Abstract: Objective: To establish a delirium prediction model for neurosurgical ICU patients and verify the sensitivity and specificity of this model. Methods: A prospective cohort study was designed to collect data from 310 patients who were admitted to Neurosurgical ICU in a tertiary hospital from January 2017 to February 2018. A risk factor prediction model was then created using multivariate logistic regression. We further collected data from 60 patients between March and June 2018 to validate the model. Results: Among the 310 patients, there were 118 cases of delirium (38.1%). The model consisted of 6 predictors, which were cognitive dysfunction on admission, having fever, hypoproteinemia, abnormal liver function, sedative use ≥ 4 times, and physical restraint during the stay in ICU. The area under the curve (AUC) of the model was M=0.80, SD=0.07, 95% CI (0.674, 0.932). The sensitivity and specificity were 0.68 and 0.83, respectively, and the critical value was 0.51. Conclusion: The study established a delirium prediction model for neurosurgical ICU patients which consisted of 6 predictors. Our model can predict the risk of delirium in neurosurgical ICU patients, thereby helping medical staff to identify patients at high risk of delirium and implement targeted interventions to reduce the incidence of delirium and its negative effects.

Key words: neurosurgical; ICU; delirium; risk factors; prediction model