A prediction for sepsis in adult patients with severe cerebro-vascular disease from neurological intensive care unit

Haiyang Sun^1*Shuyun Sun²Wanxin Wen³Yan Huang³Jingbo Sun³Chuanchuan Yu⁴Lixin Wang³Xiao Cheng³

• ¹Reproductive Center of Integrated Medicine, Affiliated Hospital of Shandong University of Traditional Chinese Medicine, Jinan, China
• ²The Affiliated Hospital of Qingdao Binhai University, Qingdao, China
• ³The Second Affiliated Hospital of Guangzhou University of Chinese Medicine, Guangzhou, China
• ⁴Sun Yat-Sen University School of Public Health, Guangzhou, China

Background: Sepsis is one of the common causes of death in the neurological intensive care unit (NICU) stroke patients, the aim of this study was to evaluate the diagnostic performance of blood biomarkers studied for the early diagnosis of sepsis in ICU hospitalized patients with acute moderate to severe stroke, and to establish a that is specifically used to predict the occurrence of sepsis or not after stroke; Methods: A prediction model was built including 157 patients with severe cerebrovascular disease (including acute ischemic stroke (AIS) or cerebral hemorrhage (ICH)) who had National Institute of Health stroke scale(NIHSS)>14 or Glasgow coma scale(GCS)<8 from January 2020 to November 2022 in NICU. Laboratory parameters and clinical characteristics of the patients were collected as well as Enzyme-Linked Immunosorbent Assay (ELISA) to detect blood biomarkers IL-10, MIP-1β, TNF-α, nNOS, iNOS, MMP-9, S-100β, and ET-1 within 48 h after symptom onset. Multi-factorial logistic regression was used to construct for predicting sepsis in patients with acute moderate-to-severe stroke, and internal validation was evaluated using bootstrap validation. The performance of the graph was assessed based on its calibration, discrimination, and clinical utility; Results: The prevalence of sepsis in acute moderate-to-severe stroke patients was 12.1%. The GCS scores of patients with comorbidity sepsis were all lower than those of patients without sepsis, and the NIHSS scores were higher than those of patients without sepsis. Logistic stepwise regression was performed to identify 4 variables Hyperlipidaemia(P<0.001), IL-10 (P<0.001), NIHSS (P=0.015), and Blood creatinine (P<0.001)), and to establish a prediction model for sepsis in acute moderate-to-severe stroke patients. The area under the curve (AUC) of the prediction model was 0.816 (95% CI: 0.721 ~ 0.911), and the calibration curve was well fitted, which has good clinical application value.