Performance analysis of a prognostic model for sepsis based on an intelligent decision system combined with white blood cell count,procalcitonin,and hypersensitive C-reactive protein levels

NIE Bing-rong; Chen Yu-bing; LIU Yu-rong; TANG Jin-he; HE Yan

doi:10.13820/j.cnki.gdyx.20250187

Abstract：Objective To evaluate the performance of a sepsis prognostic model established using an intelligent decision system combined with white blood cell count(WBC),procalcitonin(PCT),and hypersensitive C-reactive pro-tein(hs-CRP)levels.Methods A total of 122 patients with sepsis admitted to the Department of Emergency Medi-cine,Zhuhai People's Hospital,between July 1,2023 and June 30,2024 were included.Patients were divided into sur-vival(n=91)and death(n=31)groups based on clinical outcomes.Clinical characteristics,WBC,PCT,hs-CRP levels,and Sequential Organ Failure Assessment(SOFA)scores were compared.Prognostic and independent mortality-related factors were identified via multivariate Cox regression.An intelligent decision system was used to construct a prog-nostic prediction model.The model's performance was assessed using the receiver operating characteristic(ROC)curve.Results There were no significant differences in baseline characteristics between the survival and death groups(P＞0.05).However,the death group showed significantly higher WBC,PCT,hs-CRP levels,and SOFA scores(P＜0.05).Cox regression analysis revealed a negative correlation between WBC,PCT,hs-CRP levels,SOFA scores and patient survival time(P＜0.05).These four variables were identified as independent risk factors for sepsis mortality(P＜0.05).A final prognostic model was developed based on intersecting features from multivariate logistic regression analysis.ROC analysis yielded an AUC of 0.891(SE=0.018,95％CI:0.864-0.953),indicating strong predictive performance.Conclusion WBC,PCT,hs-CRP levels,and SOFA score are key predictors of sepsis prognosis and closely associated with patient survival.A prognostic model incorporating these factors based on an intelligent decision sys-tem demonstrates high predictive efficacy.