Inserm, U707, Research Unit in Epidemiology, Information Systems, and Modeling, Paris, France.
Epidemiology. 2010 Jan;21(1):124-7. doi: 10.1097/EDE.0b013e3181c04e70.
In early contextual studies, the aim was to demonstrate overall neighborhood influences rather than dissecting such influences into their components. Researchers did not need to worry about neighborhood-level confounding. However, as our interest shifts to the exploration of specific environmental effects, failure to consider neighborhood-level confounding may result in severely biased associations. We argue that neighborhood socioeconomic position and similar area structural factors may constitute powerful sources of confounding in studies of specific environmental factors and health. Controlling for neighborhood socioeconomic position is a convenient (but imperfect) adjustment strategy. Such control entails a minimal risk of overadjustment, but conveys a non-negligible risk of collider bias. Balancing the advantages and disadvantages, we suggest that researchers should often provide complementary analyses controlling for neighborhood socioeconomic position in studies of associations between specific environmental factors and health. Researchers should provide DAG-based descriptions of plausible scenarios to explore whether any decrease in the association of interest after adjustment for neighborhood socioeconomic position is likely due to neighborhood-level confounding, indirect pathway biases, or collider bias.
在早期的语境研究中，目的是展示整体邻里影响，而不是将这些影响分解为其组成部分。研究人员不必担心邻里层面的混杂因素。然而，随着我们的兴趣转移到探索特定环境影响，不考虑邻里层面的混杂因素可能会导致严重的偏倚关联。我们认为，邻里社会经济地位和类似的区域结构因素可能是特定环境因素与健康关联研究中混杂因素的有力来源。控制邻里社会经济地位是一种方便（但不完美）的调整策略。这种控制最小化了过度调整的风险，但存在不可忽视的混杂偏倚风险。权衡利弊，我们建议研究人员在研究特定环境因素与健康之间的关联时，通常应提供控制邻里社会经济地位的补充分析。研究人员应提供基于 DAG 的描述性分析，以探讨在调整邻里社会经济地位后，对感兴趣的关联的任何减少是否可能归因于邻里层面的混杂因素、间接途径偏差或混杂偏倚。
