Exceptional longevity – Predictors and methods for handling missing data.
The average lifespan has increased greatly over the last 150 years and the age group over 80 years has been the fastest growing age segment the last 50 years. Despite this marked increase of the oldest old there is very little knowledge of which predictors that are associated with exceptional long life. For the oldest old there has been some studies indicating that level and change of cognition and physical ability is associated with mortality. In addition, phenotypes such as smoking and cancer have lost their association with mortality among the oldest old.
All these studies has been conducted from questionnaire surveys, and therefore lack data from the ones who don’t participate or drop out of the study and from those who don’t perform the tests. Most of these studies didn’t take the missing data into account and since the missing data usually is from the ones with the poorest health, this could introduce bias in the results.
The main objective of this project is to get an increased insight in the ageing process and to study the factors which are associated with good functional capacity and exceptional long life. The project will use longitudinal data of men and women older than 70 years, which are already collected. Especially the analysis of Danes born from 1897 to 1914 which are older than 85 years will be the primarily focus in this project. For these cohorts there are longitudinal measures, long follow-up and almost all are diseased which gives fewer problems with censoring. Analysis of socioeconomic data, BMI, self-assessed health, depressions symptomatology, cognition and physical variables and how they interact with each other will form the basis to answer the following questions:
- Which factors predicts mortality of the oldest old, and which combination of predictors is strongest associated with exceptional long life?
- Does the association between predictors and mortality change after 70 years of age, and is the association different compared to how it is at the age of 50?
- How do the predictors change within individuals, and how is this change associated with mortality?
- Who big an influence will missing data have on the above estimates, and how much can be adjusted for by sound statistical modeling?
The Danish 1905-cohort, Glostrup 1897- and 1914-cohort and The Longitudinal Study of Aging Danish Twins (LSADT) will be the foundation of the analysis, thus including approximately 9,000 individuals at the age of 70-100, which are visited up to 7 times.