Comparison of five indices for prediction of adverse outcomes in hospitalised Mexican older adults: A cohort study

The aim of this prospective study was to investigate the ability of five indices of risk stratification to predict functional decline and prolonged length of stay (LOS) in older Mexicans hospitalized in the acute care setting. A total of 254 patients aged ≥60 years were followed up. Risk indices were constructed from baseline data collected during the first 48h of ward admission, and included: Frailty Index (FI), Hospital Admission Risk Profile (HARP), Score Hospitalier d'Evaluation du Risque de Perte d'Autonomie (SHERPA), Acute Physiology and Chronic Health Evaluation II (APACHE II) and Charlson's Co-morbidity Index (CCI). Area Under Receiver Operating Characteristic (_auROC) curves was used to compare the ability of risk indices to predict adverse outcome, with outcomes of interest being prolonged LOS, and functional decline, the latter defined as ≥10% drop in Barthel Index score across hospitalization. Mean (SD) FI score was 0.31 (0.14). Effective in predicting long LOS were FI, SHERPA and APACHE II; effective in predicting functional decline were SHERPA and HARP. Indices generally showed high specificity values (most were >80%), although all indices lacked adequate sensitivity values for outcome prediction (<80%). Geriatricians could use information from FI, SHERPA, APACHE II, HARP to guide patient management decisions. However, given that all indices lacked accuracy of prediction, results should be interpreted with caution.