Algorithm for predicting sepsis in newborns with respiratory pathology and perinatal lesions of the central nervous system on mechanical ventilation

BACKGROUND: Predicting sepsis in ventilated neonates remains a challenge in neonatology.

AIM: To increase the efficiency of predicting sepsis diagnosis in newborns by developing a decision rule for its development based on decision trees.

MATERIALS AND METHODS: This clinical study retrospectively reviewed 200 full-term newborns with respiratory pathology that are admitted to the intensive care unit and are on mechanical ventilation without clinical signs of bacterial infection.

Upon admission to the department on days 1, 3–5, and 20, an enzyme-linked immunosorbent assay determined the plasma concentration of interleukin (IL)-1β, IL-6, IL-8, tumor necrosis factor-α, granulocyte colony-stimulating factor, soluble Fas ligand, fibroblast growth factors, and nitric oxide (NO), and immunophenotyping method determined CD3⁺CD19^–, CD3^–CD19⁺, CD3⁺CD4⁺, CD3⁺CD8⁺, CD69⁺, CD71⁺, CD95⁺, HLA^–DR⁺, CD34⁺, CD14⁺, CD3^–CD56⁺; lymphocytes expressing AnnexinV-FITC⁺PI^–, and AnnexinV-FITC⁺PI⁺. The possibility of diagnosing sepsis upon intensive care unit admission was assessed by statistical cluster analysis of the total studied immunological criteria. The method of decision trees in the statistical environment R formed a diagnostic rule for predicting sepsis.

RESULTS: Visualization of the cluster analysis results of admitted patients did not exclude the presence of two clusters among them (with and without sepsis, which explain the 60.81% of the point variability).

Sepsis prediction rule are as follows: disease progression occurs if on day 1 CD95 is ≥16.8% and NO is ≤9.6 mkmol/l or CD95 is ≤16.8%, CD34 is ≤0.2%, CD69 is ≥4.12% or CD95 is ≤16.8%, CD34 is ≤0.2%, CD69 is ≤4.12%, and lymphocytes expressing AnnexinV-FITC+PI– is ≥12.3%. The diagnostic accuracy was 96.00%; sensitivity was 97.00%; specificity was 94.90%; the false-positive proportion of diagnoses was 5.10%; the false-negative proportion of diagnoses was 2.94%; the positive result accuracy was 95.19%; and the negative result was 96.88%. The disease was complicated by bacterial sepsis development on 4–5 days of observation in 45 newborns.

CONCLUSIONS: Significant importance in sepsis development belongs to the prevalence of altered immunocompetent cells over proliferation and endogenous synthesis of nitric oxide. The cumulative determination of CD95⁺, CD69⁺, AnnexinV-FITC⁺PI^–, CD34⁺, and plasma nitric oxide concentration helped diagnose sepsis development at the preclinical stage. The obtained results indirectly confirm the relevance of studies on sepsis prevention and treatment by drug correction of apoptosis and inhaled NO.