Clinical and Population Health Research Program, Graduate School of Biomedical Sciences.
Department of Quantitative Health Sciences, Division of Epidemiology of Chronic Diseases and Vulnerable Populations, University of Massachusetts Medical School, Worcester, MA.
Med Care. 2019 Mar;57(3):237-243. doi: 10.1097/MLR.0000000000001063.
The use of marginal structural models (MSMs) to adjust for time-varying confounding has increased in epidemiologic studies. However, in the setting of MSMs, recommendations for how best to handle missing data are contradictory. We present a plasmode simulation study to compare the validity and precision of MSMs estimates using complete case analysis (CC), multiple imputation (MI), and inverse probability weighting (IPW) in the presence of missing data on time-independent and time-varying confounders.
Simulations were based on a cohort substudy using data from the Osteoarthritis Initiative which estimated the marginal causal effect of intra-articular injection use on yearly changes in knee pain. We simulated 81 scenarios with parameter values varied on missing mechanisms (MCAR, MAR, and MNAR), percentages of missing (10%, 20%, and 30%), type of confounders (time-independent, time-varying, either or both), and analytical approaches (CC, IPW, and MI). The performance of CC, IPW, and MI methods was compared using relative bias, mean squared error of the estimates of interest, and empirical power.
Across scenarios defined by missing data mechanism, extent of missing data, and confounder type, MI generally produced less biased estimates (range: 1.2%-6.7%) with better precision (range: 0.17-0.18) compared with IPW (relative bias: -5.3% to 8.0%; precision: 0.19-0.53). Empirical power was constant across the scenarios using MI.
Under simple yet realistically constructed scenarios, MI seems to confer an advantage over IPW in MSMs applications.
在流行病学研究中，使用边缘结构模型（MSM）来调整时变混杂因素的方法已经越来越多。然而，在 MSM 的背景下，关于如何最好地处理缺失数据的建议是相互矛盾的。我们提出了一个 plasmode 模拟研究，比较了在存在时间独立和时间相关混杂因素缺失数据的情况下，完全案例分析（CC）、多重插补（MI）和逆概率加权（IPW）对 MSM 估计的有效性和精度。
模拟基于使用来自骨关节炎倡议（Osteoarthritis Initiative）的数据的队列子研究，该研究估计了关节内注射使用对膝关节疼痛每年变化的边际因果效应。我们模拟了 81 种情况，参数值根据缺失机制（MCAR、MAR 和 MNAR）、缺失百分比（10%、20%和 30%）、混杂因素类型（时间独立、时间相关、两者都有或都没有）和分析方法（CC、IPW 和 MI）而变化。通过相对偏差、感兴趣估计值的均方误差和经验功效比较了 CC、IPW 和 MI 方法的性能。
在缺失数据机制、缺失数据程度和混杂因素类型定义的场景中，MI 通常产生的偏差较小（范围：1.2%-6.7%），精度较好（范围：0.17-0.18），而 IPW 的偏差较大（相对偏差：-5.3%至 8.0%；精度：0.19-0.53）。使用 MI 的经验功效在所有场景中保持不变。
在简单但实际构建的场景下，MI 在 MSM 应用中似乎比 IPW 具有优势。
