Introduction: Delirium is a frequent pathology in the elderly presenting to the emergency department (ED) and is seldom recognised. This condition is associated with many medical complications and has been shown to increase the hospital length-of-stay. The objective of this study was to identify the predictor factors of developing delirium in this high-risk population. Methods: Design: This study was part of the multicenter prospective cohort INDEED study. Participants: Patients aged 65 and older, initially free of delirium and with an ED stay of 8h or longer, were followed up to 24h after ward admission. Measures: Clinical and demographic variables were collected by interview and chart review. A research professional assessed their delirium status twice daily using the Confusion Assessment Method (CAM). Analyses: A classification tree was used to select predictors and cut-points that minimized classification error of patients with incident delirium. After literature review, nineteen predictors were considered for inclusion in the model (eight non-modifiable and eleven modifiable factors). Results: Among the 605 patients included in this study, incident delirium was detected by the CAM in 69 patients (11.4%). In total, fourteen variables were included in a preliminary model, of which six were intrinsic to the patient and eight were modifiable in the ED. Variables with the greatest impact in the prediction of delirium includes age, cognitive status, ED length of stay, autonomy in daily activities, fragility and mobility during their hospital stay. The diagnostic performance of the model applied to the study sample gave a sensitivity of 78.3% (95% CI: 66.7 to 87.3), a specificity of 100.0% (95% CI: 99.3 to 100.0), a PPV of 100.0% (95% CI: 93.4 to 100.0) and a NPV of 97.3% (95% CI: 95.6 to 98.5). Conclusion: The delirium risk model developed in this study shows promising results with elevated sensitivity and specificity values. Considering the limited ability to predict and detect delirium among physicians, the potential increase in sensitivity provided by this tool could be beneficial to patients. This model will ultimately serve to identify high-risk patients with the goal of developing strategies to alter modifiable risk factors and subsequently decrease the incidence of delirium in this population.
* Views captured on Cambridge Core between <date>. This data will be updated every 24 hours.
Usage data cannot currently be displayed