Integrating serological and drainage fluid indicators: developing two predictive models for early detection of postoperative intraabdominal infections in gastrointestinal tumor patients

Junfeng Zhou ¹ Lin Lin ² Cankun He ³ Ziyi Wang ⁴ Yuping Zhan ^5,6 Sida Sun ^1* Qingliang He ^1,7*

• ¹ Department of Gastrointestinal Surgery, The First Affiliated Hospital of Fujian Medical University, Fuzhou, Fujian Province, China
• ² Nursing Department, The First Affiliated Hospital of Fujian Medical University, Fuzhou, China
• ³ Department of General Surgery, Huian County Hospital, Hui'an, China
• ⁴ Emergency Department, Zigong First People’s Hospital, Zigong, China
• ⁵ Department of Infectious diseases, The First Affiliated Hospital of Fujian Medical University, Fuzhou, China
• ⁶ Department of Infectious diseases, National Regional Medical Center, Binhai Campus of the First Affiliated Hospital, Fujian Medical University, Fuzhou, China
• ⁷ Department of gastrointestinal surgery, National Regional Medical Center, Binhai Campus of the First Affiliated Hospital, Fujian Medical University, Fuzhou, China

Purpose: This study aimed to investigate the influencing factors of postoperative intra-abdominal infection (PIAI) in gastrointestinal cancer patients by combining biomarkers in serum and drainage fluid (DF). It also intended to construct the predictive models and explore their predictive value for PIAI, offering clinical guidance. Methods: 383 patients from Institution A formed the development cohort, and 77 patients from Institution B formed the validation cohort. Independent predictors of PIAI were identified using LASSO and logistic regression analysis based on biomarkers in serum and DF, and the corresponding nomograms were constructed. The nomograms were evaluated for their performance using the calibration curve, area under the curve (AUC), decision curve analysis (DCA), and clinical impact curve (CIC). Results: The prevalence of PIAI was 15.9% in the development cohort and 24.7% in the validation cohort. There were 5 indicators included in the nomogram on postoperative day (POD) 1, and 4 indicators on POD 3, including DF lactate dehydrogenase and C-reactive protein. The AUC values of the models in the development and validation cohorts were 0.731 and 0.958 on POD 1, and 0.834 and 0.951 on POD 3, respectively. The calibration curve, DCA, and CIC demonstrated the favorable clinical applicability of the models. Conclusions: Two nomogram models including serum and DF biomarkers on POD 1 and POD 3 were developed and validated. These models can identify patients at risk of PIAI and have promise for clinical application.