Validity of the responses to self-administered questionnaires as compared with the responses to interviews using a...

A comparison of responses obtained in interviews using a structured questionnaire and those obtained using a self-administered questionnaire about parental history of liver diseases and personal smoking and drinking habits was made to evaluate whether the self-administered questionnaire is an effective means for collecting information on these variables in place of an interview. The subjects used in the present analyses were 46 male hepatocellular carcinoma (HCC) cases and 46 hospital controls who were interviewed between January, 1996 and January, 1997 for HCC cases and between January, 1996 and April, 1997 for hospital controls. Ages at start or stop of smoking and drinking were grouped into four categories as categorical variables. The cumulative number of cigarettes and cumulative alcohol consumption were also grouped into four categories. Generally speaking, the degree of agreement was substantially high for both categorical and continuous variables, particularly for parental history of liver diseases, and smoking (ICC = 0.929 and 0.912 for cases and controls, respectively) and drinking (ICC = 0.899 and 0.882, respectively) habits that are originally nominal categorical variables. On the other hand, the cumulative number of cigarettes (ICC = 0.928 and 0.716, respectively) and the cumulative alcohol consumption (ICC = 0.761 and 0.825, respectively) seemed to have a lower degree of agreement than other variables. For these variables, there was a tendency for the self-administered questionnaire to produce larger values compared with the interview. Therefore, the self-administered questionnaire seemed to be an effective means for collecting information on categorical variables, while the interview using a structured questionnaire seemed to be preferable to the self-administered questionnaire with regard to the lifetime exposure to cigarettes and alcohol.