Propensity Score Methods for Confounding Control in Nonexperimental Research
01 Sep 2013-Circulation-cardiovascular Quality and Outcomes (Circ Cardiovasc Qual Outcomes)-Vol. 6, Iss: 5, pp 604-611
TL;DR: Nonexperimental studies are increasingly used to investigate the safety and effectiveness of medical products as they are used in routine care, and confounding, systematic differences in prognosis between patients exposed to an intervention of interest and the selected comparator group is a primary challenge.
Abstract: Nonexperimental studies are increasingly used to investigate the safety and effectiveness of medical products as they are used in routine care. One of the primary challenges of such studies is confounding, systematic differences in prognosis between patients exposed to an intervention of interest and the selected comparator group. In the presence of uncontrolled confounding, any observed difference in outcome risk between the groups cannot be attributed solely to a causal effect of the exposure on the outcome.
Confounding in studies of medical products can arise from a variety of different sociomedical processes.1 The most common form of confounding arises from good medical practice, physicians prescribing medications and performing procedures on patients who are most likely to benefit from them. This leads to a bias known as confounding by indication, which can cause medical interventions to appear to cause events that they prevent.2,3 Conversely, patients who are perceived by a physician to be near the end of life may be less likely to receive preventive medications, leading to confounding by frailty or comorbidity.4–6 Additional sources of confounding bias can result from patients’ health-related behaviors. For example, patients who initiate a preventive medication may be more likely than other patients to engage in other healthy, prevention-oriented behaviors leading to bias known as the healthy user/adherer effect.7–9
Many statistical approaches can be used to remove the confounding effects of such factors if they are captured in the data. The most common statistical approaches for confounding control are based on multivariable regression models of the outcome. To yield unbiased estimates of treatment effects, these approaches require that the researcher correctly models the effect of the treatment and covariates on the outcome. However, correct specification of an outcome model can be challenging, particularly in studies …
Citations
More filters
TL;DR: It is believed that NHIRD with multiple data sources could represent a powerful research engine with enriched dimensions and could serve as a guiding light for real-world evidence-based medicine in Taiwan.
Abstract: Taiwan's National Health Insurance Research Database (NHIRD) exemplifies a population-level data source for generating real-world evidence to support clinical decisions and health care policy-making. Like with all claims databases, there have been some validity concerns of studies using the NHIRD, such as the accuracy of diagnosis codes and issues around unmeasured confounders. Endeavors to validate diagnosed codes or to develop methodologic approaches to address unmeasured confounders have largely increased the reliability of NHIRD studies. Recently, Taiwan's Ministry of Health and Welfare (MOHW) established a Health and Welfare Data Center (HWDC), a data repository site that centralizes the NHIRD and about 70 other health-related databases for data management and analyses. To strengthen the protection of data privacy, investigators are required to conduct on-site analysis at an HWDC through remote connection to MOHW servers. Although the tight regulation of this on-site analysis has led to inconvenience for analysts and has increased time and costs required for research, the HWDC has created opportunities for enriched dimensions of study by linking across the NHIRD and other databases. In the near future, researchers will have greater opportunity to distill knowledge from the NHIRD linked to hospital-based electronic medical records databases containing unstructured patient-level information by using artificial intelligence techniques, including machine learning and natural language processes. We believe that NHIRD with multiple data sources could represent a powerful research engine with enriched dimensions and could serve as a guiding light for real-world evidence-based medicine in Taiwan.
611 citations
TL;DR: This report aims to provide methodological guidance to help practitioners select the most appropriate weighting method based on propensity scores for their analysis out of many available options, and outlines recommendations for transparent reporting of studies using weighting based on the propensity scores.
Abstract: This report aims to provide methodological guidance to help practitioners select the most appropriate weighting method based on propensity scores for their analysis out of many available options (eg, inverse probability treatment weights, standardised mortality ratio weights, fine stratification weights, overlap weights, and matching weights), and outlines recommendations for transparent reporting of studies using weighting based on the propensity scores.
275 citations
TL;DR: PPI exposure associates with increased risk of incident CKD, CKD progression, and ESRD, and a graded association between duration of PPI exposure and risk of renal outcomes among those exposed for ≤30 days is detected.
Abstract: The association between proton pump inhibitors (PPI) use and risk of acute interstitial nephritis has been described. However, whether exposure to PPI associates with incident CKD, CKD progression, or ESRD is not known. We used Department of Veterans Affairs national databases to build a primary cohort of new users of PPI ( n =173,321) and new users of histamine H 2 -receptor antagonists (H 2 blockers; n =20,270) and followed these patients over 5 years to ascertain renal outcomes. In adjusted Cox survival models, the PPI group, compared with the H 2 blockers group, had an increased risk of incident eGFR 2 and of incident CKD (hazard ratio [HR], 1.22; 95% confidence interval [95% CI], 1.18 to 1.26; and HR, 1.28; 95% CI, 1.23 to 1.34, respectively). Patients treated with PPI also had a significantly elevated risk of doubling of serum creatinine level (HR, 1.53; 95% CI, 1.42 to 1.65), of eGFR decline >30% (HR, 1.32; 95% CI, 1.28 to 1.37), and of ESRD (HR, 1.96; 95% CI, 1.21 to 3.18). Furthermore, we detected a graded association between duration of PPI exposure and risk of renal outcomes among those exposed to PPI for 31–90, 91–180, 181–360, and 361–720 days compared with those exposed for ≤30 days. Examination of risk of renal outcomes in 1:1 propensity score-matched cohorts of patients taking H 2 blockers versus patients taking PPI and patients taking PPI versus controls yielded consistent results. Our results suggest that PPI exposure associates with increased risk of incident CKD, CKD progression, and ESRD.
250 citations
TL;DR: For exposures with prevalence under 5%, propensity-score stratification with fine strata, based on the exposed group propensity- score distribution, produced the best results.
Abstract: Background:When exposure is infrequent, propensity-score matching results in reduced precision because it discards a large proportion of unexposed patients. To our knowledge, the relative performance of propensity-score stratification in these circumstances has not been examined.Methods:Using an emp
153 citations
TL;DR: Dabigatran could be a safe and potentially more effective alternative to warfarin in patients with atrial fibrillation managed in routine practice settings and estimates of effectiveness and safety appeared to be mostly similar across subgroups.
Abstract: Background The recent availability of dabigatran, a novel oral anticoagulant, provided a new treatment option for stroke prevention in atrial fibrillation beyond warfarin, the main therapy for years. Little is known about their real‐world comparative effectiveness and safety, even less among patient demographic and clinical subgroups. Methods and Results Using a cohort of non‐valvular AF patients initiating anticoagulation from October 2010 to December 2012 drawn from a large US database of commercial and Medicare supplement claims, we applied propensity score weights to Cox proportional hazards regression to assess the comparative effectiveness and safety of dabigatran versus warfarin. Analyses were repeated among clinical and demographic subgroups using stratum‐specific propensity scores as an exploratory analysis. Of the 64 935 patients initiating anticoagulation, 32.5% used dabigatran. Compared with warfarin, dabigatran was associated with a lower risk of ischemic stroke or systemic embolism (composite adjusted Hazard Ratio [aHR], 95% CI: 0.86, 95% CI: 0.79 to 0.93), hemorrhagic stroke (aHR: 0.51, 0.40 to 0.65), and acute myocardial infarction (aHR: 0.88, 95% CI: 0.77 to 0.99), and no relation was seen between dabigatran and the composite harm outcome (aHR: 0.94, 95% CI: 0.87 to 1.01). However, dabigatran was associated with a higher risk of gastrointestinal bleeding (aHR: 1.11, 95% CI: 1.02 to 1.22). Estimates of effectiveness and safety appeared to be mostly similar across subgroups. Conclusions Dabigatran could be a safe and potentially more effective alternative to warfarin in patients with atrial fibrillation managed in routine practice settings.
152 citations
References
More filters
TL;DR: The authors discusses the central role of propensity scores and balancing scores in the analysis of observational studies and shows that adjustment for the scalar propensity score is sufficient to remove bias due to all observed covariates.
Abstract: : The results of observational studies are often disputed because of nonrandom treatment assignment. For example, patients at greater risk may be overrepresented in some treatment group. This paper discusses the central role of propensity scores and balancing scores in the analysis of observational studies. The propensity score is the (estimated) conditional probability of assignment to a particular treatment given a vector of observed covariates. Both large and small sample theory show that adjustment for the scalar propensity score is sufficient to remove bias due to all observed covariates. Applications include: matched sampling on the univariate propensity score which is equal percent bias reducing under more general conditions than required for discriminant matching, multivariate adjustment by subclassification on balancing scores where the same subclasses are used to estimate treatment effects for all outcome variables and in all subpopulations, and visual representation of multivariate adjustment by a two-dimensional plot. (Author)
23,744 citations
TL;DR: In this paper, the authors introduce marginal structural models, a new class of causal models that allow for improved adjustment of confounding in observational studies with exposures or treatments that vary over time, when there exist time-dependent confounders that are also affected by previous treatment.
Abstract: In observational studies with exposures or treatments that vary over time, standard approaches for adjustment of confounding are biased when there exist time-dependent confounders that are also affected by previous treatment. This paper introduces marginal structural models, a new class of causal models that allow for improved adjustment of confounding in those situations. The parameters of a marginal structural model can be consistently estimated using a new class of estimators, the inverse-probability-of-treatment weighted estimators.
4,655 citations
TL;DR: Causal diagrams can provide a starting point for identifying variables that must be measured and controlled to obtain unconfounded effect estimates and provide a method for critical evaluation of traditional epidemiologic criteria for confounding.
Abstract: Causal diagrams have a long history of informal use and, more recently, have undergone formal development for applications in expert systems and robotics. We provide an introduction to these developments and their use in epidemiologic research. Causal diagrams can provide a starting point for identifying variables that must be measured and controlled to obtain unconfounded effect estimates. They also provide a method for critical evaluation of traditional epidemiologic criteria for confounding. In particular, they reveal certain heretofore unnoticed shortcomings of those criteria when used in considering multiple potential confounders. We show how to modify the traditional criteria to correct those shortcomings.
2,983 citations
TL;DR: Propensity score methods generalize subclassification in the presence of many confounding covariates, such as age, region of the country, and sex, in a study of smoking and mortality.
Abstract: The aim of many analyses of large databases is to draw causal inferences about the effects of actions, treatments, or interventions. Examples include the effects of various options available to a p...
2,902 citations
TL;DR: This work argues that the causal structure underlying the bias in each example is essentially the same: conditioning on a common effect of 2 variables, one of which is either exposure or a cause of exposure and the other is either the outcome or acause of the outcome.
Abstract: The term "selection bias" encompasses various biases in epidemiology. We describe examples of selection bias in case-control studies (eg, inappropriate selection of controls) and cohort studies (eg, informative censoring). We argue that the causal structure underlying the bias in each example is essentially the same: conditioning on a common effect of 2 variables, one of which is either exposure or a cause of exposure and the other is either the outcome or a cause of the outcome. This structure is shared by other biases (eg, adjustment for variables affected by prior exposure). A structural classification of bias distinguishes between biases resulting from conditioning on common effects ("selection bias") and those resulting from the existence of common causes of exposure and outcome ("confounding"). This classification also leads to a unified approach to adjust for selection bias.
2,195 citations