsies (CP) registered with the population-based North of England Collaborative Cerebral Palsy Survey (NECCPS). METHOD: This is a retrospective multicentre record audit of 389 children with CP (220 males, 148 females, 21 no data; median age at time of audit 12y 3mo), born between 1995 and 2002. Data were collected on cranial magnetic resonance imaging (MRI), hip and spine surveillance and management, and pain presence and management. Variations over time and between the districts in the north of England (Northumberland, North and West Cumbria, North and South Tyneside, Newcastle-upon-Tyne, Gateshead, Sunderland, Durham, Darlington, Bishop Auckland, Hartlepool, Stockton-on-Tees, Middlesbrough, Redcar, and Cleveland), and by socio-economic status (SES) (estimated from the Index of Multiple Deprivation [IMD] 2004) were estimated by generalized estimating equations. RESULTS: There was significant variation between districts in access to MRI (p<0.001), orthopaedic surgeons (p=0.005), recording state of spine (p<0.001), and discussions about pain (p<0.001). Fifty-seven per cent (95% CI 52-62) had evidence of a reported MRI brain scan, the proportion of which increased over time (p<0.001). Sixty-seven per cent (95% CI 62-71) had a discussion about pain recorded. Of those in pain, 87% (95% CI 80-93) had a pain management plan. The proportion with documented discussion about pain increased with increasing SES (p=0.04). INTERPRETATION: The provision of care for children with CP in the north of England varies between districts. Internationally agreed, evidence-based standards are urgently needed to ensure more equitable health care and improved outcomes for all.

Langerhans cell histiocytosis in children with congenital anomalies: a population-based record linkage study.

BACKGROUND: The etiology of Langerhans cell histiocytosis (LCH), a rare cancer-like disorder of the immune system, is largely unknown although a genetic component has been suggested based on familial cases, and reports of chromosome instability and genetic mutation. Associations between various cancers and congenital anomalies have been reported and although congenital anomalies have been noted in children with LCH only one study to date has reported their frequency. An association between congenital anomalies and LCH may suggest a common etiological pathway, in particular, a genetic pathway. METHODS: Data from two coterminous registries in the same geographic region were used. All cases of LCH on the Northern Region Young Persons Malignant Disease Register diagnosed between 1985 and 2010 were cross-matched with live-born cases of congenital anomaly registered by the Northern Congenital Abnormality Survey. RESULTS: A total of 819,890 children and young people were born during 1985 to 2008. Of these, 13,799 (1.7%) had a congenital anomaly and 39 (0.005%) were diagnosed with LCH. Three LCH cases were identified among those with congenital anomalies, all three of whom had congenital heart disease. The relative risk of LCH for those with a congenital anomaly, compared with those without, was 4.87 (95% confidence interval, 1.50-15.81; p = 0.03). CONCLUSION: LCH was associated with congenital anomaly in a small but statistically significant number of patients, raising the possibility of a common genetic pathway in some cases.