Association of preoperative monocyte level with postoperative delirium in patients receiving cardiac surgery under propofol-based anesthesia: a retrospective study.

BACKGROUND: Postoperative delirium is a common complication after cardiac surgery, and perioperative inflammation may contribute to its development. However, the association between preoperative blood indicators and postoperative delirium in patients receiving propofol-based anesthesia remains unclear. OBJECTIVE: To identify preoperative blood indicators associated with postoperative delirium risk in patients undergoing cardiac surgery under propofol-based anesthesia. METHODS: Related data of 694 patients were from MIMIC-IV database. Importance ranking using 2 machine learning algorithms was performed, followed by LASSO regression, multivariable logistic regression, and trend regression analyses to identify the important factors associated with postoperative delirium risk. Receiver operating characteristic curve and decision curve analysis were used to assess their clinical value for predicting delirium. RESULTS: Among all patients, 44 patients experienced postoperative delirium (identified via ICD diagnostic coding during hospitalization). Machine learning algorithms initially implied the importance of monocyte in postoperative delirium risk as monocyte ranked first both in 2 algorithms. LASSO regression identified 7 variables, and 4 variables (red cell distribution width, red blood cell, monocyte, sepsis) showed independent association with delirium risk (all P < 0.05). Further, we found that the risk of delirium was increased with the increase of monocyte level (P for trend = 0.042), particularly when monocyte level exceeded 5.4% (P = 0.037). These results suggested monocyte as a key indicator independently associated with delirium risk and showing the most favorable discriminative performance among the evaluated markers, although its predictive utility requires prospective validation. CONCLUSIONS: Preoperative monocyte level was independently associated with coded postoperative delirium in this cohort, but prospective validation is still required before any predictive application.