Background. Greater attention is being paid to data quality in surveys of older age groups. In this paper patterns of item non-response are examined in a health risk appraisal instrument administered to an elderly cohort partIcipating in a randomized preventive intervention study. Methods. The association between demographic and health status factors with the number of non-responses out of 174 items was examined at baseline and at the 12-month follow-up on 1791 subjects. Results. Overall, non-response decreased from baseline to 12 months. The pattern was consistent across the seven major components of the questionnaire. Univariate analyses at baseline found that item non-response increased significantly (P < 0.05) with age, being female, being unmarried, lower annual Income, less education, and poorer personal health ranking. Polychotomous logistic regression identified age and personal health ranking as statistically significant at both baseline and 12-month follow-up assessments after controlling for all other factors. In addition, education was significant at baseline. Conclusions. These results help to identify subgroups of elderly participants who contribute to non-random patterns of missing data.