Abstract: Objective To develop and validate a delirium prediction model for neurosurgical ICU patients and determine its clinical value. Methods A total of 665 neurosurgical ICU patients in a tertiary grade-A Hospital in Nanjing were recruited from November 2018 to May 2019. Data of delirious patients (229 cases) were compared with those of non-delirious patients (436 cases) to identify the predictors of neurosurgical ICU delirium, and the logistic regression was used to develop the model. The discrimination of the model was measured using the area under the receiver operating characteristic curve (AUROC). Hosmer-Lemeshow goodness-of-fit test was used to evaluate the calibration ability of the model. A total of 285 patients in neurosurgical ICU from June 2019 to September 2019 were involved to validate the developed model. Results Seven independent predictors of neurosurgical ICU delirium were identified, including gender (OR=2.075), age (OR=1.047), C-reactive protein concentration (OR=3.551), use of physical restriction (OR=1.011), use of sedatives (OR=9.768), use of diuretic or dehydrating drugs (OR=3.123), and Glasgow Coma Scale (OR=0.616). Hosmer-Lemeshow goodness-of-fit test showed that the consistency between the predicted delirium occurrence probability and the actual delirium occurrence probability was good (P=0.398). The AUROC of the model was 0.919, and the Youden's index 0.705, with sensitivity and specificity of 0.847 and 0.858, respectively. For model validation,the sensitivity and specificity was 65.5% and 96.5%, respectively, and the predictive accuracy of the model 87.4%. Conclusion The developed model has good predictive power in predicting the risk of delirium in neurosurgical ICU, and provides reference for health care providers to prevent delirium at an early stage.

Key words: neurosurgical, ICU, delirium, prediction model