Reflection on modern methods: selection bias-a review of recent developments

Harnessing EHR data for health research

With the increasing availability of rich, longitudinal, real-world clinical data recorded in electronic health records (EHRs) for millions of patients, there is a growing interest in leveraging these records to improve the understanding of human health and disease and translate these insights into clinical applications. However, there is also a need to consider the limitations of these data due to various biases and to understand the impact of missing information. Recognizing and addressing these limitations can inform the design and interpretation of EHR-based informatics studies that avoid confusing or incorrect conclusions, particularly when applied to population or precision medicine. Here we discuss key considerations in the design, implementation and interpretation of EHR-based informatics studies, drawing from examples in the literature across hypothesis generation, hypothesis testing and machine learning applications. We outline the growing opportunities for EHR-based informatics studies, including association studies and predictive modeling, enabled by evolving AI capabilities-while addressing limitations and potential pitfalls to avoid.

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

Purpose of review

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

Recent findings

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

Conclusion

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

Using Exogenous Social Media Exposure Measures to Assess the Effects of Smokeless Tobacco-Related Social Media Content on Smokeless Tobacco Sales in the United States

INTRODUCTION

Prior research on the effects of social media promotion of tobacco products has predominantly relied on survey-based self-report measures of marketing exposure, which potentially introduce endogeneity, recall, and selection biases. New approaches can enhance measurement and help better understand the effects of exposure to tobacco-related messages in a dynamic social media marketing environment. We used geolocation-specific tweet rate as an exogenous indicator of exposure to smokeless tobacco (ST)-related content and employed this measure to examine the influence of social media marketing on ST sales.

AIMS AND METHODS

Autoregressive error models were used to analyze the association between the ST-relevant tweet rate (aggregated by 4-week period from February 12, 2017 to June 26, 2021 and scaled by population density) and logarithmic ST unit sales across time by product type (newer, snus, conventional) in the United States, accounting for autocorrelated errors. Interrupted time series approach was used to control for policy change effects.

RESULTS

ST product category-related tweet rates were associated with ST unit sales of newer and conventional products, controlling for price, relevant policy events, and the coronavirus disease 2019 (COVID-19) pandemic. On average, 100-unit increase in the number of newer ST-related tweets was associated with 14% increase in unit sales (RR = 1.14; p = .01); 100-unit increase in conventional ST tweets was associated with ~1% increase in unit sales (p = .04). Average price was negatively associated with the unit sales.

CONCLUSIONS

Study findings reveal that ST social media tweet rate was related to increased ST consumption and illustrate the utility of exogenous measures in conceptualizing and assessing effects in the complex media environment.

IMPLICATIONS

Tobacco control initiatives should include efforts to monitor the role of social media in promoting tobacco use. Surveillance of social media platforms is critical to monitor emerging tobacco product-related marketing strategies and promotional content reach. Exogenous measures of potential exposure to social media messages can supplement survey data to study media effects on tobacco consumption.

Family-based pediatric weight management interventions in US primary care settings targeting children ages 6-12 years old: A systematic review guided by the RE-AIM framework

Obesity is a pandemic that disproportionately affects children from vulnerable populations in the USA. Current treatment approaches in primary care settings in the USA have been reported to be insufficient at managing pediatric obesity, primarily due to implementation challenges for healthcare systems and barriers for families. While the literature has examined the efficacy of pediatric obesity interventions focused on internal validity, it lacks sufficient reporting and analysis of external validity necessary for successful translation to primary care settings. We conducted a systematic review of the primary-care-setting literature from January 2007 to March 2020 on family-based pediatric weight management interventions in both English and/or Spanish for children ages 6-12 years in the USA using the Reach, Efficacy/Effectiveness, Adoption, Implementation, Maintenance (RE-AIM) framework. A literature search, using PRISMA guidelines, was conducted in January 2022 using the following electronic databases: Medline Ovid, Embase, and Cochrane Library. 22 270 records were screened, and 376 articles were reviewed in full. 184 studies were included. The most commonly reported dimensions of the RE-AIM framework were Reach (65%), Efficacy/Effectiveness (64%), and Adoption (64%), while Implementation (47%) and Maintenance (42%) were less often reported. The prevalence of reporting RE-AIM construct indicators ranged greatly, from 1% to 100%. This systematic review underscores the need for more focus on external validity to guide the development, implementation, and dissemination of future pediatric obesity interventions based in primary care settings. It also suggests conducting additional research on sustainable financing for pediatric obesity interventions.

Tracking COVID-19 vaccination expectancies and vaccination refusal in the United States

To identify factors that predict COVID-19 vaccination refusal and show how expectancies affect vaccination acceptance for non-vaccinated adults, we used a monthly repeated cross-sectional sample from June/2021 to October/2021 to collect data on vaccination behaviors and predictor variables for 2,116 US adults over 50 years of age. Selection bias modeling - which is required when data availability is a result of behavioral choice - predicts two outcomes: (1) no vaccination vs. vaccination for the entire sample and (2) the effects of expectancy indices predicting vaccination Refuser vs. vaccination Accepters for the unvaccinated group. Vaccine refusers were younger and less educated, endorsed common misconceptions about the COVID-19 epidemic, and were Black. Vaccination expectancies were related to vaccination refusal in the unvaccinated eligible group: negative expectancies increased vaccine refusal, while positive expectancies decreased it. We conclude that behavior-related expectancies (as opposed to more stable psychological traits) are important to identify because they are often modifiable and provide a point of intervention, not just for COVID-19 vaccination acceptance but also for other positive health behaviors.

Identifiability of causal effects in test-negative design studies

Causal directed acyclic graphs (DAGs) are often used to select variables in a regression model to identify causal effects. Outcome-based sampling studies, such as the 'test-negative design' used to assess vaccine effectiveness, present unique challenges that are not addressed by the common back-door criterion. Here we discuss intuitive, graphical approaches to explain why the common back-door criterion cannot be used for identification of population average causal effects with outcome-based sampling studies. We also describe graphical rules that can be used instead in outcome-based sampling studies when the objective is limited to determining if the causal odds ratio is identifiable, and illustrate recent changes to the free online software Dagitty which incorporate these principles.

Educational patterns of health behaviors and body mass index: A longitudinal multiple correspondence analysis of a middle-aged general population, 2007-2016

Social differences in body mass index and health behaviors are a major public health challenge. The uneven distribution of unhealthy body mass index and of unhealthy behaviors such as smoking, physical inactivity, and harmful alcohol consumption has been shown to mediate social inequalities in chronic diseases. While differential exposures to these health variables have been investigated, the extent to which they vary over the lifetime in the same population and their relationship with level of education is not well understood. This study examines patterns of body mass index and multiple health behaviors (smoking, physical activity and alcohol consumption), and investigates their association with education level among adults living in Northern Norway. It presents findings from a longitudinal multiple correspondence analysis of the Tromsø Study. Longitudinal data from 8,906 adults aged 32-87 in 2007-2008, with repeated measurements in 2015-2016 were retrieved from the survey's sixth and seventh waves. The findings suggest that most in the study population remained in the same categories of body mass index and the three health behaviors at the follow-up, with a clear educational gradient in healthy patterns. That is, both healthy changes and maintained healthy categories were associated with the highest education levels. Estimating differential exposures to mediators of health inequalities could benefit policy priority setting for tackling inequalities in health.

Discriminative machine learning for maximal representative subsampling

Biased population samples pose a prevalent problem in the social sciences. Therefore, we present two novel methods that are based on positive-unlabeled learning to mitigate bias. Both methods leverage auxiliary information from a representative data set and train machine learning classifiers to determine the sample weights. The first method, named maximum representative subsampling (MRS), uses a classifier to iteratively remove instances, by assigning a sample weight of 0, from the biased data set until it aligns with the representative one. The second method is a variant of MRS - Soft-MRS - that iteratively adapts sample weights instead of removing samples completely. To assess the effectiveness of our approach, we induced artificial bias in a public census data set and examined the corrected estimates. We compare the performance of our methods against existing techniques, evaluating the ability of sample weights created with Soft-MRS or MRS to minimize differences and improve downstream classification tasks. Lastly, we demonstrate the applicability of the proposed methods in a real-world study of resilience research, exploring the influence of resilience on voting behavior. Through our work, we address the issue of bias in social science, amongst others, and provide a versatile methodology for bias reduction based on machine learning. Based on our experiments, we recommend to use MRS for downstream classification tasks and Soft-MRS for downstream tasks where the relative bias of the dependent variable is relevant.

Sensitivity analysis of selection bias: a graphical display by bias-correction index

Background

In observational studies, how the magnitude of potential selection bias in a sensitivity analysis can be quantified is rarely discussed. The purpose of this study was to develop a sensitivity analysis strategy by using the bias-correction index (BCI) approach for quantifying the influence and direction of selection bias.

Methods

We used a BCI, a function of selection probabilities conditional on outcome and covariates, with different selection bias scenarios in a logistic regression setting. A bias-correction sensitivity plot was illustrated to analyze the associations between proctoscopy examination and sociodemographic variables obtained using the data from the Taiwan National Health Interview Survey (NHIS) and of a subset of individuals who consented to having their health insurance data further linked.

Results

We included 15,247 people aged ≥20 years, and 87.74% of whom signed the informed consent. When the entire sample was considered, smokers were less likely to undergo proctoscopic examination (odds ratio (OR): 0.69, 95% CI [0.57-0.84]), than nonsmokers were. When the data of only the people who provided consent were considered, the OR was 0.76 (95% CI [0.62-0.94]). The bias-correction sensitivity plot indicated varying ORs under different degrees of selection bias.

Conclusions

When data are only available in a subsample of a population, a bias-correction sensitivity plot can be used to easily visualize varying ORs under different selection bias scenarios. The similar strategy can be applied to models other than logistic regression if an appropriate BCI is derived.

Long-term exposure to the components of fine particulate matters and disability after stroke: Findings from the China National Stroke Screening Surveys

BACKGROUND

Long-term exposure to ambient fine particulate matter (PM2.5) has been linked to an increased risk of stroke. However, the effect of long-term exposure to PM2.5 and its major components on the functional disability of stroke patients remains unclear.

METHODS

Based on China National Stroke Screening Survey data obtained from 2013 to 2019, we conducted a national multicenter longitudinal study of the associations of long-term exposure to PM2.5 and its components with the risk of disability after stroke in China. Post-stroke disability was assessed using the modified Rankin scale (mRS), which ranges from 0 to 5, with higher scores indicating greater disability. Long-term exposure to PM2.5 and its five components [sulfate (SO42-), nitrate (NO3-), ammonium salt (NH4+), organic matter (OM), and black carbon (BC)] was determined based on average concentrations during the 3 years preceding mRS administration according to the geographic coordinates of residential communities, using state-of-the-art estimates from multiple sources. We used a fixed-effect model to evaluate the associations between mRS scores and PM2.5 exposure, with adjustment for multiple covariates.

RESULTS

Every 10 µg/m3 increase in PM2.5 was associated with a 0.019 (95% confidence interval, 0.003, 0.036) increase in mRS score, but the effect was not significant after adjusting for all covariates [0.016 (95% CI, -0.003, 0.032)]. For PM2.5 components, each IQR (7.92 µg/m3) increment in OM exposure was associated significantly with 0.062 (95% CI, 0.013, 0.111) increment in the mRS score. A significant association was observed between SO42- exposure and the mRS score [0.057 (95% CI, 0.003, 0.112), per IQR: 6.28 µg/m3]. However, no significant association was found with BC, NO3-, or NH4+ exposure. Furthermore, the nonlinear curves were observed for the exposure-response relationship between PM2.5 exposure and the mRS score.

CONCLUSION

Greater PM2.5 exposure increased the mRS score and was associated with post-stroke functional disability among stroke patients. However, different chemical components showed unequal neurotoxic effects, and long-term exposure to OM and SO42- may play a more important role.

SYNOPSIS

This study reports fine particulate matter at higher concentrations damages the functional ability among specific stroke patients, and PM2.5 components have different neurotoxicities.

Estimating climate change and mental health impacts in Canada: A cross-sectional survey protocol

Climate change has severe and sweeping impacts on mental health. Although research is burgeoning on mental health impacts following climate and weather extremes, less is known about how common these impacts are outside of extreme events. Existing research exploring the prevalence of psychosocial responses to climate change primarily examines university students and uses non-random sampling methods. Herein, our protocol outlines an approach to data collection, processing, and analysis to estimate the population prevalence, magnitude, and distribution of mental health responses to climate change in Canada. A cross-sectional survey of youth and adults aged 13 years and older in Canada will be administered over the course of one year. The questionnaire will take approximately 10 minutes to complete orally and will be administered in English, French, and Inuktitut. The survey will consist of six sections: (1) self-reported past experiences of climate change; (2) self-reported climate-related emotions; (3) self-reported past and current impacts, anticipatory impacts, and vicarious experiences; (4) self-reported subclinical outcomes; (5) self-reported behavioural responses; and (6) demographics. A multi-stage, multi-stratified random probability sampling method will be used to obtain a sample representative of the Canadian population. We will use two different modes of recruitment: an addressed letter sent by postal mail or a telephone call (landlines and cellular). Population-weighted descriptive statistics, principal component analysis, and weighted multivariable regression will be used to analyse the data. The results of this survey will provide the first national prevalence estimates of subclinical mental health responses to climate change outcomes of people living in Canada.

Characterizing Acute Low Back Pain in a Community-Based Cohort

Acute low back pain (LBP) is a common experience, however, the associated pain severity, pain frequency, and characteristics of individuals with acute LBP in community settings have yet to be well understood. The purpose of this manuscript is to categorize and compare acute LBP groups for differences in the following characteristics: 1) sociodemographic, 2) general and physical health, and 3) psychological. This cross-sectional study used baseline data from 131 community-based participants with acute LBP (<4 weeks duration before screening and > 30 pain-free days before acute LBP onset). Two LBP categorization definitions were used based on LBP frequency combined with either: 1) pain interference frequency (impact-based) or 2) pain intensity (intensity-based). Descriptive associations were calculated as prevalence ratios for categorical variables and Hedges' g for continuous variables. Our analyses identified several large associations for impact-based and intensity-based categories for pain interference with activity and with enjoyment in life, global mental health, STarT Back Screening Tool risk category, general health, and Fear Avoidance Beliefs Questionnaire-Physical Activity subscale. Larger associations were found with social constructs (racially and ethnically minoritized, performance of social roles, and isolation) when using the intensity-based versus impact-based categorization. This study adds to the literature by providing standard ways to characterize community-based individuals experiencing acute LBP. The robust differences observed between these categorization approaches suggest they may be used to improve the early identification of factors potentially contributing to the development of chronic LBP.

Prepregnancy obesity and risk of placental inflammation at term: a selection bias analysis

PURPOSE

Placental histopathology is a resource for investigating obesity-associated pregnancy conditions. However, studies oversample adverse pregnancies, biasing findings. We examine the association between prepregnancy obesity (risk factor for inflammation) and histologic placental inflammation (correlated with impaired infant neurodevelopment) and how selection bias may influence the association.

METHODS

Singleton term deliveries between 2008 and 2012 from the Magee Obstetric Maternal and Infant database were analyzed. Prepregnancy body mass index (BMI) was categorized as underweight, lean (referent), overweight, and obese. Outcomes were diagnoses of acute (acute chorioamnionitis and fetal inflammation) and chronic placental inflammation (chronic villitis). Risk ratios for associations between BMI and placental inflammation were estimated using selection bias approaches: complete case, exclusion of pregnancy complications, multiple imputation, and inverse probability weighting. E-values approximated how susceptible estimates were to residual selection bias.

RESULTS

Across methods, obesity was associated with an 8-15% lower risk of acute chorioamnionitis, a 7%-14% lower risk of acute fetal inflammation, and a 12%-30% higher risk of chronic villitis relative to lean women. E-values indicated modest residual selection bias could explain away associations, though few measured indications of placental evaluations met this threshold.

CONCLUSIONS

Obesity may contribute to placental inflammation, and we highlight robust methods to analyze clinical data susceptible to selection bias.

The need for a clinical case definition in test-negative design studies estimating vaccine effectiveness

Test negative studies have been used extensively for the estimation of COVID-19 vaccine effectiveness (VE). Such studies are able to estimate VE against medically-attended illness under certain assumptions. Selection bias may be present if the probability of participation is associated with vaccination or COVID-19, but this can be mitigated through use of a clinical case definition to screen patients for eligibility, which increases the likelihood that cases and non-cases come from the same source population. We examined the extent to which this type of bias could harm COVID-19 VE through systematic review and simulation. A systematic review of test-negative studies was re-analysed to identify studies ignoring the need for clinical criteria. Studies using a clinical case definition had a lower pooled VE estimate compared with studies that did not. Simulations varied the probability of selection by case and vaccination status. Positive bias away from the null (i.e., inflated VE consistent with the systematic review) was observed when there was a higher proportion of healthy, vaccinated non-cases, which may occur if a dataset contains many results from asymptomatic screening in settings where vaccination coverage is high. We provide an html tool for researchers to explore site-specific sources of selection bias in their own studies. We recommend all groups consider the potential for selection bias in their vaccine effectiveness studies, particularly when using administrative data.

Genetic associations with parental investment from conception to wealth inheritance in six cohorts

Genetic inheritance is not the only way parents' genes may affect children. It is also possible that parents' genes are associated with investments into children's development. We examined evidence for links between parental genetics and parental investments, from the prenatal period through to adulthood, using data from six population-based cohorts in the UK, US and New Zealand, together totalling 36,566 parents. Our findings revealed associations between parental genetics-summarized in a genome-wide polygenic score-and parental behaviour across development, from smoking in pregnancy, breastfeeding in infancy, parenting in childhood and adolescence, to leaving a wealth inheritance to adult children. Effect sizes tended to be small at any given time point, ranging from RR = 1.12 (95% confidence interval (95%CI) 1.09, 1.15) to RR = 0.76 (95%CI 0.72, 0.80) during the prenatal period and infancy; β = 0.07 (95%CI 0.04, 0.11) to β = 0.29 (95%CI 0.27, 0.32) in childhood and adolescence, and RR = 1.04 (95%CI 1.01, 1.06) to RR = 1.11 (95%CI 1.07, 1.15) in adulthood. There was evidence for accumulating effects across development, ranging from β = 0.15 (95%CI 0.11, 0.18) to β = 0.23 (95%CI 0.16, 0.29) depending on cohort. Our findings are consistent with the interpretation that parents pass on advantages to offspring not only via direct genetic transmission or purely environmental paths, but also via genetic associations with parental investment from conception to wealth inheritance.

Bias assessment of a test-negative design study of COVID-19 vaccine effectiveness used in national policymaking

National test-negative-case-control (TNCC) studies are used to monitor COVID-19 vaccine effectiveness in the UK. A questionnaire was sent to participants from the first published TNCC COVID-19 vaccine effectiveness study conducted by the UK Health Security Agency, to assess for potential biases and changes in behaviour related to vaccination. The original study included symptomatic adults aged ≥70 years testing for COVID-19 between 08/12/2020 and 21/02/2021. A questionnaire was sent to cases and controls tested from 1-21 February 2021. In this study, 8648 individuals responded to the questionnaire (36.5% response). Using information from the questionnaire to produce a combined estimate that accounted for all potential biases decreased the original vaccine effectiveness estimate after two doses of BNT162b2 from 88% (95% CI: 79-94%) to 85% (95% CI: 68-94%). Self-reported behaviour demonstrated minimal evidence of riskier behaviour after vaccination. These findings offer reassurance to policy makers and clinicians making decisions based on COVID-19 vaccine effectiveness TNCC studies.

Revisiting assumptions in test-negative studies for estimating vaccine effectiveness: the need for a clinical case definition

Test negative studies have been used extensively for the estimation of COVID-19 vaccine effectiveness (VE). Such studies are able to estimate VE against medically-attended illness under certain assumptions. Selection bias may be present if the probability of participation is associated with vaccination or COVID-19, but this can be mitigated through use of a clinical case definition to screen patients for eligibility, which increases the likelihood that cases and non-cases come from the same source population. We examined the extent to which this type of bias could harm COVID-19 VE through systematic review and simulation. A systematic review of test-negative studies was re-analysed to identify studies ignoring the need for clinical criteria. Studies using a clinical case definition had a lower pooled VE estimate compared with studies that did not. Simulations varied the probability of selection by case and vaccination status. Positive bias away from the null (i.e., inflated VE consistent with the systematic review) was observed when there was a higher proportion of healthy, vaccinated non-cases, which may occur if a dataset contains many results from asymptomatic screening in settings where vaccination coverage is high. We provide an html tool for researchers to explore site-specific sources of selection bias in their own studies. We recommend all group consider the potential for selection bias in their vaccine effectiveness studies, particularly when using administrative data.

Demography and epidemiology of captive former biomedical research chimpanzees (Pan troglodytes) 1: Survival and mortality

Accurate and up-to-date data on longevity and mortality are essential for describing, analyzing, and managing animal populations in captivity. We assembled a comprehensive demography data set and analyzed survival and mortality patterns in a population of captive former biomedical research chimpanzees. The study synthesized over 51,000 life-years of demographic data collected on 2349 individuals between 1923 and 2014. Our goal was to assess the population's current age-sex composition, estimate rates of survivorship, mortality and life expectancy, and compare findings with other chimpanzee populations of interest. Results indicated an increasingly geriatric contemporary population declining in size. The median life expectancy (MLE) of the entire population was 32.6 years (males 29.1, females 36.1). For chimpanzees who reached 1 year of age, the MLE increased to 34.9 years (males 31.0, females 38.8). Survival probability was influenced by both sex and birth type. Females exhibited greater survivorship than males (β₁  = -0.34, z = -5.74, p < 0.001) and wild-born individuals exhibited greater survivorship than captive-born individuals (β₂  = -0.55, z = -5.89, p < 0.001). There was also a seasonal trend in mortality, wherein more individuals died during the winter months (December-February) compared with other seasons. Analyses of life expectancy over time showed continual increases in both median age of living individuals and median age at death, suggesting that these chimpanzees have yet to reach their full aging potential in a postresearch environment. As they continue to age, ongoing monitoring and analysis of demographic changes will be necessary for science-based population and program management until extinction occurs some decades in the future.

Neighbourhood-related socioeconomic perinatal health inequalities: An illustration of the mediational g-formula and considerations for the big data context

BACKGROUND

Advances in computing power have enabled the collection, linkage and processing of big data. Big data in conjunction with robust causal inference methods can be used to answer research questions regarding the mechanisms underlying an exposure-outcome relationship. The g-formula is a flexible approach to perform causal mediation analysis that is suited for the big data context. Although this approach has many advantages, it is underused in perinatal epidemiology and didactic explanation for its implementation is still limited.

OBJECTIVE

The aim of this was to provide a didactic application of the mediational g-formula by means of perinatal health inequalities research.

METHODS

The analytical procedure of the mediational g-formula is illustrated by investigating whether the relationship between neighbourhood socioeconomic status (SES) and small for gestational age (SGA) is mediated by neighbourhood social environment. Data on singleton births that occurred in the Netherlands between 2010 and 2017 (n = 1,217,626) were obtained from the Netherlands Perinatal Registry and linked to sociodemographic national registry data and neighbourhood-level data. The g-formula settings corresponded to a hypothetical improvement in neighbourhood SES from disadvantaged to non-disadvantaged.

RESULTS

At the population level, a hypothetical improvement in neighbourhood SES resulted in a 6.3% (95% confidence interval [CI] 5.2, 7.5) relative reduction in the proportion of SGA, that is the total effect. The total effect was decomposed into the natural direct effect (5.6%, 95% CI 5.1, 6.1) and the natural indirect effect (0.7%, 95% CI 0.6, 0.9). In terms of the magnitude of mediation, it was observed the natural indirect effect accounted for 11.4% (95% CI 9.2, 13.6) of the total effect of neighbourhood SES on SGA.

CONCLUSIONS

The mediational g-formula is a flexible approach to perform causal mediation analysis that is suited for big data contexts in perinatal health research. Its application can contribute to providing valuable insights for the development of policy and public health interventions.

Current Technology Developments Can Improve the Quality of Research and Level of Evidence for Rehabilitation Interventions: A Narrative Review

The current important limitations to the implementation of Evidence-Based Practice (EBP) in the rehabilitation field are related to the validation process of interventions. Indeed, most of the strict guidelines that have been developed for the validation of new drugs (i.e., double or triple blinded, strict control of the doses and intensity) cannot-or can only partially-be applied in rehabilitation. Well-powered, high-quality randomized controlled trials are more difficult to organize in rehabilitation (e.g., longer duration of the intervention in rehabilitation, more difficult to standardize the intervention compared to drug validation studies, limited funding since not sponsored by big pharma companies), which reduces the possibility of conducting systematic reviews and meta-analyses, as currently high levels of evidence are sparse. The current limitations of EBP in rehabilitation are presented in this narrative review, and innovative solutions are suggested, such as technology-supported rehabilitation systems, continuous assessment, pragmatic trials, rehabilitation treatment specification systems, and advanced statistical methods, to tackle the current limitations. The development and implementation of new technologies can increase the quality of research and the level of evidence supporting rehabilitation, provided some adaptations are made to our research methodology.

Toward a Clearer Definition of Selection Bias When Estimating Causal Effects

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

Prevalence of chronic non-communicable diseases in Ethiopia: A systematic review and meta-analysis of evidence

Background

Non-communicable diseases (NCDs) are a growing global health challenge disproportionately impacting low- and middle-income settings, including Ethiopia. Currently, the body of evidence describing the burden of NCDs is fragmented, inconsistent, health facility- or institution-based, and out-dated in Ethiopia. We conducted a systematic review of the literature and meta-analysis of the prevalence of NCDs in community settings in Ethiopia.

Review methodology

Community-based quantitative studies published in English between January 1st, 2012, and June 30th, 2022, that reported on the prevalence of NCDs in Ethiopia were included. A systematic search of Medline, Embase, Scopus, CINAHL, and Global Health using pretested search terms related to NCDs was conducted, and data were extracted using a piloted data extraction proforma adapted from the Joanna Briggs Institute tool. Meta-analysis was performed using Stata 16. While the pooled prevalence of Diabetes Mellitus (DM) and undiagnosed (DM) was computed and presented using forest plots, then overall prevalence of NCDs and other various types of NCDs were narratively synthesized. I2 was used to assess heterogeneity. Studies that did not fulfill the criteria (used similar tool to measure the types of NCDs) for meta-analysis were narratively synthesized.

Results

Twenty-two studies met the inclusion criteria. Five studies measured the prevalence of NCDs (all NCDs together), ranging from 29 to 35% (prevalence estimates not pooled). The pooled prevalence of Diabetes Mellitus (DM) across ten studies was 5% (95% CI: 4–7%). Three studies each reported on the prevalence of undiagnosed DM (pooled prevalence 5%, 95% CI: 4–7%) and pre-DM (pooled prevalence 7%, 95% CI: 3–14%%). In a narrative analysis the prevalence of cardiovascular conditions ranged from 13.4 to 32.2% (n = 3 studies), cancer mortality ranged from 4 to 18% (n = 3 studies) and respiratory conditions ranged from 1 to 18% (n = 3 studies). Some studies have determined more than one NCDs and that is why the total number of studies are exceeding more than twenty studies.

Conclusion and recommendations

Our analysis found that approximately one-third of Ethiopians have an NCD, with cardiovascular diseases the most common of all NCDs. The prevalence of respiratory conditions also appears high, but there are insufficient data for a pooled estimate. Whilst the prevalence of DM appears relatively low, there is evidence that the magnitude is increasing. Public health actions to address the high burden of cardiovascular and respiratory diseases, as well as the increasing magnitude of DM in Ethiopia, must be prioritized.

Systematic review registration

PROSPERO [CRD42020196815].

18-year change in serum intact fibroblast growth factor 23 from midlife to late life and risk of mortality: the ARIC Study

Objective

Fibroblast growth factor 23 (FGF23) concentration increases in response to declining kidney function to preserve normal phosphate concentrations. However, the etiological association of change in FGF23 concentration with mortality has not been examined in the general population.

Design and methods

We analyzed 5458 participants of the Atherosclerosis Risk in Communities Study who had intact FGF23 and estimated glomerular filtration rate (eGFR) assessed during midlife (visit 3, 1993-1995, mean age: 58 years) and late life (visit 5, 2011-2013, 76 years) to examine the association of FGF23 change over 18 years from mid-life to late life with the subsequent risk of mortality in late life using Cox regression models.

Results

The median 18-year change in intact FGF23 was +17.3 pg/mL. During a median follow-up of 7.2 years following visit 5, 1176 participants died. In multivariable Cox models, elevated mortality was seen in the highest quartile of FGF23 change (ΔFGF23: ≥31.3 pg/mL) (adjusted hazard ratio (aHR): 1.61 (95%CI: 1.36-1.90), or 1.37 (1.15-1.64) after additionally adjusting for eGFR change, compared with the lowest quartile (≤6.4 pg/mL)). When both FGF23 change and FGF23 in late life were simultaneously entered into the Cox model, FGF23 in late life, but not FGF23 change, was an independent predictor of mortality; however, we observed a high correlation between FGF23 change from midlife to late life and FGF23 in late life (r = 0.77).

Conclusions

Serum intact FGF23 change from midlife to late life was associated with subsequent risk of mortality independent of decline in kidney function. Our findings further support the implications of FGF23 beyond its association with kidney function.

Assessment of potential selection bias in neuroimaging studies of postoperative delirium and cognitive decline: lessons from the SAGES study

Due to cost and participant burden, neuroimaging studies are often performed in relatively small samples of voluntary participants. This may lead to selection bias. It is important to identify factors associated with participation in neuroimaging studies and understand their effect on outcome measures. We investigated the effect of postoperative delirium on long-term (over 48 months) cognitive decline (LTCD) in 560 older surgical patients (≥ 70 years), including a nested MRI cohort (n = 146). We observed a discrepancy in the effect of delirium on cognitive decline as a function of MRI participation. Although overall difference in cognitive decline due to delirium was not greater than what might be expected due to chance (p = .21), in the non-MRI group delirium was associated with a faster pace of LTCD (-0.063, 95% CI -0.094 to -0.032, p < .001); while in the MRI group the effect of delirium was less and not significant (-0.023, 95% CI -0.076, 0.030, p = .39). Since this limits our ability to investigate the neural correlates of delirium and cognitive decline using MRI data, we attempted to mitigate the observed discrepancy using inverse probability weighting for MRI participation. The approach was not successful and the difference of the effect of delirium in slope was essentially unchanged. There was no evidence that the MRI sub-group experienced delirium that differed in severity relative to MRI non-participants. We could not attribute the observed discrepancy to selection bias based on measured factors. It may reflect a power issue due to the smaller MRI subsample or selection bias from unmeasured factors.

A methodological framework for tackling confounding by indication when assessing the treatment effects of Chinese herbal injections in the real world

AIM

In the context of integrative medicine, whether Chinese herbal injections are effective in routine practice has become a question of broad interest. However, confounding by indication (i.e., indication bias) is a prevalent and highly challenging methodological issue when using routinely collected health care data to assess the real-world effectiveness of Chinese herbal injections.

METHODS AND RESULTS

We proposed a methodological approach to tackling confounding by indication in assessing the real-world effectiveness of Chinese herbal injections, incorporating empirical experiences, a literature review and interactive discussions, and a panel of external experts to finally achieve a consensus. This approach consisted of three cohesive steps, including a full understanding of treatment patterns, construction of fair comparisons by identifying appropriate combination treatments and comparators, and using statistical methods to further control for confounding. In the investigation of treatment patterns, we proposed five domains to identify treatment patterns with Chinese herbal injections, and we offered five patterns of combination treatments to characterize how Chinese herbal injections are used in conjunction with other treatments. In constructing fair comparisons, we suggested the use of both nonuse and active comparators; given the diverse combination treatments, we developed six scenarios that may form fair comparisons. In the statistical analysis, we discussed five statistical models for controlling confounding by indication, including their pros and cons. We also included a practical example to illustrate the usefulness of the methodological approach.

CONCLUSION

The proposed approach may serve as an effective tool to guide researchers to reliably assess the effectiveness of Chinese herbal injections in the context of integrative medicine.

Associations between multimorbidity and adverse health outcomes in UK Biobank and the SAIL Databank: A comparison of longitudinal cohort studies

BACKGROUND

Cohorts such as UK Biobank are increasingly used to study multimorbidity; however, there are concerns that lack of representativeness may lead to biased results. This study aims to compare associations between multimorbidity and adverse health outcomes in UK Biobank and a nationally representative sample.

METHODS AND FINDINGS

These are observational analyses of cohorts identified from linked routine healthcare data from UK Biobank participants (n = 211,597 from England, Scotland, and Wales with linked primary care data, age 40 to 70, mean age 56.5 years, 54.6% women, baseline assessment 2006 to 2010) and from the Secure Anonymised Information Linkage (SAIL) databank (n = 852,055 from Wales, age 40 to 70, mean age 54.2, 50.0% women, baseline January 2011). Multimorbidity (n = 40 long-term conditions [LTCs]) was identified from primary care Read codes and quantified using a simple count and a weighted score. Individual LTCs and LTC combinations were also assessed. Associations with all-cause mortality, unscheduled hospitalisation, and major adverse cardiovascular events (MACEs) were assessed using Weibull or negative binomial models adjusted for age, sex, and socioeconomic status, over 7.5 years follow-up for both datasets. Multimorbidity was less common in UK Biobank than SAIL (26.9% and 33.0% with ≥2 LTCs in UK Biobank and SAIL, respectively). This difference was attenuated, but persisted, after standardising by age, sex, and socioeconomic status. The association between increasing multimorbidity count and mortality, hospitalisation, and MACE was similar between both datasets at LTC counts of ≤3; however, above this level, UK Biobank underestimated the risk associated with multimorbidity (e.g., mortality hazard ratio for 2 LTCs 1.62 (95% confidence interval 1.57 to 1.68) in SAIL and 1.51 (1.43 to 1.59) in UK Biobank, hazard ratio for 5 LTCs was 3.46 (3.31 to 3.61) in SAIL and 2.88 (2.63 to 3.15) in UK Biobank). Absolute risk of mortality, hospitalisation, and MACE, at all levels of multimorbidity, was lower in UK Biobank than SAIL (adjusting for age, sex, and socioeconomic status). Both cohorts produced similar hazard ratios for some LTCs (e.g., hypertension and coronary heart disease), but UK Biobank underestimated the risk for others (e.g., alcohol-related disorders or mental health conditions). Hazard ratios for some LTC combinations were similar between the cohorts (e.g., cardiovascular conditions); however, UK Biobank underestimated the risk for combinations including other conditions (e.g., mental health conditions). The main limitations are that SAIL databank represents only part of the UK (Wales only) and that in both cohorts we lacked data on severity of the LTCs included.

CONCLUSIONS

In this study, we observed that UK Biobank accurately estimates relative risk of mortality, unscheduled hospitalisation, and MACE associated with LTC counts ≤3. However, for counts ≥4, and for some LTC combinations, estimates of magnitude of association from UK Biobank are likely to be conservative. Researchers should be mindful of these limitations of UK Biobank when conducting and interpreting analyses of multimorbidity. Nonetheless, the richness of data available in UK Biobank does offers opportunities to better understand multimorbidity, particularly where complementary data sources less susceptible to selection bias can be used to inform and qualify analyses of UK Biobank.

A nationwide cross-sectional survey of episiotomy practice in China

Background

A restrictive use of episiotomy was recommended by the obstetric society in China, but limited information on the epidemiology and the effects of restrictive use of episiotomy on maternal and fetal outcomes have been published.

Methods

The China Labor and Delivery Survey is a cross-sectional investigation with a cluster random sampling scheme in 2015-2016 with the aim to describe detailed epidemiology of labor and delivery in China. We calculated the episiotomy rate by parity using the survey data. Associated factors of episiotomy were selected using mixed models with a random effect for the hospital-level clustering. Logistic regression models were fitted to examine the effects of mediolateral episiotomy on maternal and neonatal outcomes among the propensity score-matched women.

Findings

The episiotomy rate among vaginal births was 41·7% [95% confidence interval (CI): 40·1% - 43·2%] in nulliparae and 21·5% (95% CI: 19·4% - 23·5%) in multiparae in China. Associated factors of episiotomy included prepregnancy BMI, maternal diseases, and obstetric factors. More than half of episiotomies in nulliparae and one-fourth in multiparae had no indications. Mediolateral episiotomy without indicators increased the risk of 3^rd or 4^th degree perineal laceration [odds ratio (OR) =2·64, 95% CI: 1·08-6·48)] in nulliparae without neonatal benefits.

Interpretation

Episiotomy was performed more than medically necessary in China. Mediolateral episiotomy without indications more than doubled the risk of 3^rd and 4^th degree perineal laceration in nulliparae without neonatal benefits.

Funding

National Natural Science Foundation of China and Shanghai Health Commission Programme.

Toward a better understanding about real-world evidence

BACKGROUND

There has been an interest in real-world evidence (RWE) in recent years. RWE is usually generated from data derived from routine healthcare, such as electronic healthcare records and disease registries. While RWE has many advantages, it is often open to various biases, which may distort results. Appropriate understanding and interpretation are critical to the best use of RWE in healthcare decisions.

METHODS

On the basis of a literature review and empirical research experience, we summarised the concept and methodological framework of RWE, and discussed in detail methodological issues specific to routinely collected healthcare data and observational studies using such data.

RESULTS

RWE is derived from a spectrum of data generated from the real-world setting, using two broad study designs including observational studies and pragmatic clinical trials. Real-world data may usually be collected through routine practice or sometimes actively collected with a research purpose. Observational studies using routinely collected data (RCD) are the most common type of RWE, although they are prone to biases. When planning and implementing RWE studies, coherent working steps are warranted, including definition of a clear and answerable research question, development of a research team, selection of a fit-for-purpose data source, choice of state-of-the-art study design, establishing a database with transparent data processing, performing multiple statistical analysis to control bias, and reporting results in accordance with established guidelines.

CONCLUSIONS

RWE has been mounting over the years. The appropriate interpretation and use of such evidence often warrant adequate understanding about methodology. Researchers and policymakers should be aware of the methodological pitfalls when generating and interpreting RWE.

Current challenges of severe acute respiratory syndrome coronavirus 2 seroprevalence studies among blood donors: A scoping review

BACKGROUND AND OBJECTIVES

Blood donors are increasingly being recognized as an informative resource for surveillance. We aimed to review severe acute respiratory syndrome coronavirus 2 seroprevalence studies conducted among blood donors to investigate methodological biases and provide guidance for future research.

MATERIALS AND METHODS

We conducted a scoping review of peer-reviewed and preprint publications between January 2020 and January 2021. Two reviewers used standardized forms to extract seroprevalence estimates and data on methodology pertaining to population sampling, periodicity, assay characteristics, and antibody kinetics. National data on cumulative incidence and social distancing policies were extracted from publicly available sources and summarized.

RESULTS

Thirty-three studies representing 1,323,307 blood donations from 20 countries worldwide were included (sample sizes ranged from 22 to 953,926 donations). The majority of the studies (79%) reported seroprevalence rates <10% (ranging from 0% to 76% [after adjusting for waning antibodies]). Overall, less than 1 in 5 studies reported standardized seroprevalence rates to reflect the demographics of the general population. Stratification by age and sex were most common (64% of studies), followed by region (48%). A total of 52% of studies reported seroprevalence at a single time point. Overall, 27 unique assay combinations were identified, 55% of studies used a single assay and only 39% adjusted seroprevalence rates for imperfect test characteristics. Among the nationally representative studies, case detection was most underrepresented in Kenya (1:1264).

CONCLUSION

By the end of 2020, seroprevalence rates were far from reaching herd immunity. In addition to differences in community transmission and diverse public health policies, study designs and methodology were likely contributing factors to seroprevalence heterogeneity.

Factors Associated With Advanced Colorectal Neoplasia in Patients With CKD

RATIONALE & OBJECTIVE

The risk of developing colorectal cancer in patients with chronic kidney disease (CKD) is twice that of the general population, but the factors associated with colorectal cancer are poorly understood. The aim of this study was to identify factors associated with advanced colorectal neoplasia in patients with CKD.

STUDY DESIGN

Prospective cohort study.

SETTING & PARTICIPANTS

Patients with CKD stages 3-5, including those treated with maintenance dialysis or transplantation across 11 sites in Australia, New Zealand, Canada, and Spain, were screened for colorectal neoplasia using a fecal immunochemical test (FIT) as part of the Detecting Bowel Cancer in CKD (DETECT) Study.

EXPOSURE

Baseline characteristics for patients at the time of study enrollment were ascertained, including duration of CKD, comorbidities, and medications.

OUTCOME

Advanced colorectal neoplasia was identified through a 2-step verification process with colonoscopy following positive FIT and 2-year clinical follow-up for all patients.

ANALYTICAL APPROACH

Potential factors associated with advanced colorectal neoplasia were explored using multivariable logistic regression. Sensitivity analyses were performed using grouped LASSO (least absolute shrinkage and selection operator) logistic regression.

RESULTS

Among 1,706 patients who received FIT-based screening-791 with CKD stages 3-5 not receiving kidney replacement therapy (KRT), 418 receiving dialysis, and 497 patients with a functioning kidney transplant-117 patients (6.9%) were detected to have advanced colorectal neoplasia (54 with CKD stages 3-5 without KRT, 34 receiving dialysis, and 29 transplant recipients), including 9 colorectal cancers. The factors found to be associated with advanced colorectal neoplasia included older age (OR per year older, 1.05 [95% CI, 1.03-1.07], P<0.001), male sex (OR, 2.27 [95% CI, 1.45-3.54], P<0.001), azathioprine use (OR, 2.99 [95% CI, 1.40-6.37], P=0.005), and erythropoiesis-stimulating agent use (OR, 1.92 [95% CI, 1.22-3.03], P=0.005). Grouped LASSO logistic regression revealed similar associations between these factors and advanced colorectal neoplasia.

LIMITATIONS

Unmeasured confounding factors.

CONCLUSIONS

Older age, male sex, erythropoiesis-stimulating agents, and azathioprine were found to be significantly associated with advanced colorectal neoplasia in patients with CKD.

The effect of home visits as an additional recruitment step on the composition of the final sample: a cross-sectional analysis in two study centers of the German National Cohort (NAKO)

Background

Participation in epidemiologic studies has been declining over the last decades. In addition to postal invitations and phone calls, home visits can be conducted to increase participation. The aim of this study was therefore to evaluate the effects of home visits in terms of response increase and composition of the additionally recruited and final sample.

Methods

In the framework of the German National Cohort (NAKO) recruitment process, two of 18 study centers, Halle (Saale) and Berlin-Center, performed home visits as additional recruitment step after postal invitation and reminders. Response increase was calculated and differences between participants recruited via home visits and standard recruitment were examined. Proportions are presented as percentages with 95%-confidence intervals.

Results

In the general population in Halle, 21.3-22.8% participated after postal invitation and two reminders in the five assessed recruitment waves. The increase of the overall response was 2.8 percentage points (95%confidence interval: 1.9-4.0) for home visits compared to 2.4 percentage points (95%CI: 1.7-3.3) for alternatively sent third postal reminder. Participants recruited via home visits had similar characteristics to those recruited via standard recruitment. Among persons of Turkish descent in Berlin-Center site of the NAKO, home visits conducted by native speakers increased the participation of women, persons living together with their partner, were born in Turkey, had lower German language skills, lower-income, lower education, were more often smokers and reported more often diabetes and depression to a degree which changed overall estimates for this subsample.

Conclusions

As an additional recruitment measure in the general population, home visits increased response only marginally, and the through home visits recruited participants did not differ from those already recruited. Among persons with migration background, home visits by a native speaker increased participation of persons not reached by the standard recruitment, but the effects of using a native speaker approach could not be separated from the effect of home visits.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12874-021-01357-z.

Uncovering survivorship bias in longitudinal mental health surveys during the COVID-19 pandemic

AIMS

Markedly elevated adverse mental health symptoms were widely observed early in the coronavirus disease-2019 (COVID-19) pandemic. Unlike the U.S., where cross-sectional data indicate anxiety and depression symptoms have remained elevated, such symptoms reportedly declined in the U.K., according to analysis of repeated measures from a large-scale longitudinal study. However, nearly 40% of U.K. respondents (those who did not complete multiple follow-up surveys) were excluded from analysis, suggesting that survivorship bias might partially explain this discrepancy. We therefore sought to assess survivorship bias among participants in our longitudinal survey study as part of The COVID-19 Outbreak Public Evaluation (COPE) Initiative.

METHODS

Survivorship bias was assessed in 4039 U.S. respondents who completed surveys including the assessment of mental health as part of The COPE Initiative in April 2020 and were invited to complete follow-up surveys. Participants completed validated screening instruments for symptoms of anxiety, depression and insomnia. Survivorship bias was assessed for (1) demographic differences in follow-up survey participation, (2) differences in initial adverse mental health symptom prevalence adjusted for demographic factors and (3) differences in follow-up survey participation based on mental health experiences adjusted for demographic factors.

RESULTS

Adjusting for demographics, individuals who completed only one or two out of four surveys had significantly higher prevalence of anxiety and depression symptoms in April 2020 (e.g. one-survey v. four-survey, anxiety symptoms, adjusted prevalence ratio [aPR]: 1.30, 95% confidence interval [CI]: 1.08-1.55, p = 0.0045; depression symptoms, aPR: 1.43, 95% CI: 1.17-1.75, p = 0.00052). Moreover, individuals who experienced incident anxiety or depression symptoms had significantly higher adjusted odds of not completing follow-up surveys (adjusted odds ratio [aOR]: 1.68, 95% CI: 1.22-2.31, p = 0.0015, aOR: 1.56, 95% CI: 1.15-2.12, p = 0.0046, respectively).

CONCLUSIONS

Our findings reveal significant survivorship bias among longitudinal survey respondents, indicating that restricting analytic samples to only respondents who provide repeated assessments in longitudinal survey studies could lead to overly optimistic interpretations of mental health trends over time. Cross-sectional or planned missing data designs may provide more accurate estimates of population-level adverse mental health symptom prevalence than longitudinal surveys.

Gene-environment dependencies lead to collider bias in models with polygenic scores

The application of polygenic scores has transformed our ability to investigate whether and how genetic and environmental factors jointly contribute to the variation of complex traits. Modelling the complex interplay between genes and environment, however, raises serious methodological challenges. Here we illustrate the largely unrecognised impact of gene-environment dependencies on the identification of the effects of genes and their variation across environments. We show that controlling for heritable covariates in regression models that include polygenic scores as independent variables introduces endogenous selection bias when one or more of these covariates depends on unmeasured factors that also affect the outcome. This results in the problem of conditioning on a collider, which in turn leads to spurious associations and effect sizes. Using graphical and simulation methods we demonstrate that the degree of bias depends on the strength of the gene-covariate correlation and of hidden heterogeneity linking covariates with outcomes, regardless of whether the main analytic focus is mediation, confounding, or gene × covariate (commonly gene × environment) interactions. We offer potential solutions, highlighting the importance of causal inference. We also urge further caution when fitting and interpreting models with polygenic scores and non-exogenous environments or phenotypes and demonstrate how spurious associations are likely to arise, advancing our understanding of such results.

A systematic review of quantitative bias analysis applied to epidemiological research

BACKGROUND

Quantitative bias analysis (QBA) measures study errors in terms of direction, magnitude and uncertainty. This systematic review aimed to describe how QBA has been applied in epidemiological research in 2006-19.

METHODS

We searched PubMed for English peer-reviewed studies applying QBA to real-data applications. We also included studies citing selected sources or which were identified in a previous QBA review in pharmacoepidemiology. For each study, we extracted the rationale, methodology, bias-adjusted results and interpretation and assessed factors associated with reproducibility.

RESULTS

Of the 238 studies, the majority were embedded within papers whose main inferences were drawn from conventional approaches as secondary (sensitivity) analyses to quantity-specific biases (52%) or to assess the extent of bias required to shift the point estimate to the null (25%); 10% were standalone papers. The most common approach was probabilistic (57%). Misclassification was modelled in 57%, uncontrolled confounder(s) in 40% and selection bias in 17%. Most did not consider multiple biases or correlations between errors. When specified, bias parameters came from the literature (48%) more often than internal validation studies (29%). The majority (60%) of analyses resulted in >10% change from the conventional point estimate; however, most investigators (63%) did not alter their original interpretation. Degree of reproducibility related to inclusion of code, formulas, sensitivity analyses and supplementary materials, as well as the QBA rationale.

CONCLUSIONS

QBA applications were rare though increased over time. Future investigators should reference good practices and include details to promote transparency and to serve as a reference for other researchers.

Uncovering Survivorship Bias in Longitudinal Mental Health Surveys During the COVID-19 Pandemic

Aims

Markedly elevated adverse mental health symptoms were widely observed early in the coronavirus disease 2019 (COVID-19) pandemic. Unlike the U.S., where cross-sectional data indicate anxiety and depression symptoms have remained elevated, such symptoms reportedly declined in the U.K., according to analysis of repeated measures from a largescale longitudinal study. However, nearly 40% of U.K. respondents (those who did not complete multiple follow-up surveys) were excluded from analysis, suggesting that survivorship bias might partially explain this discrepancy. We therefore sought to assess survivorship bias among participants in our longitudinal survey study as part of The COVID-19 Outbreak Public Evaluation (COPE) Initiative.

Methods

Survivorship bias was assessed 4,039 U.S. respondents who completed surveys including the assessment of mental health as part of The COPE Initiative in April 2020 and were invited to complete follow-up surveys. Participants completed validated screening instruments for symptoms of anxiety, depression, and insomnia. Survivorship bias was assessed for (1) demographic differences in follow-up survey participation, (2) differences in initial adverse mental health symptom prevalences adjusted for demographic factors, and (3) differences in follow-up survey participation based on mental health experiences adjusted for demographic factors.

Results

Adjusting for demographics, individuals who completed only one or two out of four surveys had higher prevalences of anxiety and depression symptoms in April 2020 (e.g., one-survey versus four-survey, anxiety symptoms, adjusted prevalence ratio [aPR]: 1.30, 95% confidence interval [CI]: 1.08–1.55, P=0.0045; depression symptoms, aPR: 1.43, 95% CI: 1.17–1.75, P=0.00052). Moreover, individuals who experienced incident anxiety or depression symptoms had higher odds of not completing follow-up surveys (adjusted odds ratio [aOR]: 1.68, 95% CI: 1.22–2.31, P=0.0015, aOR: 1.56, 95% CI: 1.15–2.12, P=0.0046, respectively).

Conclusions

Our findings revealed significant survivorship bias among longitudinal survey respondents, indicating that restricting analytic samples to only respondents who provide repeated assessments in longitudinal survey studies could lead to overly optimistic interpretations of mental health trends over time. Cross-sectional or planned missing data designs may provide more accurate estimates of population-level adverse mental health symptom prevalences than longitudinal surveys.

A French multicentric prospective prognostic cohort with epidemiological, clinical, biological and treatment information to improve knowledge on lymphoma patients: study protocol of the "REal world dAta in LYmphoma and survival in adults" (REALYSA) cohort

Background

Age-adjusted lymphoma incidence rates continue to rise in France since the early 80’s, although rates have slowed since 2010 and vary across subtypes. Recent improvements in patient survival in major lymphoma subtypes at population level raise new questions about patient outcomes (i.e. quality of life, long-term sequelae). Epidemiological studies have investigated factors related to lymphoma risk, but few have addressed the extent to which socioeconomic status, social institutional context (i.e. healthcare system), social relationships, environmental context (exposures), individual behaviours (lifestyle) or genetic determinants influence lymphoma outcomes, especially in the general population. Moreover, the knowledge of the disease behaviour mainly obtained from clinical trials data is partly biased because of patient selection.

Methods

The REALYSA (“REal world dAta in LYmphoma and Survival in Adults”) study is a real-life multicentric cohort set up in French areas covered by population-based cancer registries to study the prognostic value of epidemiological, clinical and biological factors with a prospective 9-year follow-up. We aim to include 6000 patients over 4 to 5 years. Adult patients without lymphoma history and newly diagnosed with one of the following 7 lymphoma subtypes (diffuse large B-cell, follicular, marginal zone, mantle cell, Burkitt, Hodgkin, mature T-cell) are invited to participate during a medical consultation with their hematologist. Exclusion criteria are: having already received anti-lymphoma treatment (except pre-phase) and having a documented HIV infection. Patients are treated according to the standard practice in their center. Clinical data, including treatment received, are extracted from patients’ medical records. Patients’ risk factors exposures and other epidemiological data are obtained at baseline by filling out a questionnaire during an interview led by a clinical research assistant. Biological samples are collected at baseline and during treatment. A virtual tumor biobank is constituted for baseline tumor samples. Follow-up data, both clinical and epidemiological, are collected every 6 months in the first 3 years and every year thereafter.

Discussion

This cohort constitutes an innovative platform for clinical, biological, epidemiological and socio-economic research projects and provides an opportunity to improve knowledge on factors associated to outcome of lymphoma patients in real life.

Trial registration

2018-A01332–53, ClinicalTrials.gov identifier: NCT03869619.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12889-021-10433-4.

Quantification of selection bias in studies of risk factors for birth defects among livebirths

BACKGROUND

Risk factors for birth defects are frequently investigated using data limited to liveborn infants. By conditioning on survival, results of such studies may be distorted by selection bias, also described as "livebirth bias." However, the implications of livebirth bias on risk estimation remain poorly understood.

OBJECTIVES

We sought to quantify livebirth bias and to investigate the conditions under which it arose.

METHODS

We used data on 3994 birth defects cases and 11 829 controls enrolled in the National Birth Defects Prevention Study to compare odds ratio (OR) estimates of the relationship between three established risk factors (antiepileptic drug use, smoking, and multifetal pregnancy) and four birth defects (anencephaly, spina bifida, omphalocele, and cleft palate) when restricted to livebirths as compared to among livebirths, stillbirths, and elective terminations. Exposures and birth defects represented varying strengths of association with livebirth; all controls were liveborn. We performed a quantitative bias analysis to evaluate the sensitivity of our results to excluding terminated and stillborn controls.

RESULTS

Cases ranged from 33% liveborn (anencephaly) to 99% (cleft palate). Smoking and multifetal pregnancy were associated with livebirth among anencephaly (crude OR [cOR] 0.61 and cOR 3.15, respectively) and omphalocele cases (cOR 2.22 and cOR 5.22, respectively). For analyses of the association between exposures and birth defects, restricting to livebirths produced negligible differences in estimates except for anencephaly and multifetal pregnancy, which was twofold higher among livebirths (adjusted OR [aOR] 4.93) as among all pregnancy outcomes (aOR 2.44). Within tested scenarios, bias analyses suggested that results were not sensitive to the restriction to liveborn controls.

CONCLUSIONS

Selection bias was generally limited except for high mortality defects in the context of exposures strongly associated with livebirth. Findings indicate that substantial livebirth bias is unlikely to affect studies of risk factors for most birth defects.

Does the association between physical capability and mortality differ by deprivation? Findings from the UK Biobank population-based cohort study

BACKGROUND

To investigate whether the excess risk of adverse health outcomes associated with a lower physical capability in adulthood differs by deprivation levels.

METHODS

279,030 participants from the UK Biobank were included. Handgrip strength and walking pace were the exposures. All-cause mortality, CVD mortality and incidence were the outcomes. Townsend deprivation index was treated as a potential effect modifier. The associations were investigated using Cox-regression models with years of follow-up as the time-varying covariate.

RESULTS

A significant interaction between deprivation and handgrip strength was found for all-cause mortality (p = 0.024), CVD mortality (p = 0.006) and CVD incidence (p = 0.001). The hazard ratio for all-cause mortality was 1.18 [1.09; 1.29] per 1-tertile higher level of grip strength in the least deprived group, whereas it was 1.30 [1.18; 1.43] in the most deprived individuals. Similar results were found for CVD mortality and incidence per tertile increment in handgrip strength in the least and most deprived quintiles, respectively. No significant interactions between deprivation and walking pace were found for any of the outcomes.

CONCLUSION

Low handgrip strength is a stronger predictor of morbidity and mortality in individuals living in more deprived areas.

The effectiveness of conservative interventions for the management of syndromic hypermobility: a systematic literature review

Introduction

‘Syndromic hypermobility’ encompasses heritable connective tissue disorders such as hypermobile Ehlers-Danlos syndrome and hypermobility spectrum disorders which are characterised by excessive joint range of motion and pain. Conservative interventions such as exercise are the cornerstone of management, yet their effectiveness is unclear.

Aim

To systematically appraise the effectiveness of conservative management for people with syndromic hypermobility.

Method

A systematic online database search was conducted (AMED, BND, CINAHL Plus, MEDLINE, PEDro, PsychINFO and SportDiscus). Potential articles were assessed for eligibility by two researchers against the following criteria: adults and children with a hEDS/HSD diagnosis (or equivalent diagnosis using specific criteria); non-pharmacological or non-surgical interventions; outcomes related to pain, physical function, psychological well-being or quality of life. Controlled trials and cohort studies were included. Critical Appraisal Skills Programme checklists were used to assess methodological quality.

Results

Eleven studies were included, comprising eight controlled trials and three cohort studies. All studies investigated interventions that had exercise as the primary component. Three small controlled studies demonstrated superior effects of conservative management relative to a control group. However, those studies only focused on a single area of the body, only recruited women, and had no long-term follow-up. All studies reported improvements in a wide range of outcomes over time.

Conclusion

Controlled trial evidence for the superiority of conservative management over comparators is weak. There is some evidence that people improve over time. Robust randomised controlled trial research of the long-term effectiveness of ‘whole-body’ (rather than individual joints or body areas) conservative management is required.

Key Points • Conservative management is the cornerstone of management of syndromic hypermobility. • The review found that evidence for the effectiveness of conservative management relative to no treatment or other conservative comparators was weak. • However, there was consistent evidence for effectiveness from pre- to post-treatment. • Further robust randomised controlled trial evidence is required.

A novel framework for classification of selection processes in epidemiological research

BACKGROUND

Selection and selection bias are terms that lack consistent definitions and have varying meaning and usage across disciplines. There is also confusion in current definitions between underlying mechanisms that lead to selection and their consequences. Consequences of selection on study validity must be judged on a case-by-case basis depending on research question, study design and analytical decisions. The overall aim of the study was to develop a simple but general framework for classifying various types of selection processes of relevance for epidemiological research.

METHODS

Several original articles from the epidemiological literature and from related areas of observational research were reviewed in search of examples of selection processes, used terminology and description of the underlying mechanisms.

RESULTS

We classified the identified selection processes in three dimensions: i) selection level (selection at the population level vs. study-specific selection), ii) type of mechanism (selection in exposure vs. selection in population composition), iii) timing of the selection (at exposure entry, during exposure/follow-up or post-outcome).

CONCLUSIONS

Increased understanding of when, how, and why selection occur is an important step towards improved validity of epidemiological research.

Extrapulmonary small cell carcinoma: Prognostic factors, patterns of care, and overall survival

BACKGROUND

Extrapulmonary small cell carcinoma is a rare malignancy with scarce data. Treatment paradigms extrapolate from pulmonary small cell carcinoma and single institution studies. We analyzed the epidemiology, patterns of care, prognostic factors, and overall survival (OS) of EPSCC patients.

METHODS

The cohort included EPSCC. Patients with <2 months follow-up, unknown demographic/treatment variables were excluded. Descriptive statistics were performed to characterize the cohort. Kaplan-Meier methods were used to estimate OS. Cox proportional hazard modeling was done to analyze the influence of prognostic variables on OS.

RESULTS

5747 patients were included. Median OS was 1.2 years. Head and neck (HR: 0.60, 95% CI 0.53-0.67, p < 0.0001) and breast (HR: 0.69, 95% CI 0.53-0.89, p = .0046) were associated with improved OS; gastrointestinal (HR: 1.19, 95% CI 1.09-1.29, p < .0001) worse OS; and gynecologic (HR: 1.04, 95% CI 0.92-1.17, p = .5660) showing no difference, all compared to genitourinary (reference). Surgery was associated with improved overall survival (HR: 0.84, 95% CI 0.79-0.91, p=<.0001). Chemoradiation showed a decreased HR (HR: 0.91, 95% CI 0.83-0.99, p = .0363) when compared to chemotherapy alone (reference).

CONCLUSION

EPSCC occurs throughout the body with poor survival. Anatomic subsite was predictive for survival. Surgical resection may improve survival. Concurrent chemoradiation appears to improve survival over chemotherapy alone.

Long-term use of hydrocodone vs. oxycodone in primary care

BACKGROUND

Hydrocodone and oxycodone are the Schedule II opioids most often prescribed in primary care. Notwithstanding the dangers of prescription opioid use, the likelihood of long-term use with either drug is presently unknown.

METHODS

Using a retrospective cohort design and data from a commerical healthcare claims repository, we compared the likelihood of long-term use of hydrocodone and oxycodone in primary care patients presenting with acute back pain. Treatment was categorized as long-term if the prescription dates spanned ≥90 days from initial prescription to the run-out date of the last prescription, and included ≥120 days' supply or ≥10 fills. Instrumental variable methods and probit regression were used to model the effect of drug choice on long-term use, estimate the average treatment effect, and correct for confounding by indication.

RESULTS

A total of 3,983 patients who were prescribed only hydrocodone or only oxycodone were followed for 270 days in 2016. Long-term opioid use was observed in 320 patients (8%). Controlling for potential confounders including morphine milligram equivalents and dosage, an estimated 12% (95 CI, 10%-14%) treated with hydrocodone transitioned to long-term use vs. 2% (95 CI, 1%-3%) on oxycodone. Among patients who received more than one prescription (n = 1,866), an estimated 23% (95 CI, 19%-26%) treated with hydrocodone transitioned to long-term use vs. 5% (95 CI, 3%-7%) on oxycodone. The difference between drugs was supported in sensitivity and subgroup analyses. Sample selection bias was not detected.

CONCLUSIONS

Long-term use was substantially greater for patients treated with hydrocodone than oxycodone, despite equianalgesia.

Generalizing causal inferences from individuals in randomized trials to all trial-eligible individuals

We consider methods for causal inference in randomized trials nested within cohorts of trial-eligible individuals, including those who are not randomized. We show how baseline covariate data from the entire cohort, and treatment and outcome data only from randomized individuals, can be used to identify potential (counterfactual) outcome means and average treatment effects in the target population of all eligible individuals. We review identifiability conditions, propose estimators, and assess the estimators' finite-sample performance in simulation studies. As an illustration, we apply the estimators in a trial nested within a cohort of trial-eligible individuals to compare coronary artery bypass grafting surgery plus medical therapy vs. medical therapy alone for chronic coronary artery disease.

Commentary: Causal Inference for Social Exposures

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