论文部分内容阅读
选择偏性是病例对照研究(Case-Control Studies)中的重要问题,1946年Berkson已作了有关研究并发表了文章。为了分析在队列研究((Cohort Studies)中的回答偏性和随访偏性问题,作者在文章中举了一个假设的例子,并作了分析和说明。 假定起初把250名暴露者和750名非暴露者确定为调查对象,但两组各有13％及14％的人无回答,都有14％失访。因此,实际观察对象数为183[=250×(1-27％)]和540[=750×(1-28％)],资料如表1。鉴于暴露组和非暴露组无回答及失访比例相仿,许多调查者就会认为此调查资料对于这两组是均衡的、无偏性的。事实并非如此。现假定暴露组和非暴露组全部对象的实际经历如表2,从表2和表1,又可得无回答及失访者的经历(表3)。 从表1、2、3可见,漏失了表3这批数据后,使应获的表2资料变成了实获的表1资料,从相对危险性1.69、不显著的x2值3.40,变成了相对危险性2.05和显著的x2值
Selective bias is an important issue in Case-Control Studies. In 1946 Berkson has made relevant research and published articles. In order to analyze the response bias and follow-up bias in cohort studies (Cohort Studies), the authors gave a hypothetical example in the article, and analyzed and explained. Assume that 250 exposed persons and 750 non- The exposed persons were identified as the survey subjects, but 13% and 14% of the two groups did not respond and 14% were lost. Therefore, the actual number of observed objects was 183 [=250×(1-27%)] and 540. [=750×(1-28%)], the data is shown in Table 1. Since the exposure group and the non-exposed group have no answer and the proportion of lost followers is similar, many investigators will think that the survey data is balanced and non-existent for the two groups. This is not the case, assuming that the actual experience of all subjects in the exposed and non-exposed groups is shown in Table 2. From Table 2 and Table 1, there are no answers and the experiences of the lost followers (Table 3). 1,2,3 can be seen, after missing the data in Table 3, the data in Table 2 that should be obtained has become the data obtained in Table 1, from the relative risk of 1.69, the insignificant x2 value of 3.40, into a relative Hazardous 2.05 and significant x2 values