Abstract
BACKGROUND: Postoperative delirium (POD) is common after head and neck tumor surgery with free flaps, compromising immobilization and flap survival. However, robust tools for early POD severity prediction and systematic studies on onset-time heterogeneity are lacking. METHODS: In a retrospective cohort of 65 POD patients, demographic, perioperative, and postoperative data were analyzed. POD was subclassified by severity (CAM-ICU-7 score) and onset time (24-hour threshold). Descriptive, univariate and multivariate analyses identified independent influencing factors, and a prediction model for POD severity was developed and evaluated. RESULTS: Severity: 39 severe and 26 non-severe, differing in mental state, clinical manifestations, onset time, duration, and olanzapine dose. Preoperative platelet-to-lymphocyte ratio (PLR) and intraoperative tropisetron dosage were independent influencing factors. The nomogram showed strong discrimination (AUC = 0.830), good calibration and clinical benefit. Onset time: 33 acute and 32 delayed, differing in POD duration and olanzapine dose. Preoperative diastolic blood pressure and PLR were independent influencing factors. CONCLUSIONS: POD after head and neck tumor surgery shows marked heterogeneity in severity and onset. Neutrophils and platelets may respectively drive neuroinflammation occurrence and progression, synergizing with blood pressure to regulate POD subtyping. Identified factors enable early multidimensional warning. The prediction model performs well with optimal threshold 0.45-0.55.