Construction of a risk prediction model for sepsis-associated delirium in the intensive unit patients: A retrospective cohort study

Abstract Background: To identify the risk factors associated with sepsis-associated delirium and to develop a nomogram for predicting the risk of delirium in patients with sepsis on the basis of the Medical Information Mart for Intensive Care IV database. Methods: A total of 12029 patients from the Medical Information Mart for Intensive Care-IV database (MIMIC-IV 2.0) were divided into a training set and a testing set at a ratio of 7:3. The primary outcome was the occurrence of sepsis-associated delirium (SAD) in ICU patients with sepsis. Hierarchical logistic regression analyses were conducted in the training set, sequentially analyzing basic information and clinical features (Model 1) followed by additional nursing management factors (Model 2), to identify factors associated with the risk of SAD. These identified factors were utilized to construct a nomogram for predicting the risk of SAD. The predictive performance of the nomogram was evaluated via receiver operating characteristic curves and calibration curves. Results: A total of 12,029 patients were enrolled, of whom 1,666 patients (13.3%) died in the hospital and 4,807 patients (39.96%) had SAD. In-hospital mortality in the SAD group was greater than that in the non-SAD group (23.4% vs. 7.5%, p < 0.001). By hierarchical logistic regression, two models were compared, and a nomogram integrating basic information, clinical features and easily overlooked nursing management factors was constructed. The risk factors for sepsis-associated delirium include age ≥ 65 years, tobacco use, SOFA score, intensive care unit (ICU) type at admission, hypernatremia, hypocalcemia and diet type on the first day. CRRT during sepsis and receiving skin care decrease the risk of sepsis-associated delirium. The AUCs of the developed model were 0.832 (95% CI 0.823–0.841) in the training set and 0.832 (95% CI 0.818–0.846) in the testing set. The calibration curve showed good concordance between the predicted and observed delirium rates among sepsis patients. Conclusions: We developed a nomogram incorporating ten features of clinical work,especially adding factors related to nursing management. The constructed nomogram has good performance and clinical applicability and can help identify SAD patients in a timely manner for early intervention and improve neurological outcomes.