Bias: Considerations for research practice
Bias: Considerations for research practice
复制标题
DOI：
10.2146/ajhp070369
复制
发表时间：
2008-11-15
影响因子：
2.7
通讯作者：
Gerhard, Tobias
中科院分区：
文献类型：
--
作者：
Gerhard, Tobias
目的.综述了偏倚的概念及其对研究结果的影响。理解偏见对于进行合理的研究至关重要。在研究方法论的背景下，偏倚是指研究中存在系统性误差。本报告介绍了偏倚的概念，并描述了将偏倚引入医疗保健干预研究的三个主要领域：（1）与患者在人群中接受治疗相关的因素，（2）影响患者纳入研究的因素，以及（3）与评估和测量相关的因素。在研究的设计和分析阶段的方法，以解决偏见，包括限制，匹配，多变量分析，工具变量估计，并防止错误分类的信息。虽然某些偏倚是特定于某些研究设计和方法的，但如何将偏倚引入研究的一般概念同样适用于随机对照试验、前瞻性观察性或准实验性研究以及回顾性研究。设计之间的一个重要区别是研究者有多少机会避免偏倚和影响测量的准确性，范围从随机对照试验中理想的完全控制到自动数据库的回顾性研究中几乎没有控制。研究者应致力于避免研究设计中的偏倚，如果偏倚无法避免，应调整研究分析中的偏倚，并量化和讨论剩余偏倚对研究结果的影响。
Purpose. The concept of bias and its effects on study results are reviewed.Summary. Understanding bias is paramount for the conduct of sound research studies. In the context of research methodology, bias refers to the presence of systematic error in a study. This report serves as an introduction to the concept of bias and describes three major areas through which bias is introduced into research studies of health care interventions: (1) factors that relate to the exposure of patients to treatments in the population, (2) factors that influence inclusion of patients in the study, and (3) factors related to assessment and measurement. Methods to address bias in both design and analysis stages of a study are presented, including restriction, matching, multivariate analysis, instrumental variable estimation, and the prevention of misclassification of information. While some biases are specific to certain study designs and approaches, the general concepts of how bias is introduced into a study apply equally to randomized controlled trials, prospective observational or quasiexperimental studies, and retrospective studies. An important difference between the designs is how much opportunity the investigator has to avoid bias and influence accuracy of measurements, ranging from ideally full control in randomized controlled trials to almost no control in retrospective studies of automated databases.Conclusion. Investigators should aim to avoid bias in the design of a study, adjust for bias in the study analysis if bias cannot feasibly be avoided, and quantify and discuss the effects of residual bias on study results.