Identifying a fitness tool in early old-age to predict long-term risk of disability, severe disability, and mortality.

Population ageing has led to an increase in prevalence of old-age disability but whether the risk of disability can be detected early remains unclear. We used ten functioning/fitness measures in early old-age to identify their predictive ability for disability at older ages. A total of 4593 participants of the Whitehall II study, mean age 65.3 years, were followed for a median of 11.00 (IQR 7.25-12.67) years for incident disability [≥ 1 limitation in activities of daily living (ADL)], and severe disability (≥ 2 ADL). We first examined whether the C-statistic for each predictor, considered individually, improved the C-statistic of a model containing age and sex. We then used LASSO regression, using decrements of 0.0001 in a step-wise manner in the lambda regularization parameter to select the most important predictors. Improvement in C-statistic of a new set of LASSO predictors was tested to select the final set of predictors. Among the functioning/fitness measures considered individually, waist circumference had the highest C-statistic for disability [0.6557, 95% confidence interval (CI) 0.6367-0.6747] and severe disability (0.6868, 0.6582-0.7153). The best set of LASSO predictors (0.6617, 0.6431-0.6804) for incident disability included age, sex, waist circumference, walking speed, timed chair rises, and balance. For severe disability, the predictors were the same without walking speed (0.6955, 0.6678-0.7233). Our findings highlight the importance of obesity measures for risk of disability, and show that a small set of functioning/fitness measures can be useful in identifying individuals at higher risk of disability at older ages.