Predictive validity of a functional classification method in older adults

Abstract

Objective. Evaluate the predictive validity of a functional classification (FC) method for the use of emergency services and hospitalization, mortality, and health care costs among older adults. Methods. Retrospective cohort study that included 2 168 older adults in a chronic noncommunicable disease (CNCD) care program in Medellin, Colombia. Patients were stratified according to a FC method based on functional status, presence of risk factors, and control of comorbidity. During one year of follow-up, the predictive validity of the FC method was assessed for the studied outcomes. Discrimination and calibration were measured with the C-statistic and Hosmer-Lemeshow (HL) test, respectively. Results. The average age was 74.6 ± 7.9 years; 40.8% (n = 884) were men and 7.7% (n = 168) died. The risk of death (odds ratio [OR]: 1.767; 3.411; 8.525), hospitalization (OR: 1.397; 2.172; 3.540) and high cost of health care (OR: 1.703; 2.369; 5.073) increased in proportion to a deterioration in functional classification (classes 2B, 3, and 4, respectively). The predictive model for the outcome of death showed good capacity for discrimination (C-statistic = 0.721) and calibration (HL statistic 10.200; P = 0.251). Conclusion. There is a dose-response relationship between deterioration in FC and a higher risk of death, hospitalization, and high cost. FC has predictive validity for the mortality rate and could be used to stratify older adults in CNCD care programs with a view to guiding interventions.