Estimating treatment effects on healthcare costs under exogeneity: is there a 'magic bullet'?
Estimating treatment effects on healthcare costs under exogeneity: is there a 'magic bullet'?
复制标题
DOI：
10.1007/s10742-011-0072-8
复制
发表时间：
2011-07-01
影响因子：
1.5
通讯作者：
Manning, Willard G.
中科院分区：
文献类型：
--
作者：
Basu, Anirban;Polsky, Daniel;Manning, Willard G.
在没有未测量混杂因素的假设下，估计平均治疗效应（ATE）的方法包括回归模型;使用分层、加权或匹配的倾向评分（PS）调整;以及双重稳健估计量（两者的组合）。研究人员继续争论医疗保健成本数据等结果的最佳估计量，因为它们通常具有不对称分布和异质性治疗效果的特点。在文献中有很好的记录，在寻找回归模型的正确规格的挑战。倾向得分估计提出了替代克服这些挑战。使用模拟，我们发现在中等规模的样本（n = 5，000）中，平衡从饱和规格估计的PS可以平衡治疗组之间的协变量均值，但无法平衡协变量之间的高阶矩和协方差。因此，与回归模型不同，即使不需要用于结果的正式模型，PS估计器在最好的情况下可能是低效的，并且在最坏的情况下对于医疗保健成本数据是有偏差的。我们的模拟研究，旨在采取“矛盾证明”的方法，证明没有一个估计可以被认为是最好的所有数据生成过程中的结果，如成本。逆倾向加权估计量是最有可能是无偏的替代数据生成过程下，但在PS模型的误设定下容易出现偏差，并且与无偏回归估计量相比效率低下。我们的研究结果表明，在估计医疗保健费用的治疗效果时，没有“灵丹妙药”。在天真地应用任何一个估计量来估计这些数据中的ATE之前，应该小心。我们说明了替代方法在乳腺癌治疗成本数据集中的性能。
Methods for estimating average treatment effects (ATEs), under the assumption of no unmeasured confounders, include regression models; propensity score (PS) adjustments using stratification, weighting, or matching; and doubly robust estimators (a combination of both). Researchers continue to debate about the best estimator for outcomes such as health care cost data, as they are usually characterized by an asymmetric distribution and heterogeneous treatment effects,. Challenges in finding the right specifications for regression models are well documented in the literature. Propensity score estimators are proposed as alternatives to overcoming these challenges. Using simulations, we find that in moderate size samples (n = 5,000), balancing on PSs that are estimated from saturated specifications can balance the covariate means across treatment arms but fails to balance higher-order moments and covariances amongst covariates. Therefore, unlike regression model, even if a formal model for outcomes is not required, PS estimators can be inefficient at best and biased at worst for health care cost data. Our simulation study, designed to take a 'proof by contradiction' approach, proves that no one estimator can be considered the best under all data generating processes for outcomes such as costs. The inversepropensity weighted estimator is most likely to be unbiased under alternate data generating processes but is prone to bias under misspecification of the PS model and is inefficient compared to an unbiased regression estimator. Our results show that there are no 'magic bullets' when it comes to estimating treatment effects in health care costs. Care should be taken before naively applying any one estimator to estimate ATEs in these data. We illustrate the performance of alternative methods in a cost dataset on breast cancer treatment.