ObjectiveTo analyze the main risk factors of sudden delirium in ICU patients, construct logistic regression model and decision-tree model through a prospective verification to establish the best prevention mechanism for early identification of delirium.

MethodsA total of 198 ICU patients from January 2018 to January 2020 were retrospectively summarized as the model group.According to the diagnostic criteria of delirium within 24 hours, they were divided into delirium group (n=88) and non-delirium group (n=110).The risk factors were screened by multivariate logistic regression analysis.In addition, 87 patients from February 2020 to February 2021 were choosed as the validation group.Receiver operating characteristic curve was used to compare the predictive efficacy of the two models.

ResultsCompared with the patients without delirium, the age, acute physiology and chronic health (APACHE Ⅱ) score and complications of the patients with delirium were increased, sedation time and mechanical ventilation time were prolonged, serum neuro-specific enolase(NSE) and arterial blood lactic acid were increased, and oxygenation index was decreased (P < 0.01).Logistic regression analysis showed that APACHE Ⅱ score, serum NSE and arterial blood lactate level were independent risk factors for delirium (P < 0.01).The area under the curve of the decision-tree model was higher than that of logistic regression model (P < 0.01).

ConclusionsThe incidence of sudden delirium within 24 hours in ICU patients is high, and multiple risk factors may be involved in the occurrence of delirium, including APACHE Ⅱ score, serum NSE and arterial blood lactic acid level.The decision-tree model may have a higher predictive efficacy than traditional logistic regression model, which provides a better evaluation method for guiding clinical medical staff to correctly identify high-risk groups of delirium in early stage.