Dissonant Conclusions When Testing the Validity of an Instrumental Variable

Valid instrumental variable selection method using negative control outcomes and constructing efficient estimator

In observational studies, instrumental variable (IV) methods are commonly applied when there are unmeasured covariates. In Mendelian randomization, constructing an allele score using many single nucleotide polymorphisms is often implemented; however, estimating biased causal effects by including some invalid IVs poses some risks. Invalid IVs are those IV candidates that are associated with unobserved variables. To solve this problem, we developed a novel strategy using negative control outcomes (NCOs) as auxiliary variables. Using NCOs, we are able to select only valid IVs and exclude invalid IVs without knowing which of the instruments are invalid. We also developed a new two-step estimation procedure and proved the semiparametric efficiency of our estimator. The performance of our proposed method was superior to some previous methods through simulations. Subsequently, we applied the proposed method to the UK Biobank dataset. Our results demonstrate that the use of an auxiliary variable, such as an NCO, enables the selection of valid IVs with assumptions different from those used in previous methods.

Evaluating a Key Instrumental Variable Assumption Using Randomization Tests

Instrumental variable (IV) analyses are becoming common in health services research and epidemiology. Most IV analyses use naturally occurring instruments, such as distance to a hospital. In these analyses, investigators must assume that the instrument is as-if randomly assigned. This assumption cannot be tested directly, but it can be falsified. Most IV falsification tests compare relative prevalence or bias in observed covariates between the instrument and exposure. These tests require investigators to make covariate-by-covariate judgments about the validity of the IV design. Often, only some covariates are well-balanced, making it unclear whether as-if randomization can be assumed for the instrument. We propose an alternative falsification test that compares IV balance or bias with the balance or bias that would have been produced under randomization. A key advantage of our test is that it allows for global balance measures as well as easily interpretable graphical comparisons. Furthermore, our test does not rely on parametric assumptions and can be used to validly assess whether the instrument is significantly closer to being as-if randomized than the exposure. We demonstrate our approach using data from (SPOT)light, a prospective cohort study carried out in 48 National Health Service hospitals in the United Kingdom between November 1, 2010, and December 31, 2011. This study used bed availability in the intensive care unit as an instrument for admission to the intensive care unit.

Does Transfer to Intensive Care Units Reduce Mortality? A Comparison of an Instrumental Variables Design to Risk Adjustment

BACKGROUND

Instrumental variable (IV) analysis can estimate treatment effects in the presence of residual or unmeasured confounding. In settings wherein measures of baseline risk severity are unavailable, IV designs are, therefore, particularly appealing, but, where established measures of risk severity are available, it is unclear whether IV methods are preferable.

OBJECTIVE

We compared regression with an IV design to estimate the effect of intensive care unit (ICU) transfer on mortality in a study with well-established measures of risk severity.

RESEARCH DESIGN

We use ICU bed availability at the time of assessment for ICU transfer as an instrument. Bed availability increases the chance of ICU admission, contains little information about patient characteristics, and it is unlikely that bed availability has any direct effect on in-hospital mortality.

SUBJECTS

We used a cohort study of deteriorating ward patients assessed for critical care unit admission, in 49 UK National Health Service hospitals between November 1, 2010, and December 31, 2011.

MEASURES

Detailed demographic, physiological, and comorbidity data were collected for all patients.

RESULTS

The risk adjustment methods reported that, after controlling for all measured covariates including measures of risk severity, ICU transfer was associated with higher 28-day mortality, with a risk difference of 7.2% (95% confidence interval=5.3%-9.1%). The IV estimate of ICU transfer was -5.4% (95% confidence interval=-47.1% to 36.3%) and applies to the subsample of patients whose transfer was "encouraged" by bed availability.

CONCLUSIONS

IV estimates indicate that ICU care is beneficial but are imprecisely estimated. Risk-adjusted estimates are more precise but, even with a rich set of covariates, report that ICU care is harmful.

Long-term use of hydrocodone vs. oxycodone in primary care

BACKGROUND

Hydrocodone and oxycodone are the Schedule II opioids most often prescribed in primary care. Notwithstanding the dangers of prescription opioid use, the likelihood of long-term use with either drug is presently unknown.

METHODS

Using a retrospective cohort design and data from a commerical healthcare claims repository, we compared the likelihood of long-term use of hydrocodone and oxycodone in primary care patients presenting with acute back pain. Treatment was categorized as long-term if the prescription dates spanned ≥90 days from initial prescription to the run-out date of the last prescription, and included ≥120 days' supply or ≥10 fills. Instrumental variable methods and probit regression were used to model the effect of drug choice on long-term use, estimate the average treatment effect, and correct for confounding by indication.

RESULTS

A total of 3,983 patients who were prescribed only hydrocodone or only oxycodone were followed for 270 days in 2016. Long-term opioid use was observed in 320 patients (8%). Controlling for potential confounders including morphine milligram equivalents and dosage, an estimated 12% (95 CI, 10%-14%) treated with hydrocodone transitioned to long-term use vs. 2% (95 CI, 1%-3%) on oxycodone. Among patients who received more than one prescription (n = 1,866), an estimated 23% (95 CI, 19%-26%) treated with hydrocodone transitioned to long-term use vs. 5% (95 CI, 3%-7%) on oxycodone. The difference between drugs was supported in sensitivity and subgroup analyses. Sample selection bias was not detected.

CONCLUSIONS

Long-term use was substantially greater for patients treated with hydrocodone than oxycodone, despite equianalgesia.

Falsification Tests for Instrumental Variable Designs With an Application to Tendency to Operate

BACKGROUND

Instrumental variable (IV) methods are becoming an increasingly important tool in health services research as they can provide consistent estimates of causal effects in the presence of unobserved confounding. However, investigators must provide justifications that the IV is independent with any unmeasured confounder and its effect on the outcome occurs only through receipt of the exposure. These assumptions, while plausible in some contexts, cannot be verified from the data.

METHODS

Falsification tests can be applied to provide evidence for the key IV assumptions. A falsification test cannot prove the assumptions hold, but can provide decisive evidence when the assumption fails. We provide a general overview of falsification tests for IV designs. We highlight a falsification test that utilizes a subpopulation of the data where an overwhelming proportion of units are treated or untreated. If the IV assumptions hold, we should find the intention-to-treat effect is zero within these subpopulations.

RESULTS

We demonstrate the usage of falsification tests for IV designs using an IV known as tendency to operate from health services research. We show that the falsification test provides no evidence against the IV assumptions in this application.

Evaluation of subset matching methods and forms of covariate balance

This paper conducts a Monte Carlo simulation study to evaluate the performance of multivariate matching methods that select a subset of treatment and control observations. The matching methods studied are the widely used nearest neighbor matching with propensity score calipers and the more recently proposed methods, optimal matching of an optimally chosen subset and optimal cardinality matching. The main findings are: (i) covariate balance, as measured by differences in means, variance ratios, Kolmogorov-Smirnov distances, and cross-match test statistics, is better with cardinality matching because by construction it satisfies balance requirements; (ii) for given levels of covariate balance, the matched samples are larger with cardinality matching than with the other methods; (iii) in terms of covariate distances, optimal subset matching performs best; (iv) treatment effect estimates from cardinality matching have lower root-mean-square errors, provided strong requirements for balance, specifically, fine balance, or strength-k balance, plus close mean balance. In standard practice, a matched sample is considered to be balanced if the absolute differences in means of the covariates across treatment groups are smaller than 0.1 standard deviations. However, the simulation results suggest that stronger forms of balance should be pursued in order to remove systematic biases due to observed covariates when a difference in means treatment effect estimator is used. In particular, if the true outcome model is additive, then marginal distributions should be balanced, and if the true outcome model is additive with interactions, then low-dimensional joints should be balanced. Copyright © 2016 John Wiley & Sons, Ltd.

Instrumental variable methods for causal inference

A goal of many health studies is to determine the causal effect of a treatment or intervention on health outcomes. Often, it is not ethically or practically possible to conduct a perfectly randomized experiment, and instead, an observational study must be used. A major challenge to the validity of observational studies is the possibility of unmeasured confounding (i.e., unmeasured ways in which the treatment and control groups differ before treatment administration, which also affect the outcome). Instrumental variables analysis is a method for controlling for unmeasured confounding. This type of analysis requires the measurement of a valid instrumental variable, which is a variable that (i) is independent of the unmeasured confounding; (ii) affects the treatment; and (iii) affects the outcome only indirectly through its effect on the treatment. This tutorial discusses the types of causal effects that can be estimated by instrumental variables analysis; the assumptions needed for instrumental variables analysis to provide valid estimates of causal effects and sensitivity analysis for those assumptions; methods of estimation of causal effects using instrumental variables; and sources of instrumental variables in health studies.