Retrospective Information From Questionnaires

The author compared medical, smoking, and dietary consumption data obtained from cases or controls and their respective next-of-kin as part of a study of colon cancer in the five Pennsylvania counties of the Philadelphia metropolitan area. The case population consisted of whites aged 45-69 years who had resided in the region for at least two years prior to diagnosis and diagnosed with colon cancer after July 1, 1976. Controls were selected using an area probability sampling scheme and were frequency-matched to the case group. Questionnaires for cases and controls were administered by interviewers; questionnaires for their next-of-kin were randomly allocated to be self- or interviewer-administered. Agreement when both respondents received the interviewer-administered questionnaire was greater than when the next-of-kin received the self-administered questionnaire for all variables. Medical and cigarette smoking variables exhibited high agreement, the percentage agreement exceeded 80 for 80% of the comparisons, the kappa statistic exceeded 0.6 for 54% of the comparisons. Diet histories were more variable (average agreement ranging from 54% to 82% and average kappa values from 0.17 to 0.59). A subanalysis of subjects and spouses showed that husbands and wives gave equally reliable responses.