Development and internal validation of a Viremia-Integrated Risk Assessment (VIRA score) for predicting mortality at diagnosis of COVID-19 pneumonia.

backgroundCURB-65 score usefulness has been assessed in COVID-19 pneumonia, but not in cohorts including immunocompromised patients. SARS-CoV-2 RNAemia is associated with mortality and unfavorable clinical outcomes in COVID-19. This study aimed to develop the VIRA score (Viremia-Integrated Risk Assessment), incorporating CURB-65 and RNAemia, for predicting 30-day COVID-19 pneumonia mortality.

methodsWe included two multicenter cohorts of COVID-19 pneumonia: 539 immunocompetent and 280 immunocompromised patients, vaccinated against SARS-CoV-2 in 191 (35.4%) and 240 (85.7%), respectively. We employed multivariable logistic regression to identify predictors of 30-day mortality, with model performance assessed through area under the receiver operating characteristic curve (AUROC), and internal validation using calibration plots and bootstrap resampling. CURB-65 was available in 422 immunocompetent and 222 immunocompromised patients, constituting the derivation cohorts for the VIRA score. Integrated discrimination improvement was calculated for the addition of RNAemia, dyspnea, and male sex to CURB-65. Logistic regression coefficients were converted to the VIRA score for clinical application.

resultsMortality rates were 5.2% in immunocompetent and 15.0% in immunocompromised patients. Although CURB-65 ≥ 2 was associated with 30-day mortality in both groups (p < 0.001 and p = 0.001), the score consistently underestimated observed mortality across risk categories, most markedly in immunocompromised patients, where predicted mortality for the high-risk category (3 points: 14.5%) contrasted sharply with the observed 50.0%. At the high-risk threshold (CURB-65 score ≥ 3), sensitivity was poor in both groups (9.1% in immunocompetent and 8.6% in immunocompromised patients), with NPV of 95.2% and 85.3%, respectively. Incorporating RNAemia improved CURB-65's discriminatory performance, raising AUROCs in immunocompetent (0.806 [95% CI, 0.701-0.910] vs. 0.771 [95% CI, 0.661-0.881]; IDI 0.023) and immunocompromised (0.745 [95% CI, 0.639-0.851] vs. 0.637 [95% CI, 0.532-0.742]; IDI 0.121) cohorts. At the equivalent high-risk cut-off (score ≥ 2), the VIRA score substantially improved sensitivity and NPV over CURB-65 alone: from 9.1% to 72.7% and from 95.2% to 98.2% in immunocompetent patients, and from 8.6% to 60.0% and from 85.3% to 91.9% in immunocompromised patients.

conclusionsRNAemia strengthens CURB-65's predictive accuracy for 30-day mortality of SARS-CoV-2 pneumonia. The VIRA score, freely accessible at virascore.com, improves risk stratification especially in higher-risk categories and in immunocompromised patients. CLINICAL TRIAL NUMBER: Not applicable.