Spreading interviews over time in health surveys: Do temporal variations of self-reported alcohol consumption affect measurement?
Spreading interviews over time in health surveys: Do temporal variations of self-reported alcohol consumption affect measurement?
复制标题
DOI：
10.1081/ja-200030718
复制
发表时间：
2005-01-01
影响因子：
2
通讯作者：
Gmel, G
中科院分区：
文献类型：
--
作者：
Heeb, JL;Gmel, G
客观的。解决电话健康调查中酒精消费自我报告中与访谈当天相关的系统性差异。调查包括时间聚类效应、酒精消耗量的预测以及受访者特征和访谈期间的不同天数的变化，以及对测量仪器变化的敏感性。方法。使用了 1999 年春季从瑞士一项纵向概率一般人口调查中 2846 名参与者收集的基线数据。该研究对瑞士饮酒者具有代表性。酒精消耗量测量包括 6 个月数量频率和 1 周渐进频率测量。结果。即使在控制了样本特征之后，在分级频率测量中也发现了与访谈当天相关的自我报告的系统变化的证据。在数量频率测量上也发现了类似的变化，但在对样本特征进行统计控制后不再显着。结论。基于短参考期的分级频率测量的电话访谈酒精调查研究中的统计推断可能会受到与访谈当天的聚类效应相关的错误的困扰。因此，在统计分析中应考虑进行实地工作的时间方面。
Objective. To address systematic variations related to the day of the interview in self-reports of alcohol consumption in telephone health surveys. The investigations include temporal clustering effects, prediction of alcohol consumption and variations across days by characteristics of respondents and interviewing period, and sensitivity, to variations of measurements instruments. Method. Data at baseline collected in Spring 1999 from 2846 participants in a longitudinal probabilistic general-population survey in Switzerland were used. The study is representative for drinkers in Switzerland. Alcohol consumption measures include a 6-month quantity frequency and a 1-week graduated frequency measure. Results. Evidence for systematic variations in self-reports related to the day of interview was found on the graduated frequency measure even after controlling for sample characteristics. Similar variations on the quantity frequency measure were found, but were no longer significant after statistical control of the sample characteristics. Conclusions. Statistical inference in alcohol survey research by telephone interviews based on graduated frequency measures with short reference period may be plagued with errors related to clustering effects of the day of the interview. Temporal aspects of conducting the fieldwork should therefore be accounted for in statistical analysis.