RESUMO
PURPOSE: To develop a predictive model to identify hospitalized older patients at risk of functional decline. METHODS: This retrospective cohort study recruited participants aged 65 years and over admitted to internal medicine wards of a tertiary medical center in Taiwan during May to October 2017 for developing predictive model (n = 1698) and those admitted during November to December 2017 for validation study (n = 530) of the model. Demographic data, geriatric assessments and hospital conditions (admission route and length of hospital stay) were collected for analysis. RESULTS: Overall, of the 1698 participants (mean age 75.8 ± 7.9 years, 60.9% male) enrolled in the development study, 20.1% had functional decline. Results of multivariate logistic regression showed that older age, hearing impairment, history of falls within one year, risk of malnutrition, physical restraint, admission via emergency department and hospital stay ≥ 5 days were independent predictive factors for decline. A scoring system, HAD-FREE Score, constructed from the above predictive factors ranged from 0 to 18 points and ≥ 6 points was chosen as the cut-off point. The area under the receiver operating characteristic analysis was 0.748 (95% confidence interval: 0.720-0.776), the sensitivity was 65.3% and the specificity was 71.3%. Validation of the HAD-FREE Score showed moderate discriminative ability in the validation study. CONCLUSION: A HAD-FREE Score developed from seven independent factors could predict functional decline with moderate discriminative ability and good validation. This scoring system can be the basis of an automatic dynamic tracking within the electronic medical record to identify those older patients at risk of functional decline during hospitalization.