Prediction of Mortality in Older Hospitalized Patients after Discharge as Determined by Comprehensive Geriatric Assessment

Several dimensional impairments regarding Comprehensive Geriatric Assessment (CGA) have been shown to be associated with the prognosis of older patients. The purpose of this study is to investigate mortality prediction factors based upon clinical characteristics and test in CGA, and then subsequently develop a prediction model to classify both short- and long-term mortality risk in hospitalized older patients after discharge. A total of 1565 older patients with a median age of 81 years (74.0–86.0) were consecutively enrolled. The CGA, which included assessment of clinical, cognitive, functional, nutritional, and social parameters during hospitalization, as well as clinical information on each patient was recorded. Within the one-year follow up period, 110 patients (7.0%) had died. Using simple Cox regression analysis, it was shown that a patient’s Length of Stay (LOS), previous hospitalization history, admission Barthel Index (BI) score, Instrumental Activity of Daily Living (IADL) score, Mini Nutritional Assessment (MNA) score, and Charlson’s Comorbidity Index (CCI) score were all associated with one-year mortality after discharge. When these parameters were dichotomized, we discovered that those who were aged ≥90 years, had a LOS ≥ 12 days, an MNA score < 17, a CCI ≥ 2, and a previous admission history were all independently associated with one-year mortality using multiple cox regression analyses. By applying individual scores to these risk factors, the area under the receiver operating characteristics curve (AUC) was 0.691 with a cut-off value score ≧ 3 for one year mortality, 0.801 for within 30-day mortality, and 0.748 for within 90-day mortality. It is suggested that older hospitalized patients with varying risks of mortality may be stratified by a prediction model, with tailored planning being subsequently implemented.