Abstract
BACKGROUND: There are a very few studies focusing on the individual-based survival with a long follow-up time. AIM: To identify predictors and determine their joint predictive value for longevity using individual-based outcome measures. METHODS: Data were drawn from Tampere Longitudinal Study on Aging (TamELSA), a study of individuals' age 60-89 years (N = 1450) with a mortality follow-up of up to 35 years. Two measures of longevity were used: the longevity difference (LD) and realized probability of dying (RPD), both of which compare each individual's longevity with their life expectancy as derived from population life tables. Independent variables were categorized into five domains: sociodemographic, health and functioning, subjective experiences, social activities, and living conditions. Linear regression models were used in three steps: bivariate analysis for each variable, multivariate analysis based on backward elimination for each domain, and one final model. RESULTS: The most important predictors of both outcomes were marital status, years smoked regularly, mobility, self-rated health, endocrine and metabolic diseases, respiratory diseases, and unwillingness to do things or lack of energy. The explained variance in longevity was 13.8% for LD and 14.1% for RPD. This demonstrated a large proportion of unexplained error margins for the prediction of individual longevity, even though many known predictors were used. DISCUSSION AND CONCLUSIONS: Several predictors associated with longer life were found. Yet, on an individual level, it remains difficult to predict who will live longer than their age peers. The stochastic element in the process of aging and in death may affect this prediction.