评估疾病管理项目的有效性：工具变量介绍

Evaluating disease management programme effectiveness: an introduction to instrumental variables.

作者信息

Linden Ariel, Adams John L

机构信息

Linden Consulting Group, Portland, OR 97124, USA.

出版信息

J Eval Clin Pract. 2006 Apr;12(2):148-54. doi: 10.1111/j.1365-2753.2006.00615.x.

DOI:10.1111/j.1365-2753.2006.00615.x

PMID:16579823

Abstract

This paper introduces the concept of instrumental variables (IVs) as a means of providing an unbiased estimate of treatment effects in evaluating disease management (DM) programme effectiveness. Model development is described using zip codes as the IV. Three diabetes DM outcomes were evaluated: annual diabetes costs, emergency department (ED) visits and hospital days. Both ordinary least squares (OLS) and IV estimates showed a significant treatment effect for diabetes costs (P = 0.011) but neither model produced a significant treatment effect for ED visits. However, the IV estimate showed a significant treatment effect for hospital days (P = 0.006) whereas the OLS model did not. These results illustrate the utility of IV estimation when the OLS model is sensitive to the confounding effect of hidden bias.

摘要

本文介绍了工具变量（IVs）的概念，作为在评估疾病管理（DM）项目有效性时提供无偏治疗效果估计的一种方法。描述了使用邮政编码作为工具变量的模型开发过程。评估了三种糖尿病疾病管理结果：年度糖尿病费用、急诊科（ED）就诊次数和住院天数。普通最小二乘法（OLS）和工具变量估计均显示糖尿病费用有显著治疗效果（P = 0.011），但两种模型对急诊科就诊次数均未产生显著治疗效果。然而，工具变量估计显示住院天数有显著治疗效果（P = 0.006），而OLS模型则没有。这些结果说明了当OLS模型对隐藏偏差的混杂效应敏感时，工具变量估计的效用。