ABSTRACT
Background: Delirium is a prevalent problem among older patients and it is frequently underdiagnosed. Aim: To develop and validate a clinical predictive model to identify patients at high risk of delirium. Material and Methods: Two consecutive prospective cohort studies were used to develop and validate the model. The development cohort included 542 consecutive medical inpatients, 65 years or older. The validation cohort included 85 comparable patients. A predictive score was constructed with a multivariate analysis, using variables independently associated with delirium and subsequently tested in the new cohort. Patients were assessed within the first 48 hours of admission, and every 48 hours thereafter, using the Confusion Assessment Method to diagnose delirium, evaluating also the severity of underlying disease, comorbidities, functionality, and laboratory data. Results: Delirium occurred in 192 patients (35.4%) of the development cohort and was independently associated with age and functional status assessed using the Barthel Index. With these two variables, the predictive score for delirium was developed and tested rendering an area under the receiver operating characteristic (ROC) of 0.80 (confidence intervals 0.77-0.85). Cut-off points were chosen to establish low, intermediate, and high-risk groups for delirium. According to these cut-off points, delirium frequencies in the development cohort were 8%, 23%, and 69%, and in the validation cohort 5%, 34%, and 66%, respectively (c² p < 0.05). Conclusions: This simple predictive model based on age and functional status may be a useful tool for identifying older patients risking delirium.