Identification of high-risk individuals for the development of disability in activities of daily living. A ten-year follow-up study

Independence in activities of daily living (ADL) is important in an aging population. ADL disability is a multifactorial problem, therefore a multifactorial approach is needed in the prediction of ADL disability. Our objective is to identify predictors for the development of ADL disability over a course of ten years in middle-aged and older persons. In a prospective cohort study, 478 middle-aged and older persons (61.2years, range 40–78years) without ADL disability at baseline were included. ADL disability was measured using the Katz-questionnaire. We included the following candidate predictors: number of chronic diseases, MMSE, Short Physical Performance Battery, leg strength, handgrip strength, physical activity, cholesterol/HDL ratio, BMI, pulse wave velocity, the degree of urbanization, age, gender and socioeconomic status. Associations between candidate predictors and ADL disability were examined using Poisson regression analysis. Performance of the prediction model was assessed with calibration and discrimination measures. The number of chronic diseases, muscle strength, age, gender and socioeconomic status were predictors of ADL disability at ten-year follow-up. The model showed a good calibration and discrimination (c-statistic: 0.83) between persons who will and will not develop ADL disability. In conclusion, the present study showed that using a multifactorial prediction model – based on easily and readily available measurements – individuals who are at high risk of developing ADL disability could be identified. The prediction model could be used as a screening tool to identify which persons most likely benefit from preventive strategies and interventions. ► We developed a model to predict the risk of ADL disability over a course of ten years. ► Aim prediction model: to timely identify high-risk individuals ► Predictors included chronic diseases, muscle strength, age, gender and socioeconomic status. ► The model's c-statistic was 0.84, which indicates a good discriminative ability.