Goldberg Marcel, Chastang Jean François, Zins Marie, Niedhammer Isabelle, Leclerc Annette
INSERM Unité 687-IFR 69, Hôpital National de Saint-Maurice, 14, rue du Val d'Osne, F-94415 Saint-Maurice Cedex, France.
J Clin Epidemiol. 2006 Nov;59(11):1213-21. doi: 10.1016/j.jclinepi.2006.02.020. Epub 2006 Aug 30.
This study examined socioeconomic, lifestyle, and health factors associated with response to annual mail questionnaires in a longitudinal study from 1990 through 2000 within the French GAZEL cohort.
Twenty thousand six hundred twenty-four participants volunteered in 1989, and received each year a questionnaire. As responding one given year was not independent of responding in other years, mixed models were used to analyze the variables associated with response.
Higher response rates were associated with gender (male), age (older), managerial status, and retirement. Smoking and alcohol drinking at baseline were associated with lower participation. Subjects who had at least three sick leaves were less likely to respond, especially for absence for psychiatric and alcohol-related diseases among men. Those who had rated their health as bad at baseline were less prone to respond during the follow-up. Attrition in subsequent response to questionnaires was associated with cancer diagnosis and with episodes of coronary heart disease for men.
Most of the variables that predicted initial participation were also associated with continued participation during follow-up. Health problems strongly predicted attrition, whereas socioeconomic factors played a weaker role. Withdrawing is a reversible state and considering only one episode of nonparticipation could be misleading.
本研究在法国GAZEL队列1990年至2000年的一项纵向研究中，调查了与年度邮件调查问卷回复情况相关的社会经济、生活方式和健康因素。
2624名参与者于1989年自愿参与研究，并每年收到一份问卷。由于某一年的回复并非独立于其他年份的回复，因此使用混合模型分析与回复相关的变量。
较高的回复率与性别（男性）、年龄（较大）、管理职位和退休有关。基线时吸烟和饮酒与较低的参与度有关。至少有三次病假的受试者回复的可能性较小，尤其是男性因精神疾病和酒精相关疾病缺勤的情况。那些在基线时将自己的健康状况评为不佳的人在随访期间回复的可能性较小。男性后续问卷回复的流失与癌症诊断和冠心病发作有关。
大多数预测初始参与的变量也与随访期间的持续参与有关。健康问题强烈预测流失，而社会经济因素的作用较弱。退出是一种可逆状态，仅考虑一次不参与情况可能会产生误导。
