Abstract / Summary
Abstract Background: Postoperative intensive care unit (ICU) admission after gastrointestinal tumor surgery represents a clinically important escalation of perioperative care. Conventional risk assessment is dominated by demographic, comorbidity, and procedural information, whereas the host biological state before surgery is less directly captured. We aimed to develop and evaluate a parsimonious coagulation-inflammation-nutrition score (CINS) for postoperative ICU admission. Methods: This retrospective cohort study used the INSPIRE perioperative database. Adults aged 18-90 years who underwent non-emergency gastrointestinal tumor surgery under general anesthesia were included. CINS was constructed from three routinely available preoperative biomarkers: albumin, fibrinogen, and C-reactive protein (CRP). One point was assigned for low albumin (<=3.7 g/dL), high fibrinogen (>=397 mg/dL), and high CRP (>=7.7 mg/L), generating a 0-3 score. Multivariable logistic regression adjusted for age, sex, body mass index, American Society of Anesthesiologists physical status, procedure group, and tumor site. Performance was benchmarked against the fibrinogen-to-albumin ratio (FAR), a continuous CIN-z index, and several immune-extended composite indices using discrimination, calibration, reclassification, decision curve analysis, bootstrap resampling, and repeated cross-validation. Results: Among 4,776 eligible patients, 677 experienced postoperative ICU admission. ICU admission increased stepwise across CINS categories, from 8.7% in patients with score 0 to 26.7% in patients with score 3. Compared with score 0, adjusted odds ratios were 1.86 for score 1, 2.53 for score 2, and 3.30 for score 3; each 1-point increase in CINS was associated with 51% higher odds of ICU admission. The clinical model had an apparent AUC of 0.823. Adding FAR increased the AUC to 0.832, whereas adding CINS increased it to 0.837. Bootstrap analysis showed that CINS improved AUC compared with FAR by 0.0052, with a 95% bootstrap interval of 0.0018 to 0.0087. CIN-z achieved a similar AUC of 0.839. In a common complete-case comparison of multiple composite indices, more complex immune-extended scores did not outperform the three-component CINS/CIN-z construct. Repeated 5-fold cross-validation preserved the same ranking. Conclusions: A simple preoperative score integrating fibrinogen, CRP, and albumin was independently associated with postoperative ICU admission after gastrointestinal tumor surgery and performed better than FAR and more complex immune-extended indices. CINS may provide a practical, biologically grounded adjunct to perioperative risk stratification.