Development and validation of a predictive score for ICU delirium in critically ill patients

Abstract

Background: The incidence of delirium in intensive care unit (ICU) patients is high and associated with a poor prognosis. We validated the risk factors of delirium to identify relevant early and predictive clinical indicators and developed an optimized model. Methods: In the derivation cohort, 223 patients were assigned to two groups (with or without delirium) based on the CAM-ICU results. Multivariate logistic regression analysis was conducted to identify independent risk predictors, and the accuracy of the predictors was then validated in a prospective cohort of 81 patients. Results: A total of 304 patients were included: 223 in the derivation group and 81 in the validation group, 64(21.1%)developed delirium. The model consisted of six predictors assessed at ICU admission: history of hypertension (RR = 4.367; P = 0.020), hypoxaemia (RR = 3.382; P = 0.018), use of benzodiazepines (RR = 5.503; P = 0.013), deep sedation (RR = 3.339; P = 0.048), sepsis (RR = 3.480; P = 0.018) and mechanical ventilation (RR = 3.547; P = 0.037). The mathematical model predicted ICU delirium with an accuracy of 0.862 (P < 0.001) in the derivation cohort and 0.739 (P < 0.001) in the validation cohort. No significant difference was found between the predicted and observed cases of ICU delirium in the validation cohort (P > 0.05). Conclusions: Patients’ risk of delirium can be predicted at admission using the early prediction score, allowing the implementation of early preventive interventions aimed to reduce the incidence and severity of ICU delirium.