The main objective of this study was to analyze how the identification of determinants of survival is dependent on age at the time of data collection and on the length of the predictive period. The study is part of a gerontological and geriatric population investigation in Göteborg (Gothenburg), Sweden, called H70, which is a longitudinal study based on a large random sample of men and women born in 1901/1902 in Göteborg. They were first examined at the age of 70, and have then been re-examined at the ages of 75 and 79. Twelve variables from different areas were selected for use in the present study. As physical health has proven to be highly correlated to survival, the sample was split into two subgroups, one consisting of elderly with few health problems and the other of less healthy elderly. All analyses were performed on each gender separately, and on the whole samples of men and women, respectively, as well as on each health subgroup. The results show that the statistically significant determinants of survival differ substantially between the two subgroups. In addition, age at examination and length of predictive period proved to be very important in identifying determinants of survival. Lung capacity, measured be peak flow, was the only variable, among those selected for this study, that had predictive power for both genders, in both health groups and at each of the three observation ages.