Validity of questionnaire information on frequency of coitus
Hornsby, P.P.; Wilcox, A.J.
American Journal of Epidemiology 130(1): 94-99
1989
ISSN/ISBN: 0002-9262 PMID: 2787113 DOI: 10.1093/oxfordjournals.aje.a115326Document Number: 363909
Researchers interviewed 91 women of reproductive age who were married or living with a partner when they enrolled in a prospective study to compare the frequency of coitus on the questionnaire with the frequency on prospective daily record cards (mean, 70 days of records). 89% of the women reported a higher frequency of intercourse on the questionnaire than on the prospective records. The mean frequency of intercourse on the questionnaire was 2.5 times/week while the daily record cards indicated a mean of 1.7 times/week--a statistically significant mean difference of .8 times/week (p.0001). .4 times/week was the mean lowest value for all women and 3.3 times/week was the mean highest frequency. In addition, 75% of the women reported no intercourse for a least 1 week. The researchers guessed that the most likely explanation for the overestimate may be that the frequency reported on the questionnaire doe not take into account menstruation, illness, travel, or other conditions that might reduce the frequency of intercourse. When they calculated a frequency of coitus on all days, excluding menses, however, the mean difference was reduced only slightly to .6 times/week. Therefore the researchers could not explain the bulk of the difference. The myriad of potential reasons stated in the article for differences in the 2 sources of intercourse data suggest possible improvements in questionnaire design are needed, such as asking about factors that affect their usual intercourse. In addition, despite the fact that the differences constitutes an information bias, the bias occurs in the same direction and magnitude in all the various subgroups and thus is nondifferential.