Ekaterina I. Glushko

Objective . To substantiate the concept of perioperative infectious risk by analyzing current approaches to classifying complications, methods for predicting them, and the key factors influencing surgical outcomes. Materials and Methods. We conducted an analytical review based on expert judgement and a systematic search of publications from the preceding 10 years in PubMed and Lens.org. The analysis included data on the epidemiology of infectious complications, classification systems (CDC, Clavien-Dindo, Accordion), prognostic scores (ASA, SOFA, ACS-NSQIP, SORT), laboratory markers (CRP, IL-6, procalcitonin, neutrophil-tolymphocyte ratio), and machine learning models (PERISCOPE AI, Random Forest, SVM, ensemble methods). Results . The risk of infectious complications was shown to be determined by the cumulative effect of modifiable and nonmodifiable factors related to both the patient and the healthcare facility. Laboratory markers demonstrated high prognostic value: IL-6 levels above 432 pg/mL and CRP 150 mg/L on postoperative day 3 were statistically significantly associated with adverse outcomes. Advanced prognostic tools provided high accuracy: the AUROC for ACS-NSQIP reached 0.80, and for PERISCOPE AI, 0.82–0.91. The use of scoring systems and machine learning made it possible to identify high-risk patients before clinical symptoms appeared. Conclusion . Risk of perioperative infections is multifactorial and requires an integrated approach to risk stratification, incorporating clinical data, risk assessment scales, laboratory markers, and AI models. The use of structured prognostic tools enables early identification of complications, optimize prevention, and reduce mortality.

INTRODUCTION: Despite the widespread use of rapid risk stratification scales, their prognostic value for the COVID-19 (COronaVIrus Disease 2019) remains unclear. OBJECTIVE: To evaluate the prognostic performance of rapid scoring systems in predicting mortality and adverse outcomes (ICU admission, organ support, or death) in patients with COVID-19. MATERIALS AND METHODS: A literature search was conducted in PubMed, eLIBRARY.RU and ScienceGate for studies published from January 2019 to January 2025. Prospective and retrospective observational studies were included. Prognostic accuracy was assessed via meta-analysis of AUROC (Area Under the Receiver Operating Characteristic Curve). Data synthesis and quality assessment were performed using MedCalc 20.027, Microsoft Excel 2019 and the GRADE approach. RESULTS: Sixty observational studies (mostly cohort and retrospective designs) with 619,494 patients were included. Five scoring systems were analyzed for predicting in-hospital mortality: NEWS, NEWS2, REMS, qSOFA, and SIRS. REMS showed the highest prognostic value (AUC = 0.808; 95% CI: 0.776–0.839), while SIRS showed the lowest (AUC = 0.662; 95% CI: 0.596–0.728). The NEWS, NEWS2, and qSOFA scores demonstrated fair performance (AUC ranging from 0.722 to 0.782). High heterogeneity was observed across studies (I² > 96%, p < 0.1). The same five tools were evaluated for predicting severe COVID-19. NEWS and NEWS2 were the most accurate (AUC = 0.778; 95% CI: 0.707–0.849 and 0.738–0.819, respectively). REMS also showed fair performance (AUC = 0.733; 95% CI: 0.708–0.757) with minimal heterogeneity (I² = 0%, p > 0.1), whereas qSOFA and SIRS were less accurate. Heterogeneity remained high for NEWS, NEWS2, and qSOFA. CONCLUSIONS: This systematic review and meta-analysis indicate that REMS, NEWS, NEWS2, and qSOFA have acceptable prognostic value for predicting mortality and severe outcomes in COVID-19 patients. REMS was most effective for mortality prediction, while NEWS and NEWS2 were more suitable for identifying patients at risk of severe disease. These findings support the use of rapid scoring systems in clinical risk stratification.