Prediction Model of 30-Day Mortality in Elderly Patients Admitted to Geriatric Acute Ward Using Comprehensive Geriatric Assessment Domain

Introduction. Mortality of hospitalized elderly patients remains high. This study aimed to develop and to determine the performance of a prediction model for 30-day mortality in elderly patients hospitalized in geriatric acute ward using comprehensive geriatric assessment (CGA) domain. Methods. A retrospective cohort study was conducted using medical records of elderly patients (> 60 years) hospitalized in acute geriatric ward. Nine predictors (age, sex, delirium, comorbidity, albumin level, comorbidity [CIRS-G], psycho-affective status, cognitive status, and nutrition status [MNA]) were analyzed. Multivariate analysis using cog regression of significant predictors was done to determine hazard ratio (HR) for each predictor and to develop prediction model. The model’s calibration performance and its discrimination ability were respectively determined by Hosmer-Lemeshow test and area under the receiver-operating-characteristic curve (AUC). Results. There were 530 subjects with the median age was 69 (range 60-96)) years old. The 30-day mortality was 28.1%. Delirium (HR 4.11 [95% CI 1.83-9.11]), albumin < 3 mg/dl (HR 2,18 [95% CI 1.23-3.85]), Barthel index < 9 (HR 2.21 [95%CI 1.23-3.85]), and malnutrition (MNA < 17) (HR 1,77 (95% CI 1.19-2.63)] were significant predictors of 30-day mortality. Prediction model of mortality was stratified into 3 groups: lower risk (4.4%), medium risk (24.8%), and high risk (64.3%). The Hosmer-Lemeshow showed good precision (p = 0.409) and the AUC revealed good discrimination ability (84.3% [95% CI 80.7-87.9]). Conclusion. Prediction model of 30-day mortality based on CGA domain has good precision and discrimination ability.