Estimation of causal effects of binary treatments in unconfounded studies

Estimation of causal effects in non‐randomized studies comprises two distinct phases: design, without outcome data, and analysis of the outcome data according to a specified protocol. Recently, Gutman and Rubin (2013) proposed a new analysis‐phase method for estimating treatment effects when the outcome is binary and there is only one covariate, which viewed causal effect estimation explicitly as a missing data problem. Here, we extend this method to situations with continuous outcomes and multiple covariates and compare it with other commonly used methods (such as matching, subclassification, weighting, and covariance adjustment). We show, using an extensive simulation, that of all methods considered, and in many of the experimental conditions examined, our new ‘multiple‐imputation using two subclassification splines’ method appears to be the most efficient and has coverage levels that are closest to nominal. In addition, it can estimate finite population average causal effects as well as non‐linear causal estimands. This type of analysis also allows the identification of subgroups of units for which the effect appears to be especially beneficial or harmful. Copyright © 2015 John Wiley & Sons, Ltd.     

Multiple imputation procedures for estimating causal effects with multiple treatments with application to the comparison of healthcare providers

Choosing between multiple healthcare providers requires us to simultaneously compare the expected outcomes under each provider. This comparison is complex because the composition of patients treated by each provider may differ. Similar issues arise when simultaneously comparing the adverse effects of interventions using non-randomized data. To simultaneously estimate the effects of multiple providers/interventions we propose procedures that explicitly impute the set of potential outcomes for each subject. The procedures are based on different specifications of the generalized additive models (GAM) and the Bayesian additive regression trees (BART). We compare the performance of the proposed procedures to previously proposed matching and weighting procedures using an extensive simulation study for continuous outcomes. Our simulations show that when the distributions of the covariates across treatment groups have adequate overlap, the multiple imputation procedures based on separate BART or GAM models in each treatment group are generally superior to weighting based methods and have similar and sometimes better performance than matching on the logit of the generalized propensity score. Another advantage of these multiple imputation procedures is the ability to provide point and interval estimates to a wide range of causal effect estimands. We apply the proposed procedures to comparing multiple nursing homes in Massachusetts for readmission outcomes. The proposed approach can be applied to other causal effects applications with multiple treatments.     

Weighted causal inference methods with mismeasured covariates and misclassified outcomes

Inverse probability weighting (IPW) estimation has been widely used in causal inference. Its validity relies on the important condition that the variables are precisely measured. This condition, however, is often violated, which distorts the IPW method and thus yields biased results. In this paper, we study the IPW estimation of average treatment effects for settings with mismeasured covariates and misclassified outcomes. We develop estimation methods to correct for measurement error and misclassification effects simultaneously. Our discussion covers a broad scope of treatment models, including typically assumed logistic regression models and general treatment assignment mechanisms. Satisfactory performance of the proposed methods is demonstrated by extensive numerical studies.     

Multiply robust estimation of causal quantile treatment effects

In causal inference, often the interest lies in the estimation of the average causal effect. Other quantities such as the quantile treatment effect may be of interest as well. In this article, we propose a multiply robust method for estimating the marginal quantiles of potential outcomes by achieving mean balance in (a) the propensity score, and (b) the conditional distributions of potential outcomes. An empirical likelihood or entropy measure approach can be utilized for estimation instead of inverse probability weighting, which is known to be sensitive to the misspecification of the propensity score model. Simulation studies are conducted across different scenarios of correctness in both the propensity score models and the outcome models. Both simulation results and theoretical development indicate that our proposed estimator is consistent if any of the models are correctly specified. In the data analysis, we investigate the quantile treatment effect of mothers' smoking status on infants' birthweight.     

Matching algorithms for causal inference with multiple treatments

Randomized clinical trials are ideal for estimating causal effects, because the distributions of background covariates are similar in expectation across treatment groups. When estimating causal effects using observational data, matching is a commonly used method to replicate the covariate balance achieved in a randomized clinical trial. Matching algorithms have a rich history dating back to the mid-1900s but have been used mostly to estimate causal effects between two treatment groups. When there are more than two treatments, estimating causal effects requires additional assumptions and techniques. We propose several novel matching algorithms that address the drawbacks of the current methods, and we use simulations to compare current and new methods. All of the methods display improved covariate balance in the matched sets relative to the prematched cohorts. In addition, we provide advice to investigators on which matching algorithms are preferred for different covariate distributions.     

Principal stratification for quantile causal effects under partial compliance

Within the principal stratification framework in causal inference, the majority of the literature has focused on binary compliance with an intervention and modelling means. Yet in some research areas, compliance is partial, and research questions—and hence analyses—are concerned with causal effects on (possibly high) quantiles rather than on shifts in average outcomes. Modelling partial compliance is challenging because it can suffer from lack of identifiability. We develop an approach to estimate quantile causal effects within a principal stratification framework, where principal strata are defined by the bivariate vector of (partial) compliance to the two levels of a binary intervention. We propose a conditional copula approach to impute the missing potential compliance and estimate the principal quantile treatment effect surface at high quantiles, allowing the copula association parameter to vary with the covariates. A bootstrap procedure is used to estimate the parameter to account for inflation due to imputation of missing compliance. Moreover, we describe precise assumptions on which the proposed approach is based, and investigate the finite sample behavior of our method by a simulation study. The proposed approach is used to study the 90th principal quantile treatment effect of executive stay-at-home orders on mitigating the risk of COVID-19 transmission in the United States.     

A mediation analysis for a nonrare dichotomous outcome with sequentially ordered multiple mediators

Mediation analyses can help us to understand the biological mechanism in which an exposure or treatment affects an outcome. Single mediator analyses have been used in various applications, but may not be appropriate for analyzing intricate mechanisms involving multiple mediators that affect each other. Thus, in this article, we studied multiple sequentially ordered mediators for a dichotomous outcome and presented the identifiability assumptions for the path-specific effects on the outcome, that is, the effect of an exposure on the outcome mediated by a specific set of mediators. We proposed a closed-form estimator for the path-specific effects by modeling the dichotomous outcome using a probit model. Asymptotic variance of the proposed estimator is derived and can be approximated via delta method or bootstrapping. Simulations under a finite sample showed the validity of our method in capturing the path-specific effects when the probability of each potential counterfactual outcome is not small and demonstrated the utility of a computationally efficient alternative to bootstrapping for calculating variance. The method is applied to investigate the effects of polycystic ovarian syndrome on live birth rates mediated by estradiol levels and the number of oocytes retrieved in a large electronic in vitro fertilization database. We implemented the method into an R package SOMM , which is available at https://github.com/roqe/SOMM .     

Variable selection for multiply‐imputed data with application to dioxin exposure study

Multiple imputation (MI) is a commonly used technique for handling missing data in large‐scale medical and public health studies. However, variable selection on multiply‐imputed data remains an important and longstanding statistical problem. If a variable selection method is applied to each imputed dataset separately, it may select different variables for different imputed datasets, which makes it difficult to interpret the final model or draw scientific conclusions. In this paper, we propose a novel multiple imputation‐least absolute shrinkage and selection operator (MI‐LASSO) variable selection method as an extension of the least absolute shrinkage and selection operator (LASSO) method to multiply‐imputed data. The MI‐LASSO method treats the estimated regression coefficients of the same variable across all imputed datasets as a group and applies the group LASSO penalty to yield a consistent variable selection across multiple‐imputed datasets. We use a simulation study to demonstrate the advantage of the MI‐LASSO method compared with the alternatives. We also apply the MI‐LASSO method to the University of Michigan Dioxin Exposure Study to identify important circumstances and exposure factors that are associated with human serum dioxin concentration in Midland, Michigan. Copyright © 2013 John Wiley & Sons, Ltd.     

The design versus the analysis of observational studies for causal effects: parallels with the design of randomized trials

For estimating causal effects of treatments, randomized experiments are generally considered the gold standard. Nevertheless, they are often infeasible to conduct for a variety of reasons, such as ethical concerns, excessive expense, or timeliness. Consequently, much of our knowledge of causal effects must come from non-randomized observational studies. This article will advocate the position that observational studies can and should be designed to approximate randomized experiments as closely as possible. In particular, observational studies should be designed using only background information to create subgroups of similar treated and control units, where 'similar' here refers to their distributions of background variables. Of great importance, this activity should be conducted without any access to any outcome data, thereby assuring the objectivity of the design. In many situations, this objective creation of subgroups of similar treated and control units, which are balanced with respect to covariates, can be accomplished using propensity score methods. The theoretical perspective underlying this position will be presented followed by a particular application in the context of the US tobacco litigation. This application uses propensity score methods to create subgroups of treated units (male current smokers) and control units (male never smokers) who are at least as similar with respect to their distributions of observed background characteristics as if they had been randomized. The collection of these subgroups then 'approximate' a randomized block experiment with respect to the observed covariates.     

Optimal probability weights for estimating causal effects of time-varying treatments with marginal structural Cox models

Marginal structural Cox models have been used to estimate the causal effect of a time-varying treatment on a survival outcome in the presence of time-dependent confounders. These methods rely on the positivity assumption, which states that the propensity scores are bounded away from zero and one. Practical violations of this assumption are common in longitudinal studies, resulting in extreme weights that may yield erroneous inferences. Truncation, which consists of replacing outlying weights with less extreme ones, is the most common approach to control for extreme weights to date. While truncation reduces the variability in the weights and the consequent sampling variability of the estimator, it can also introduce bias. Instead of truncated weights, we propose using optimal probability weights, defined as those that have a specified variance and the smallest Euclidean distance from the original, untruncated weights. The set of optimal weights is obtained by solving a constrained quadratic optimization problem. The proposed weights are evaluated in a simulation study and applied to the assessment of the effect of treatment on time to death among people in Sweden who live with human immunodeficiency virus and inject drugs.     

On the aggregation of published prognostic scores for causal inference in observational studies

As real world evidence on drug efficacy involves nonrandomized studies, statistical methods adjusting for confounding are needed. In this context, prognostic score (PGS) analysis has recently been proposed as a method for causal inference. It aims to restore balance across the different treatment groups by identifying subjects with a similar prognosis for a given reference exposure (“control”). This requires the development of a multivariable prognostic model in the control arm of the study sample, which is then extrapolated to the different treatment arms. Unfortunately, large cohorts for developing prognostic models are not always available. Prognostic models are therefore subject to a dilemma between overfitting and parsimony; the latter being prone to a violation of the assumption of no unmeasured confounders when important covariates are ignored. Although it is possible to limit overfitting by using penalization strategies, an alternative approach is to adopt evidence synthesis. Aggregating previously published prognostic models may improve the generalizability of PGS, while taking account of a large set of covariates—even when limited individual participant data are available. In this article, we extend a method for prediction model aggregation to PGS analysis in nonrandomized studies. We conduct extensive simulations to assess the validity of model aggregation, compared with other methods of PGS analysis for estimating marginal treatment effects. We show that aggregating existing PGS into a “meta-score” is robust to misspecification, even when elementary scores wrongfully omit confounders or focus on different outcomes. We illustrate our methods in a setting of treatments for asthma.     

Combining multiple imputation and meta‐analysis with individual participant data

Multiple imputation is a strategy for the analysis of incomplete data such that the impact of the missingness on the power and bias of estimates is mitigated. When data from multiple studies are collated, we can propose both within‐study and multilevel imputation models to impute missing data on covariates. It is not clear how to choose between imputation models or how to combine imputation and inverse‐variance weighted meta‐analysis methods. This is especially important as often different studies measure data on different variables, meaning that we may need to impute data on a variable which is systematically missing in a particular study. In this paper, we consider a simulation analysis of sporadically missing data in a single covariate with a linear analysis model and discuss how the results would be applicable to the case of systematically missing data. We find in this context that ensuring the congeniality of the imputation and analysis models is important to give correct standard errors and confidence intervals. For example, if the analysis model allows between‐study heterogeneity of a parameter, then we should incorporate this heterogeneity into the imputation model to maintain the congeniality of the two models. In an inverse‐variance weighted meta‐analysis, we should impute missing data and apply Rubin's rules at the study level prior to meta‐analysis, rather than meta‐analyzing each of the multiple imputations and then combining the meta‐analysis estimates using Rubin's rules. We illustrate the results using data from the Emerging Risk Factors Collaboration. © 2013 The Authors. Statistics in Medicine published by John Wiley & Sons Ltd.     

On the conditional logistic estimator in two‐arm experimental studies with non‐compliance and before–after binary outcomes

The behavior of the conditional logistic estimator is analyzed under a causal model for two‐arm experimental studies with possible non‐compliance in which the effect of the treatment is measured by a binary response variable. We show that, when non‐compliance may only be observed in the treatment arm, the effect (measured on the logit scale) of the treatment on compliers and that of the control on non‐compliers can be identified and consistently estimated under mild conditions. The same does not happen for the effect of the control on compliers. A simple correction of the conditional logistic estimator is then proposed, which allows us to considerably reduce the bias in estimating this quantity and the causal effect of the treatment over control on compliers. A two‐step estimator results on the basis of which we can also set up a Wald test for the hypothesis of absence of a causal effect of the treatment. The asymptotic properties of the estimator are studied by exploiting the general theory on maximum likelihood estimation of misspecified models. Finite‐sample properties of the estimator and of the related Wald test are studied by simulation. The extension of the approach to the case of missing responses is also outlined. The approach is illustrated by an application to a dataset deriving from a study on the efficacy of a training course on the breast self examination practice. Copyright © 2010 John Wiley & Sons, Ltd.     

Bias in causal estimates from Mendelian randomization studies with weak instruments

Mendelian randomization studies using genetic instrumental variables (IVs) are now being commonly used to estimate the causal association of a phenotype on an outcome. Even when the necessary underlying assumptions are valid, estimates from analyses using IVs are biased in finite samples. The source and nature of this bias appear poorly understood in the epidemiological field. We explain why the bias is in the direction of the confounded observational association, with magnitude relating to the statistical strength of association between the instrument and phenotype. We comment on the size of the bias, from simulated data, showing that when multiple instruments are used, although the variance of the IV estimator decreases, the bias increases. We discuss ways to analyse Mendelian randomization studies to alleviate the problem of weak instrument bias.     

Testing treatment effects in unconfounded studies under model misspecification: Logistic regression,discretization, and their combination

Logistic regression is commonly used to test for treatment effects in observational studies. If the distribution of a continuous covariate differs between treated and control populations, logistic regression yields an invalid hypothesis test even in an uncounfounded study if the link is not logistic. This flaw is not corrected by the commonly used technique of discretizing the covariate into intervals. A valid test can be obtained by discretization followed by regression adjustment within each interval. Copyright © 2009 John Wiley & Sons, Ltd.     

Estimating causal effects of treatment in RCTs with provider and subject noncompliance

Subject noncompliance is a common problem in the analysis of randomized clinical trials (RCTs). With cognitive behavioral interventions, the addition of provider noncompliance further complicates making causal inference. As a motivating example, we consider an RCT of a motivational interviewing (MI)-based behavioral intervention for treating problem drug use. Treatment receipt depends on compliance of both a therapist (provider) and a patient (subject), where MI is received when the therapist adheres to the MI protocol and the patient actively participates in the intervention. However, therapists cannot be forced to follow protocol and patients cannot be forced to cooperate in an intervention. In this article, we (1) define a causal estimand of interest based on a principal stratification framework, the average causal effect of treatment among provider-subject pairs that comply with assignment or ACE(cc); (2) explore possible assumptions that identify ACE(cc); (3) develop novel estimators of ACE(cc); (4) evaluate estimators' statistical properties via simulation; and (5) apply our proposed methods for estimating ACE(cc) to data from our motivating example.     

Causal inference with noisy data: Bias analysis and estimation approaches to simultaneously addressing missingness and misclassification in binary outcomes

Causal inference has been widely conducted in various fields and many methods have been proposed for different settings. However, for noisy data with both mismeasurements and missing observations, those methods often break down. In this paper, we consider a problem that binary outcomes are subject to both missingness and misclassification, when the interest is in estimation of the average treatment effects (ATE). We examine the asymptotic biases caused by ignoring missingness and/or misclassification and establish the intrinsic connections between missingness effects and misclassification effects on the estimation of ATE. We develop valid weighted estimation methods to simultaneously correct for missingness and misclassification effects. To provide protection against model misspecification, we further propose a doubly robust correction method which yields consistent estimators when either the treatment model or the outcome model is misspecified. Simulation studies are conducted to assess the performance of the proposed methods. An application to smoking cessation data is reported to illustrate the use of the proposed methods.     

The use of prognostic scores for causal inference with general treatment regimes

In nonrandomised studies, inferring causal effects requires appropriate methods for addressing confounding bias. Although it is common to adopt propensity score analysis to this purpose, prognostic score analysis has recently been proposed as an alternative strategy. While both approaches were originally introduced to estimate causal effects for binary interventions, the theory of propensity score has since been extended to the case of general treatment regimes. Indeed, many treatments are not assigned in a binary fashion and require a certain extent of dosing. Hence, researchers may often be interested in estimating treatment effects across multiple exposures. To the best of our knowledge, the prognostic score analysis has not been yet generalised to this case. In this article, we describe the theory of prognostic scores for causal inference with general treatment regimes. Our methods can be applied to compare multiple treatments using nonrandomised data, a topic of great relevance in contemporary evaluations of clinical interventions. We propose estimators for the average treatment effects in different populations of interest, the validity of which is assessed through a series of simulations. Finally, we present an illustrative case in which we estimate the effect of the delay to Aspirin administration on a composite outcome of death or dependence at 6 months in stroke patients.     

Quantifying degrees of necessity and of sufficiency in cause-effect relationships with dichotomous and survival outcomes

We suggest measures to quantify the degrees of necessity and of sufficiency of prognostic factors for dichotomous and for survival outcomes. A cause, represented by certain values of prognostic factors, is considered necessary for an event if, without the cause, the event cannot develop. It is considered sufficient for an event if the event is unavoidable in the presence of the cause. Necessity and sufficiency can be seen as the two faces of causation, and this symmetry and equal relevance are reflected by the suggested measures. The measures provide an approximate, in some cases an exact, multiplicative decomposition of explained variation as defined by Schemper and Henderson for censored survival and for dichotomous outcomes. The measures, ranging from zero to one, are simple, intuitive functions of unconditional and conditional probabilities of an event such as disease or death. These probabilities often will be derived from logistic or Cox regression models; the measures, however, do not require any particular model. The measures of the degree of necessity implicitly generalize the established attributable fraction or risk for dichotomous prognostic factors and dichotomous outcomes to continuous prognostic factors and to survival outcomes. In a setting with multiple prognostic factors, they provide marginal and partial results akin to marginal and partial odds and hazard ratios from multiple logistic and Cox regression. Properties of the measures are explored by an extensive simulation study. Their application is demonstrated by three typical real data examples.     

Sensitivity analysis of treatment effect to unmeasured confounding in observational studies with survival and competing risks outcomes

No unmeasured confounding is often assumed in estimating treatment effects in observational data, whether using classical regression models or approaches such as propensity scores and inverse probability weighting. However, in many such studies collection of confounders cannot possibly be exhaustive in practice, and it is crucial to examine the extent to which the resulting estimate is sensitive to the unmeasured confounders. We consider this problem for survival and competing risks data. Due to the complexity of models for such data, we adapt the simulated potential confounder approach of Carnegie et al (2016), which provides a general tool for sensitivity analysis due to unmeasured confounding. More specifically, we specify one sensitivity parameter to quantify the association between an unmeasured confounder and the exposure or treatment received, and another set of parameters to quantify the association between the confounder and the time-to-event outcomes. By varying the magnitudes of the sensitivity parameters, we estimate the treatment effect of interest using the stochastic expectation-maximization (EM) and the EM algorithms. We demonstrate the performance of our methods on simulated data, and apply them to a comparative effectiveness study in inflammatory bowel disease. An R package “survSens” is available on CRAN that implements the proposed methodology.