Prognosis/Risk Indices for Function/Mortality in Elders
Funded Grant
Overview
Affiliation
View All
Overview
description
The goal of this project is to develop and validate two prognostic indices that will stratify community living elders into groups according to risk for two outcomes: mortality and functional deterioration. Our results will be important to efforts to improve outcomes in older people because prognostic indices facilitate efforts to improve healthcare on the clinical level (physician counseling of patients), the policy level (risk adjustment and comparing outcomes across providers), and the public health level (evaluating the effect of a risk factor on outcomes). In spite of the important role prognostic indices could play in improving outcomes in older people, there are few appropriately validated indices available for use in general groups of community-dwelling elders. This project is guided by a conceptual framework that recognizes that outcomes in older people are the result of multiple domains of risk that interact together. These domains include demographic factors (age and gender), biomedical factors (disease), functional status (physical, cognitive, and psychological), and social factors (social support, ethnicity and SES). Our consideration of multiple domains that are clearly important to outcomes in the elderly will be an innovative component of this study. We will use a unique database of 7447 subjects (age less than or equal to 70) compiled from the AHEAD study to develop and validate our prognostic indices. This database is unique 166th in terms of the diversity of its subjects and the availability of information on most of the important domains of baseline risk. We will first divide the dataset into derivation and validation components. Next, we will describe the relation of each predictor variable to each outcome (survival time over five years, and increased dependence in ADL function over two years) in the derivation set. Within each risk domain, we will determine the variables that best predict each outcome. Next, we will use multivariate methods to determine which variables independently predict each outcome. We will use these multivariable models to develop a simple point scoring system to stratify subjects into categories at variable risk of each outcome. Finally, we will use the validation set to test the accuracy and transportability of our prognostic indices, as measured by their calibration and discrimination. The results of this project will be prognostic indices useful to clinicians, epidemiologic and health services investigators, and policy makers.
