Agreement in self-reported personal risk factor information collected by different modes in Mongolia

This study compares two methods (a self-administered paper survey and a face-to-face interview) of collecting information about personal risk behaviours important for studies of HIV and other blood-borne pathogen transmissions in a developing country. In the framework of an epidemiological study conducted among 2504 donors in the Blood Center at the Ministry of Health, Mongolia, 2250 participants completed a short paper survey and 923 participants were interviewed concerning risk factors for hepatitis infections. A total of 669 individuals completed both surveys. McNemar's test and Kappa statistics were used to compare responses from both types of questionnaire. Kappa coefficients for health-care factors ranged from 0.57 (injection outside of hospital) to 0.81 (previous blood donation). Alcohol use and smoking were both reported more often in the interview than in the survey; the kappa coefficient was lowest (0.61) for alcohol use. While the prevalence of these behaviours depended on the mode of data collection, the association between behaviour and an outcome, hepatitis B surface antigen, was not substantially different between the two data collection methods. The results indicate that misclassification of risk behaviours is likely regardless of data collection method. However, in this study we found that biased estimates of prevalence likely did not substantially bias the estimates of association between risk factors and blood-borne infection.