Prediction of Mortality in ICU Patients with SARS-CoV-2-Associated Pneumonia

Aim: to determine the predictive value of selected routine clinical and laboratory parameters and to assess their prognostic significance for modeling mortality risk in intensive care unit (ICU) patients with SARS-CoV-2-associated pneumonia. Materials and Methods. A retrospective case-control analysis of 73 medical records was performed. The control group included 20 records of surviving patients, while the primary group comprised 53 records of non-survivors treated between January and February 2022. The study parameters included leukocyte differential count, C-reactive protein (CRP), ferritin, blood oxygen saturation (SpO₂) via pulse oximetry, and the neutrophil ratio (NR) defined as the percentage of band neutrophils divided by the percentage of segmented neutrophils. The prognostic value of identified predictors was assessed using receiver operating characteristic (ROC) curve analysis. The area under the curve (AUC), 95% confidence interval (CI), sensitivity (Se), specificity (Sp), and cutoff point (CP) were determined, with CP defined as the predictor value yielding the highest sum of sensitivity and specificity. Results. The most informative predictors of mortality in SARS-CoV-2-associated pneumonia were: On the day of hospital admission: Ferritin levels (AUC=0.826; 95% CI: 0.717–0.905; P <0.001, CP≤0.473 mg/L; Se=78%; Sp=75%). On ICU day 1: Granulocyte count (GRA, AUC=0.711; 95% CI: 0.589–0.814; P <0.002, CP>6×10⁹/L; Se=94%; Sp=75%), NR (AUC=0.713; 95% CI: 0.541–0.850; P <0.016, CP>18; Se=91%; Sp=62%). On the final day in ICU: CRP (AUC=0.825; 95% CI: 0.522–0.973; P <0.013, CP>14 mg/L; Se=75%; Sp=100%); NR (AUC=0.862; 95% CI: 0.724–0.947; P <0.0001, CP>16; Se=94%; Sp=82%); SpO₂ (AUC=0.909; 95% CI: 0.819–0.963; P <0.0001, CP≤91%; Se=77%; Sp=100%); White blood cell count (WBC, AUC=0.833; 95% CI: 0.725–0.912; P <0.001, CP>12.2 × 10⁹/L; Se=80%; Sp=81%). Using a stepwise elimination approach, a mathematical model was proposed for predicting mortality probability ( P ) in SARS-CoV-2-associated pneumonia. Conclusion. The most valuable prognostic model for predicting mortality risk is represented by the equation: P =1/(1+е –Z )×100% using routine laboratory parameters such as ferritin, neutrophil ratio and blood oxygen saturation. The model showed a sensitivity of 84.0% and a specificity of 94.1%.