Bias In Observational Studies Research Articles

Confounding and bias in observational studies in inflammatory bowel disease: a meta‐epidemiological study

SummaryBackgroundObservational research concerning inflammatory bowel disease (IBD) is highly susceptible to spurious findings because of confounding and bias.AimTo investigate how these issues were reported in this research field.MethodsWe identified and appraised a random sample of 160 observational studies concerning IBD published in high‐impact gastroenterology journals and the most respected specialty journals of the condition. We applied a standardised methodology to assess how confounding and bias were reported and discussed, and investigated the association between yearly citations and study characteristics using mixed‐effect multivariable regression analysis.ResultsThe authors of 67 out of 160 articles (41.9%) mentioned confounding, and in 89 cases (55.6%) reported any bias. Although most authors applied strategies to minimise confounding or bias (n = 139; 86.9%) and acknowledged at least one unadjusted confounder (n = 116; 72.5%), a minority commented about whether the main findings could have been affected (n = 60; 37.5%). Very few authors (n = 7; 4.4%) called for caution in interpreting the results in the discussion. Reporting of confounding and bias was particularly lacking for case‐control studies, those not using routinely collected data, those employing laboratory analyses as the primary method of assessment and studies investigating non‐modifiable exposures. In adjusted analyses, mentioning or alluding to confounding was positively associated with yearly citations (P = 0.010), whereas calling for a cautious interpretation of the findings was not.ConclusionsReporting of confounding is inadequate and its acknowledgement is often neglected in interpreting high‐impact observational research in IBD. These results encourage a more careful evaluation of the consequences of confounding and bias.

Can statistical adjustment guided by causal inference improve the accuracy of effect estimation? A simulation and empirical research based on meta-analyses of case\u2013control studies

BackgroundStatistical adjustment is often considered to control confounding bias in observational studies, especially case–control studies. However, different adjustment strategies may affect the estimation of odds ratios (ORs), and in turn affect the results of their pooled analyses. Our study is aimed to investigate how to deal with the statistical adjustment in case–control studies to improve the validity of meta-analyses.MethodsThree types of adjustment strategies were evaluated including insufficient adjustment (not all preset confounders were adjusted), full adjustment (all confounders were adjusted under the guidance of causal inference), and improper adjustment (covariates other than confounders were adjusted). We carried out a series of Monte Carlo simulation experiments based on predesigned scenarios, and assessed the accuracy of effect estimations from meta-analyses of case–control studies by combining ORs calculated according to different adjustment strategies. Then we used the data from an empirical review to illustrate the replicability of the simulation results.ResultsFor all scenarios with different strength of causal relations, combining ORs that were comprehensively adjusted for confounders would get the most precise effect estimation. By contrast, combining ORs that were not sufficiently adjusted for confounders or improperly adjusted for mediators or colliders would easily introduce bias in causal interpretation, especially when the true effect of exposure on outcome was weak or none. The findings of the simulation experiments were further verified by the empirical research.ConclusionsStatistical adjustment guided by causal inference are recommended for effect estimation. Therefore, when conducting meta-analyses of case–control studies, the causal relationship formulated by exposure, outcome, and covariates should be firstly understood through a directed acyclic graph, and then reasonable original ORs could be extracted and combined by suitable methods.

Evaluating sources of bias in observational studies of angiotensin-converting enzyme inhibitor/angiotensin II receptor blocker use during COVID-19: beyond confounding.

Concerns over ACE inhibitor or ARB use to treat hypertension during COVID-19 remain unresolved. Although studies using more robust methodologies provided some clarity, sources of bias persist and it remains critical to quickly address this question. In this review, we discuss pernicious sources of bias using a causal model framework, including time-varying confounder, collider, information, and time-dependent bias, in the context of recently published studies. We discuss causal inference methodologies that can address these issues, including causal diagrams, time-to-event analyses, sensitivity analyses, and marginal structural modeling. We discuss effect modification and we propose a role for causal mediation analysis to estimate indirect effects via mediating factors, especially components of the renin--angiotensin system. Thorough knowledge of these sources of bias and the appropriate methodologies to address them is crucial when evaluating observational studies to inform patient management decisions regarding whether ACE inhibitors or ARBs are associated with greater risk from COVID-19.

Causal graphical views of fixed effects and random effects models.

Despite the long-standing discussion on fixed effects (FE) and random effects (RE) models, how and under what conditions both methods can eliminate unmeasured confounding bias has not yet been widely understood in practice. Using a simple pretest-posttest design in a linear setting, this paper translates the conventional algebraic formalization of FE and RE models into causal graphs and provides intuitively accessible graphical explanations about their data-generating and bias-removing processes. The proposed causal graphs highlight that FE and RE models consider different data-generating models. RE models presume a data-generating model that is identical to a randomized controlled trial, while FE models allow for unobserved time-invariant treatment-outcome confounding. Augmenting regular causal graphs that describe data-generating processes by adding the computational structures of FE and RE estimators, the paper visualizes how FE estimators (gain score and deviation score estimators) and RE estimators (quasi-deviation score estimators) offset unmeasured confounding bias. In contrast to standard regression or matching estimators that reduce confounding bias by blocking non-causal paths via conditioning, FE and RE estimators offset confounding bias by deliberately creating new non-causal paths and associations of opposite sign. Though FE and RE estimators are similar in their bias-offsetting mechanisms, the augmented graphs reveal their subtle differences that can result in different biases in observational studies.

Dental Implant Failure Rate and Marginal Bone Loss in Transplanted Patients: A Systematic Review and Meta-Analysis

This systematic review investigates the failure rate and marginal bone loss (MBL) of dental implants placed in patients undergoing solid-organ transplant (SOT) compared to healthy controls. Three databases (PubMed, Web of Sciences, and the Cochrane Library) were searched up to June 2020 (PROSPERO CRD42019124896). Case-control and cohort studies reporting data failure rate and marginal bone loss (MBL) of dental implants placed in SOT patients were included. The risk of bias of observational studies was assessed through the Newcastle-Ottawa Scale (NOS). Four case-control studies fulfilled the inclusion criteria; all had low risk of bias. Meta-analyses revealed consistently lower implant failure rate than control populations at patient and implant levels. SOT patients had a significant difference of −18% (p-value < 0.001) MBL compared to healthy patients. SOT status poses no serious threat to implant survival. Overall, this group of patients presented lower levels of dental implant failure rate and marginal bone loss compared to otherwise healthy patients. Further intervention trials with larger sample size and longer follow-ups are necessary to confirm these summarized results.

Assessing risk of bias: a proposal for a unified framework for observational studies and randomized trials

BackgroundEvidence based medicine aims to integrate scientific evidence, clinical experience, and patient values and preferences. Individual health care professionals need to appraise the evidence from randomized trials and observational studies when guidelines are not yet available. To date, tools for assessment of bias and terminologies for bias are specific for each study design. Moreover, most tools appeal only to methodological knowledge to detect bias, not to subject matter knowledge, i.e. in-depth medical knowledge about a topic. We propose a unified framework that enables the coherent assessment of bias across designs.MethodsEpidemiologists traditionally distinguish between three types of bias in observational studies: confounding, information bias, and selection bias. These biases result from a common cause, systematic error in the measurement or common effect of the intervention and outcome respectively. We applied this conceptual framework to randomized trials and show how it can be used to identify bias. The three sources of bias were illustrated with graphs that visually represent researchers’ assumptions about the relationships between the investigated variables (causal diagrams).ResultsCritical appraisal of evidence started with the definition of the research question in terms of the population of interest, the compared interventions and the main outcome. Next, we used causal diagrams to illustrate how each source of bias can lead to over- or underestimated treatment effects. Then, we discussed how randomization, blinded outcome measurement and intention-to-treat analysis minimize bias in trials. Finally, we identified study aspects that can only be appraised with subject matter knowledge, irrespective of study design.ConclusionsThe unified framework encompassed the three main sources of bias for the effect of an assigned intervention on an outcome. It facilitated the integration of methodological and subject matter knowledge in the assessment of bias. We hope that graphical diagrams will help clarify debate among professionals by reducing misunderstandings based on different terminology for bias.

Selection Bias in Observational Studies of Palliative Care: Lessons Learned

ContextPalliative care (PC) programs are typically evaluated using observational data, raising concerns about selection bias. ObjectivesTo quantify selection bias because of observed and unobserved characteristics in a PC demonstration program. MethodsProgram administrative data and 100% Medicare claims data in two states and a 20% sample in eight states (2013–2017). The sample included 2983 Medicare fee-for-service beneficiaries aged 65+ participating in the PC program and three matched cohorts: regional; two states; and eight states. Confounding because of observed factors was measured by comparing patient baseline characteristics. Confounding because of unobserved factors was measured by comparing days of follow-up and six-month and one-year mortality rates. ResultsAfter matching, evidence for observed confounding included differences in observable baseline characteristics, including race, morbidity, and utilization. Evidence for unobserved confounding included significantly longer mean follow-up in the regional, two-state, and eight-state comparison cohorts, with 207 (P < 0.001), 192 (P < 0.001), and 187 (P < 0.001) days, respectively, compared with the 162 days for the PC cohort. The PC cohort had higher six-month and one-year mortality rates of 53.5% and 64.5% compared with 43.5% and 48.0% in the regional comparison, 53.4% and 57.4% in the two-state comparison, and 55.0% and 59.0% in the eight-state comparison. ConclusionThis case study demonstrates that selection of comparison groups impacts the magnitude of measured and unmeasured confounding, which may change effect estimates. The substantial impact of confounding on effect estimates in this study raises concerns about the evaluation of novel serious illness care models in the absence of randomization. We present key lessons learned for improving future evaluations of PC using observational study designs.

Evidence-based approaches to comparing the effectiveness of modern cardiology interventions: trends, bias and prospects

The article discusses the limitations of the evidence from observational studies. Modern approaches to reducing bias in observational studies are discussed in detail, in particular, propensity score matching, which has become popular in recent years. The main differences between randomized and observational studies are discussed. Arguments against the observational studies and improved methods of analysis to compare the treatments’ effectiveness in clinical practice are presented. The role of observational studies as a source of evidence is discussed. The article points out the validity of performing large-scale prospective observational studies to assess the effects of postmarketing drug use in clinical practice, as well as to obtain data on drug use in patients that differ from those in randomized clinical trials.

To use or not to use propensity score matching?

Propensity score matching (PSM) has been widely used to reduce confounding biases in observational studies. Its properties for statistical inference have also been investigated and well documented. However, some recent publications showed concern of using PSM, especially on increasing postmatching covariate imbalance, leading to discussion on whether PSM should be used or not. We review empirical and theoretical evidence for and against its use in practice and revisit the property of equal percent bias reduction and adapt it to more practical situations, showing that PSM has some additional desirable properties. With a small simulation, we explore the impact of caliper width on biases due to mismatching in matched samples and due to the difference between matched and target populations and show some issue of PSM may be due to inadequate caliper selection. In summary, we argue that the right question should be when and how to use PSM rather than to use or not to use it and give suggestions accordingly.

Anticoagulation, immortality, and observations of COVID‐19

Anticoagulation, immortality, and observations of COVID‐19

Measuring dementia incidence within a cohort of 267,153 older Australians using routinely collected linked administrative data

To estimate dementia incidence rates using Australian administrative datasets and compare the characteristics of people identified with dementia across different datasets. This data linkage study used a cohort of 267,153 from the Australian 45 and Up Study. Participants completed a survey in 2006–2009 and subsequent dementia was identified through pharmaceutical claims, hospitalisations, aged care eligibility assessments, care needs at residential aged care entry and death certificates. Age-specific, and age-standardised incidence rates, incidence rate ratios and survival from first dementia diagnosis were estimated. Estimated age-standardised dementia incidence rates using all linked datasets was 16.8 cases per 1000 person years for people aged 65+. Comparing incidence rates to the global published rates suggested 77% of cases were identified but this varied by age with highest coverage among those aged 80–84 years (92%). Incidence rate ratios were inconsistent across datasets for: sex, socio-economic disadvantage, size of support network, marital status, functional limitations and diabetes. Median survival from first dementia diagnosis ranged from 1.80 years in the care needs dataset to 3.74 years in the pharmaceutical claims dataset. Characteristics of people identified with dementia in different administrative datasets reflect the factors that drive interaction with specific services; this may introduce bias in observational studies using a single data-source to identify dementia.

Prognostic factors for mortality among patients with visceral leishmaniasis in East Africa: Systematic review and meta-analysis.

BackgroundVisceral leishmaniasis (VL) is a vector-borne disease that is deadly if left untreated. Understanding which factors have prognostic value may help to focus clinical management and reduce case fatality. However, information about prognostic factors is scattered and conflicting. We conducted a systematic review and meta-analysis to identify prognostic factors for mortality among VL patients in East Africa.Methodology/Principal findingsThe review protocol was registered in PROSPERO (CRD42016043112). We included studies published in English after 1970 describing VL patients treated in East African health facilities. To be included, studies had to report on associations between clinical or laboratory factors and mortality during admission or during VL treatment, with a minimal study size of ten patients. Conference abstracts and evaluations of genetic or immunological prognostic factors were excluded. We searched for studies in MEDLINE and four other databases in December 2018. To assess the risk of bias in observational studies and clinical trials, we used the Quality in Prognostic Studies (QUIPS) tool. We included 48 studies in the systematic review, describing 150,072 VL patients of whom 7,847 (5.2%) died. Twelve prognostic factors were evaluated in five or more studies and these results were submitted to meta-analysis producing one pooled crude odds ratio (OR) per prognostic factor. The following factors were strongly (OR>3) and significantly (P-value<0.05) associated with mortality: jaundice (OR = 8.27), HIV (OR = 4.60), tuberculosis (OR = 4.06), age >45 years (OR = 3.69), oedema (OR = 3.52), bleeding (OR = 3.37), and haemoglobin ≤6.5 g/dl (OR = 3.26). Factors significantly and moderately (OR between one and three) associated with death were severe malnutrition, long duration of illness, young age (<5 years), and large spleen size.Conclusions/SignificanceThese prognostic factors can be identified by health professionals in resource-constrained settings. They should be considered as “core” prognostic factors in future studies that aim at improving the prognosis of VL patients.

多変量解析で何を調整するべきか―観察研究におけるバイアスの整理―

多変量解析で何を調整するべきか―観察研究におけるバイアスの整理―

A step-by-step guide to Propensity Score Matching method using R program in dental research

The propensity score matching method is a statistical method used to reduce selection bias in observational studies and to show effects similar to random allocation. There are many observational studies in dentistry research, and differences in baseline covariates between the control and case groups affect the outcome. In order to reduce the bias due to confounding variables, the propensity scores are used by equating groups based on the baseline covariates. This method is effective, especially when there are many covariates or the sample size is small. In this paper, the propensity score matching method was explained in a simple way with a dental example by using R software. This simulated data were obtained from one of retrospective study. The control group and the case group were matched according to the propensity score and compared before and after treatment. The propensity score matching method could be an alternative to compensate for the disadvantage of the observation study by reducing the bias based on the covariates with the propensity score.

Potential Selection Bias in Observational Studies Comparing Cervical Disc Arthroplasty to Anterior Cervical Discectomy and Fusion.

Retrospective study using a national administrative database. To define the cohort differences in patient characteristics between patients undergoing cervical disc arthroplasty (CDA) and anterior cervical discectomy and fusion (ACDF) in a large national sample, and to describe the impact of those baseline patient characteristics on analyses of costs and complications. CDA was initially studied in high quality, randomized trials with strict inclusion criteria. Recently a number of non-randomized, observational studies have been published an attempt to expand CDA indications. These trials are predisposed to falsely attributing differences in outcomes to an intervention due to selection bias. Adults undergoing ACDF or CDA between 2004 and 2014 were identified using International Classification of Diseases, 9, Clinical Modification (ICD-9-CM) diagnosis and procedure codes. Perioperative demographics, comorbidities, complications, and costs were queried. Patient characteristics were compared via chi-square and t tests. Cost, mortality, and complications were compared between ACDF and CDA cohorts using models that adjusted for demographics and comorbidities, as well as "naïve" models that did not. A total of 290,419 procedures, 98.2% ACDF and 1.8% CDA, were included in the sample. Compared with ACDF patients, CDA patients were younger, healthier as evidenced by number of comorbidities, and had an improved socioeconomic status as measured by income and insurance. The naïve logistic regression model showed that hospital costs for CDA were $549 lower than ACDF. In the fully specified model, CDA was $574 more expensive. The naïve model for medical complications suggests a protective advantage for CDA over ACDF, odds ratio of 0.627, P < 0.01. No statistically significant difference was found in the fully specified model in terms of complications. Patients undergoing CDA were younger and healthier with higher socioeconomic statuses compared with ACDF patients. Accounting for these baseline differences significantly attenuated the apparent benefit for CDR on costs and medical complications. 3.

Point-of-Care Clinical Trials in Sports Medicine Research: Identifying Effective Treatment Interventions Through Comparative Effectiveness Research.

Recently, calls to conduct comparative effectiveness research (CER) in athletic training to better support patient care decisions have been circulated. Traditional research methods (eg, randomized controlled trials [RCTs], observational studies) may be ill suited for CER. Thus, innovative research methods are needed to support CER efforts. To discuss the limitations of traditional research designs in CER studies, describe a novel methodologic approach called the point-of-care clinical trial (POC-CT), and highlight components of the POC-CT (eg, incorporation of an electronic medical record [EMR], Bayesian adaptive feature) that allow investigators to conduct scientifically rigorous studies at the point of care. Practical concerns (eg, high costs and limited generalizability of RCTs, the inability to control for bias in observational studies) may stall CER efforts in athletic training. In short, the aim of the POC-CT is to embed a randomized pragmatic trial into routine care; thus, patients are randomized to minimize potential bias, but the study is conducted at the point of care to limit cost and improve the generalizability of the findings. Furthermore, the POC-CT uses an EMR to replace much of the infrastructure associated with a traditional RCT (eg, research team, patient and clinician reminders) and a Bayesian adaptive feature to help limit the number of patients needed for the study. Together, the EMR and Bayesian adaptive feature can improve the overall feasibility of the study and preserve the typical clinical experiences of the patient and clinician. The POC-CT includes the basic tenets of practice-based research because studies are conducted at the point of care, in real-life settings, and during routine clinical practice. If implemented effectively, the POC-CT can be seamlessly integrated into daily clinical practice, allowing investigators to establish patient-reported evidence that may be quickly applied to patient care decisions. This design appears to be a promising approach for CER investigations and may help establish a "learning health care system" in the sports medicine community.

Time-dependent biases in observational studies of comparative effectiveness research in rheumatology. A methodological review

ObjectiveTo assess to what extent time-dependent biases (ie, immortal time bias (ITB) and time-lag bias (TLB)) occur in the latest rheumatology observational studies, describe their main mechanisms and increase the...

Use of Negative Control Exposure Analysis to Evaluate Confounding: An Example of Acetaminophen Exposure and Attention-Deficit/Hyperactivity Disorder in Nurses' Health Study II.

Frequent maternal use of acetaminophen in pregnancy has been linked to attention-deficit/hyperactivity disorder (ADHD) in children, but concerns regarding uncontrolled confounding remain. In this article, we illustrate use of the negative control exposure (NCE) approach to evaluate uncontrolled confounding bias in observational studies on pregnancy drug safety and explain the causal assumptions behind the method. We conducted an NCE analysis and evaluated the associations between maternal acetaminophen use during different exposure periods and ADHD among 8,856 children born in 1993-2005 to women enrolled in the Nurses' Health Study II cohort. Information on regular maternal acetaminophen use was collected prospectively in biennial questionnaires. A total of 721 children (8.1%) in the cohort had been diagnosed with ADHD as reported by the mothers. Our NCE analysis suggested that only acetaminophen use at the time of pregnancy was associated with childhood ADHD (odds ratio = 1.34, 95% confidence interval: 1.05, 1.72), and the effect estimates for the 2 NCE periods (about 4 years before and 4 years after the pregnancy) were null. Our findings corroborate those of prior reports suggesting that prenatal acetaminophen exposure may influence neurodevelopment. The lack of an association between acetaminophen use in the pre- and postpregnancy exposure periods and ADHD provides assurance that uncontrolled time-invariant factors do not explain this association.

Effective Promotional Strategies Selection in Social Media: A Data-Driven Approach

Nowdays, many companies, organizations and individuals are using the function of sharing or retweeting information to promote their products, policies, and ideas on social media. While a growing body of research has focused on identifying the promoters from millions of users, the promoters themselves are seeking to know which strategy can improve promotional effectiveness, which is rarely studied in the literature. In this work, we investigate an open problem of effective promotional strategy selection via causal analysis which is challenging in identifying and quantifying promotional strategies as well as the selection bias when estimating the causal effect of promotional strategies from observational data. We study the promotional strategies not only on the content level (what to promote) but also on the context level (when and how to promote). To alleviate the issue of selection bias in observational studies, we propose a data-driven approach that is a Propensity Score Matching (PSM) based method, which helps to evaluate the causal effect of each promotional strategy and discover the set of effective strategies to predict the promotional effectiveness (i.e., the number of users infected by the promotion). We evaluate our proposed method on a real social dataset including 194 million users and 5 million promoted messages. Experimental results show that (1) the top-ranked strategies by our PSM based method significantly and consistently outperform the correlation based feature selection methods in predicting promotional effectiveness; (2) we conclude our observations from the real data with three interpretable and practical ideas for steering social media promotion.

The risk of bias in observational studies of exposures (ROBINS-E) tool: concerns arising from application to observational studies of exposures

BackgroundSystematic reviews, which assess the risk of bias in included studies, are increasingly used to develop environmental hazard assessments and public health guidelines. These research areas typically rely on evidence from human observational studies of exposures, yet there are currently no universally accepted standards for assessing risk of bias in such studies. The risk of bias in non-randomised studies of exposures (ROBINS-E) tool has been developed by building upon tools for risk of bias assessment of randomised trials, diagnostic test accuracy studies and observational studies of interventions. This paper reports our experience with the application of the ROBINS-E tool.MethodsWe applied ROBINS-E to 74 exposure studies (60 cohort studies, 14 case-control studies) in 3 areas: environmental risk, dietary exposure and drug harm. All investigators provided written feedback, and we documented verbal discussion of the tool. We inductively and iteratively classified the feedback into 7 themes based on commonalities and differences until all the feedback was accounted for in the themes. We present a description of each theme.ResultsWe identified practical concerns with the premise that ROBINS-E is a structured comparison of the observational study being rated to the ‘ideal’ randomised controlled trial. ROBINS-E assesses 7 domains of bias, but relevant questions related to some critical sources of bias, such as exposure and funding source, are not assessed. ROBINS-E fails to discriminate between studies with a single risk of bias or multiple risks of bias. ROBINS-E is severely limited at determining whether confounders will bias study outcomes. The construct of co-exposures was difficult to distinguish from confounders. Applying ROBINS-E was time-consuming and confusing.ConclusionsOur experience suggests that the ROBINS-E tool does not meet the need for an international standard for evaluating human observational studies for questions of harm relevant to public and environmental health. We propose that a simpler tool, based on empirical evidence of bias, would provide accurate measures of risk of bias and is more likely to meet the needs of the environmental and public health community.