Instrumental Variables II: Instrumental Variable Application—In 25 Variations, the Physician Prescribing Preference Generally was Strong and Reduced Covariate Imbalance

ArticleinJournal of clinical epidemiology 62(12):1233-41 · May 2009with23 Reads
DOI: 10.1016/j.jclinepi.2008.12.006 · Source: PubMed
An instrumental variable (IV) is an unconfounded proxy for a study exposure that can be used to estimate a causal effect in the presence of unmeasured confounding. To provide reliably consistent estimates of effect, IVs should be both valid and reasonably strong. Physician prescribing preference (PPP) is an IV that uses variation in doctors' prescribing to predict drug treatment. As reduction in covariate imbalance may suggest increased IV validity, we sought to examine the covariate balance and instrument strength in 25 formulations of the PPP IV in two cohort studies. We applied the PPP IV to assess antipsychotic medication (APM) use and subsequent death among two cohorts of elderly patients. We varied the measurement of PPP, plus performed cohort restriction and stratification. We modeled risk differences with two-stage least square regression. First-stage partial r(2) values characterized the strength of the instrument. The Mahalanobis distance summarized balance across multiple covariates. Partial r(2) ranged from 0.028 to 0.099. PPP generally alleviated imbalances in nonpsychiatry-related patient characteristics, and the overall imbalance was reduced by an average of 36% (+/-40%) over the two cohorts. In our study setting, most of the 25 formulations of the PPP IV were strong IVs and resulted in a strong reduction of imbalance in many variations. The association between strength and imbalance was mixed.
    • "Indeed, in the example earlier, one could use the CUE to estimate the causal effects of lean and fat mass jointly. Outside the world of Mendelian randomization, recent pharmaco-epidemiological studies have found that including multiple physicians' previous prescriptions as instruments for their prescribing behavior increases the precision of estimates [57][58][59]. These results may be improved upon by using the CUE. "
    [Show abstract] [Hide abstract] ABSTRACT: Instrumental variable estimates of causal effects can be biased when using many instruments that are only weakly associated with the exposure. We describe several techniques to reduce this bias and estimate corrected standard errors. We present our findings using a simulation study and an empirical application. For the latter, we estimate the effect of height on lung function, using genetic variants as instruments for height. Our simulation study demonstrates that, using many weak individual variants, two-stage least squares (2SLS) is biased, whereas the limited information maximum likelihood (LIML) and the continuously updating estimator (CUE) are unbiased and have accurate rejection frequencies when standard errors are corrected for the presence of many weak instruments. Our illustrative empirical example uses data on 3631 children from England. We used 180 genetic variants as instruments and compared conventional ordinary least squares estimates with results for the 2SLS, LIML, and CUE instrumental variable estimators using the individual height variants. We further compare these with instrumental variable estimates using an unweighted or weighted allele score as single instruments. In conclusion, the allele scores and CUE gave consistent estimates of the causal effect. In our empirical example, estimates using the allele score were more efficient. CUE with corrected standard errors, however, provides a useful additional statistical tool in applications with many weak instruments. The CUE may be preferred over an allele score if the population weights for the allele score are unknown or when the causal effects of multiple risk factors are estimated jointly. © 2014 The Authors. Statistics in Medicine published by John Wiley & Sons Ltd.
    Full-text · Article · Nov 2014
    • "rate control therapy for atrial fibrillation. Finally, Rassen et al. [23] found that restricting their analysis to high-volume practices maximized the strength of their instrument for prescribing antipsychotic medications, and consequently the precision of their results. We add to these findings in three ways. "
    [Show abstract] [Hide abstract] ABSTRACT: To investigate whether physicians' prescribing preferences were valid instrumental variables for the antidepressant prescriptions they issued to their patients. We investigated whether physicians' previous prescriptions of (1) tricyclic antidepressants (TCAs) vs. selective serotonin reuptake inhibitors (SSRIs) and (2) paroxetine vs. other SSRIs were valid instruments. We investigated whether the instrumental variable assumptions are likely to hold and whether TCAs (vs. SSRIs) were associated with hospital admission for self-harm or death by suicide using both conventional and instrumental variable regressions. The setting for the study was general practices in the United Kingdom. Prior prescriptions were strongly associated with actual prescriptions: physicians who previously prescribed TCAs were 14.9 percentage points (95% confidence interval [CI], 14.4, 15.4) more likely to prescribe TCAs, and those who previously prescribed paroxetine were 27.7 percentage points (95% CI, 26.7, 28.8) more likely to prescribe paroxetine, to their next patient. Physicians' previous prescriptions were less strongly associated with patients' baseline characteristics than actual prescriptions. We found no evidence that the estimated association of TCAs with self-harm/suicide using instrumental variable regression differed from conventional regression estimates (P-value = 0.45). The main instrumental variable assumptions held, suggesting that physicians' prescribing preferences are valid instruments for evaluating the short-term effects of antidepressants.
    Full-text · Article · Sep 2013
    • "As illustrated inFigure 1 reverse causality is actually a highly plausible bias. In order to solve this problem, instrumental variables techniques have been used to deal with these potential biases35363738. Consequently, given the strong possibility of endogeneity of poverty incidence, malaria prevalence and protection, and given that the error terms are not necessarily independent of each other, the 3SLS estimates are useful, providing estimates that are free of endogeneity bias. The complete system of structural equations was estimated with a heteroskedastic-efficient 3SLS two step generalized method of moments (3SLS GMM) described by Wooldridge [35]. "
    [Show abstract] [Hide abstract] ABSTRACT: Background In spite of massive efforts to generalize efficient prevention, such as insecticide-treated mosquito nets (ITN) or long-lasting insecticidal nets (LLINs), malaria remains prevalent in many countries and ITN/LLINs are still only used to a limited extent. Methods This study proposes a new model for malaria economic analysis by combining economic epidemiology tools with the literature on poverty traps. A theoretical model of rational protective behaviour in response to malaria is designed, which includes endogenous externalities and disease characteristics. Survey data available for Uganda provide empirical support to the theory of prevalence-elastic protection behaviours, once endogeneity issues related to epidemiology and poverty are solved. Results Two important conclusions emerge from the model. First, agents increase their protective behaviour when malaria is more prevalent in a society. This is consistent with the literature on "prevalence-elastic behaviour". Second, a ‘malaria trap’ defined as the result of malaria reinforcing poverty while poverty reduces the ability to deal with malaria can theoretically exist and the conditions of existence of the malaria trap are identified. Conclusions These results suggest the possible existence of malaria traps, which provides policy implications. Notably, providing ITN/LLINs at subsidized prices is not sufficient. To be efficient an ITN/LLINs dissemination campaigns should include incentive of the very poor for using ITN/LLINs.
    Full-text · Article · Jun 2013
Show more