The purpose of the present study was to investigate the baseline characteristics related to the participation status in a longitudinal epidemiological study for the elderly. Nine hundred twenty-six self-supporting older adults lived in a local city located southwest in Japan were classified into the following 3 groups; non-participants (NP, n=170), participants for the baseline measurements (PB, n=278), and participants for both the baseline and follow-up measurements after two-years (PBF, n=478). The basic attributes including socioeconomic status, depressive symptoms, cognitive function, sleep quality, fitness, objectively- measured physical activity, habitual behavior, psychosocial characteristics and functional problems in daily life were assessed. Regarding the NP group, data were collected by questionnaire or home visits. According to the results of a multinomial logistic regression analysis adjusted for age, sex and years of education, significant odds ratios which indicate a poor status in most of the assessment items were observed in both the NP and PB groups in comparison to the PBF group. We concluded that sex (female), a higher age, fewer years of education, a cognitive function decline, depressive symptoms, a low fitness level, physical activity and daily living function, and poor psychosocial backgrounds were related to the non-participation and dropouts. These results should be considered in a longitudinal data analysis of the present study, as well as in the design setting of epidemiological studies dealing with physical activity indexes and care risk factors. Furthermore, these findings are also considered to be useful as basic data for such care prevention activities in the community.