Prediction of survival in 70-year olds

The purpose of this study was to find predictors of 10-year survival among 70-year olds and to show that there might be more than one good set of predictors. All variables found to be significantly related to survival in univariate tests (predictors) are given a risk value (0 = no risk, 1 = risk), and classified into high-order domains. The number of domains containing predictors with risk values are then counted, giving the individual a certain risk index. The method was applied on data from the longitudinal population study of 70-year-old people in Gothenburg, Sweden, the H70-study. In a random sample of 474 men and 562 women predictors for survival over a 10-year period, from the age of 70, were selected for men and women, respectively. The predictors were classified into seven domains: socio-economic status, life style, social networks, need of help, biomarkers of ageing, physical health and cognitive function. The analysis revealed a consistent pattern in which survival was related to the number of risk domains. In the no-risk sub-sample 12.5% of the men were deceased after 10 years, in the sub-sample with five or more risk domains involved 78.0% were deceased. The corresponding figures for women were 10.0% and 60.0%. The results show that it is possible to find variables that can predict survival. However, there is not only one set of good predictors. As an alternative to use of specific variables as predictors the presumptive predictors can be grouped into a few main domains and good predictions be made on the basis of these domains.