Wen S W, Kramer M S
The Bureau of Reproductive and Child Health, Laboratory Centre for Disease Control, Health Canada, Ottawa, Ontario.
J Clin Epidemiol. 1999 Jan;52(1):7-12. doi: 10.1016/s0895-4356(98)00136-x.
Because of the potential for confounding by indication (disease severity) in individual-level observational studies of intended treatment effects, a treatment designed to prevent an adverse event may appear to cause it. We use a hypothetical example to show that despite substantial variation in the frequency of treatment among patients residing in different geographic areas, a constant area-specific mortality rate can be observed, indicating the absence of confounding by indication at the ecologic level. The advantage of ecologic over individual-level observational studies in the assessment of intended treatment effects holds even if variations in disease severity, socioeconomic status, and other unmeasured factors are taken into account, as long as treatment utilization is influenced by practice style in the local medical community independently of disease severity. Ecologic studies can suggest the need for changes in practice, help resolve ethical issues, and indicate priorities for randomized trials.
由于在针对预期治疗效果的个体水平观察性研究中存在因指征（疾病严重程度）导致的混杂可能性，一种旨在预防不良事件的治疗方法可能看起来反而会导致该不良事件。我们用一个假设的例子来说明，尽管居住在不同地理区域的患者接受治疗的频率存在很大差异，但仍可观察到特定区域的死亡率恒定，这表明在生态水平上不存在因指征导致的混杂。在评估预期治疗效果时，生态研究相对于个体水平观察性研究的优势依然存在，即使考虑到疾病严重程度、社会经济地位及其他未测量因素的差异，只要治疗的使用受当地医疗社区的实践方式影响，且独立于疾病严重程度。生态研究可以提示实践变革的必要性，有助于解决伦理问题，并指明随机试验的重点。
