Instruments for Causal Inference: An Epidemiologist's Dream?

Department of Epidemiology, Harvard School of Public Health, Boston, Massachusetts 02115, USA.
Epidemiology (Impact Factor: 6.2). 08/2006; 17(4):360-72. DOI: 10.1097/01.ede.0000222409.00878.37
Source: PubMed


The use of instrumental variable (IV) methods is attractive because, even in the presence of unmeasured confounding, such methods may consistently estimate the average causal effect of an exposure on an outcome. However, for this consistent estimation to be achieved, several strong conditions must hold. We review the definition of an instrumental variable, describe the conditions required to obtain consistent estimates of causal effects, and explore their implications in the context of a recent application of the instrumental variables approach. We also present (1) a description of the connection between 4 causal models-counterfactuals, causal directed acyclic graphs, nonparametric structural equation models, and linear structural equation models-that have been used to describe instrumental variables methods; (2) a unified presentation of IV methods for the average causal effect in the study population through structural mean models; and (3) a discussion and new extensions of instrumental variables methods based on assumptions of monotonicity.

1 Follower
14 Reads
  • Source
    • "The use of genetic variants as instrumental variables in observational data has been termed 'Mendelian randomization' [6] [7]. Many reviews exist on the use of instrumental variables, in particular on the assumptions necessary to be an instrumental variable [8] [9], assessing the validity of instrumental variables [10], assumptions necessary for estimation of a causal effect parameter using instrumental variables [11] [12] [13], methods for effect estimation with multiple instrumental variables [14], with binary outcomes [15] [16], methods for the estimation of odds ratios [17], and guidelines for the reporting of instrumental variable analysis [18]. We seek to complement this literature by contributing a review to compare methods for instrumental variable analysis, with accompanying practical guidelines on their use. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Instrumental variable analysis is an approach for obtaining causal inferences on the effect of an exposure (risk factor) on an outcome from observational data. It has gained in popularity over the past decade with the use of genetic variants as instrumental variables, known as Mendelian randomization. An instrumental variable is associated with the exposure, but not associated with any confounder of the exposure-outcome association, nor is there any causal pathway from the instrumental variable to the outcome other than via the exposure. Under the assumption that a single instrumental variable or a set of instrumental variables for the exposure is available, the causal effect of the exposure on the outcome can be estimated. There are several methods available for instrumental variable estimation; we consider the ratio method, two-stage methods, likelihood-based methods, and semi-parametric methods. Techniques for obtaining statistical inferences and confidence intervals are presented. The statistical properties of estimates from these methods are compared, and practical advice is given about choosing a suitable analysis method. In particular, bias and coverage properties of estimators are considered, especially with weak instruments. Settings particularly relevant to Mendelian randomization are prioritized in the paper, notably the scenario of a continuous exposure and a continuous or binary outcome. © The Author(s) 2015.
    Statistical Methods in Medical Research 07/2015; DOI:10.1177/0962280215597579 · 4.47 Impact Factor
  • Source
    • "Intuitively, the IV method seeks to extract variation in treatment that is free of unmeasured confounders and uses this variation to estimate the treatment effect. For more information on IV methods, see Angrist et al. (1996), Newhouse and McClellan (1998), Greenland (2000), Hernán and Robins (2006), Cheng et al. (2009a), Baiocchi et al. (2014) and Imbens (2014). This paper is motivated by provider preference IV (PP IV) which is commonly used as an IV in health studies . "
    [Show abstract] [Hide abstract]
    ABSTRACT: Instrumental variable (IV) methods are widely used to adjust for the bias in estimating treatment effects caused by unmeasured confounders in observational studies. In this manuscript, we provide empirical and theoretical evidence that the IV methods may result in biased treatment effects if applied on a data set in which subjects are preselected based on their received treatments. We frame this as a selection bias problem and propose a procedure that identifies the treatment effect of interest as a function of a vector of sensitivity parameters. We also list assumptions under which analyzing the preselected data does not lead to a biased treatment effect estimate. The performance of the proposed method is examined using simulation studies. We applied our method on The Health Improvement Network (THIN) database to estimate the comparative effect of metformin and sulfonylureas on weight gain among diabetic patients.
  • Source
    • "In recent years, the instrumental variable (IV) design has gained popularity in epidemiology, as a strategy to recover an unbiased estimate of an exposure or treatment causal effect in settings where unobserved confounding is suspected to be present (Greenland, 2000; Davey Smith and Ebrahim, 2003; Hernán and Robins, 2006; Lawlor et al., 2008; Palmer et al., 2011). For continuous outcomes, the most common IV estimator used in practice is two stage least squares which involves fitting a linear regression of the outcome on an estimate of the exposure mean, obtained by regressing the exposure on the IV in a first stage linear model (Wooldridge, 2002). "
    [Show abstract] [Hide abstract]
    ABSTRACT: Unobserved confounding is a well known threat to causal inference in non-experimental studies. The instrumental variable design can under certain conditions be used to recover an unbiased estimator of a treatment effect even if unobserved confounding cannot be ruled out with certainty. For continuous outcomes, two stage least squares is the most common instrumental variable estimator used in epidemiologic applications. For a rare binary outcome, an analogous linear-logistic two-stage procedure can be used. Alternatively, a control function approach is sometimes used which entails entering the residual from the first stage linear model as a covariate in a second stage logistic regression of the outcome on the treatment. Both strategies for binary response have previously formally been justified only for continuous exposure, which has impeded widespread use of the approach outside of this setting. In this note, we consider the important setting of binary exposure in the context of a binary outcome. We provide an alternative motivation for the control function approach which is appropriate for binary exposure, thus establishing simple conditions under which the approach may be used for instrumental variable estimation when the outcome is rare. In the proposed approach, the first stage regression involves a logistic model of the exposure conditional on the instrumental variable, and the second stage regression is a logistic regression of the outcome on the exposure adjusting for the first stage residual. In the event of a non-rare outcome, we recommend replacing the second stage logistic model with a risk ratio regression.
    12/2014; 3(1):107-112. DOI:10.1515/em-2014-0009
Show more

Preview (2 Sources)

14 Reads