Hatch M C, Misra D, Kabat G C, Kartzmer S
Division of Epidemiology, Columbia University School of Public Health, New York, NY 10032.
Am J Epidemiol. 1991 Apr 15;133(8):826-31. doi: 10.1093/oxfordjournals.aje.a115962.
The quality of proxy reporting was assessed among 136 prenatal patients and their spouse/partners recruited from the obstetric services of a New Jersey hospital between 1985 and 1987. The concordance, sensitivity, and specificity of proxy reports about partners' occupation, smoking, and drinking were examined in relation to self-reports. Overall, private patients provided better proxy data than did clinic patients, and women provided better data than did men. No consistent effects on the quality of proxy reports were found in relation to age, level of education, marital status, or length of cohabitation. Partners' recent job titles appeared to be quite accurately reported, whereas partners' smoking and drinking patterns were less well-reported. For alcohol use in particular, there was evidence of considerable misclassification resulting from proxy reports even when kappa statistics and intraclass correlation coefficients suggested good agreement. Use of proxy respondents is unnecessary in reproductive studies and should be avoided when it may produce misleading results. Our data indicate that private prenatal patients and their partners can give reasonable proxy reports about job titles and smoking, but not about alcohol use. The high proportion of clinic patients who did not refer a partner (or whose partners could not be contacted) limits the generalizability of our results for this group and gives cause for concern about collecting proxy information from clinic populations.
1985年至1987年间，从新泽西一家医院的产科服务部门招募了136名产前患者及其配偶/伴侣，对代理报告的质量进行了评估。考察了代理报告中关于伴侣职业、吸烟和饮酒情况与自我报告的一致性、敏感性和特异性。总体而言，自费患者提供的代理数据比门诊患者更好，女性提供的数据比男性更好。在年龄、教育程度、婚姻状况或同居时间方面，未发现对代理报告质量有一致影响。伴侣最近的职位似乎报告得相当准确，而伴侣的吸烟和饮酒模式报告得较差。特别是对于饮酒情况，即使kappa统计量和组内相关系数显示一致性良好，代理报告仍存在相当程度的错误分类证据。在生殖研究中，使用代理受访者是不必要的，当可能产生误导性结果时应避免使用。我们的数据表明，自费产前患者及其伴侣可以就职位和吸烟情况给出合理的代理报告，但对于饮酒情况则不然。未提及伴侣（或其伴侣无法联系到）的门诊患者比例很高，这限制了我们的结果对此类人群的普遍性，也引发了对从门诊人群中收集代理信息的担忧。
