Background and objective: The main aims of this study were to evaluate the separate and joint effects of respiratory, sensory, and psychomotor function, muscle strength, and mobility in predicting mortality in older men and women, and to find a way to control multicollinearity in a multivariate Cox regression model.
Methods: Mortality was followed for 10 years (1990-2000) in an entire cohort of 75-year-old residents of the city of Jyvaskyla, Finland (born in 1914; N=388). Cox regression models and principal component estimation were employed to study the association between the covariates and mortality.
Results: The study indicated that, after adjustment for fatal diseases and cognitive capacity, vital capacity, tapping rate, muscle strength, and walking speed predict mortality in elderly men and women to a similar extent. The hazard ratio for mortality for those in the lowest tertile of the principal component representing all the significant physiological functions was from three to four times higher than for those in the highest tertile.
Conclusion: Our results suggest that a poor level in any of the measured functions has a similar impact on the risk for mortality, which is further increased by combining the different results. No single, dominant predictor of mortality emerged.