Integrated Clinical, Respiratory and Biomarker Prediction of Mortality in Patients with Severe COVID-19 Pneumonia Treated with Convalescent Plasma: Development and Internal Validation of Prognostic Models
Gianluca Gessoni *
Division of Transfusion Medicine, Ospedale dell’Angelo, Mestre, Venice, Italy.
Sara Valverde
Division of Laboratory Medicine, Ospedale Madonna della Navicella, Chioggia, Venice, Italy.
Francesca Gessoni
Division of Radiation Oncology, Centro di Riferimento Oncologico di Aviano (CRO), Aviano, Italy.
Giovanni Roveroni
Division of Transfusion Medicine, Ospedale di Mirano, Mirano, Venice, Italy.
Lauretta Moro
Division of Transfusion Medicine, Ospedale di Mirano, Mirano, Venice, Italy.
Carla Rossi
Division of Transfusion Medicine, Ospedale SS Giovanni e Paolo, Venice, Italy.
Giulia DE Fusco
Division of Transfusion Medicine, Ospedale dell’Angelo, Mestre, Venice, Italy.
Enza Coluccia
Division of Transfusion Medicine, Ospedale dell’Angelo, Mestre, Venice, Italy.
Gianni Tidore
Division of Transfusion Medicine, Ospedale dell’Angelo, Mestre, Venice, Italy.
Luca Collodel
Division of Transfusion Medicine, Ospedale dell’Angelo, Mestre, Venice, Italy.
Rosaria Piatti
Division of Transfusion Medicine, Ospedale dell’Angelo, Mestre, Venice, Italy.
Federica Mazzetto
Division of Transfusion Medicine, Ospedale dell’Angelo, Mestre, Venice, Italy.
Gabriella Mazzaro
Division of Transfusion Medicine, Ospedale Madonna Della Navicella, Chioggia, Venice, Italy.
Michela Pivetta
Division of Transfusion Medicine, Ospedale dell’Angelo, Mestre, Venice, Italy.
Cecilia Callari
Division of Transfusion Medicine, Ospedale di Mirano, Mirano, Venice, Italy.
Silvia DA Ros
Division of Transfusion Medicine, Ospedale dell’Angelo, Mestre, Venice, Italy.
*Author to whom correspondence should be addressed.
Abstract
Background: Mortality risk among patients with severe COVID-19 is determined by interacting clinical, respiratory and systemic factors. This study evaluated whether routinely available respiratory and biomarker information could provide incremental prognostic information in a cohort of patients treated with convalescent plasma (CCP).
Methods: A secondary prognostic modelling analysis was performed using data from 376 consecutive patients with severe COVID-19 pneumonia treated with CCP at ULSS 3 Serenissima, Italy. The primary outcome was 30-day all-cause mortality. Sequential multivariable models were developed using Firth penalised logistic regression. Two parallel severity pathways were evaluated: Berlin severity (Model 5A) and ventilatory support (Model 5B). Both pathways incorporated tachypnoea >30/min, PaO₂/FiO₂ <200, raised Cystatin C, BNP and hs-cTnI. Internal validation used 2,000 bootstrap resamples with optimism correction; calibration and decision-curve analysis were also assessed.
Results: There were 97 deaths (25.8%). Tachypnoea and PaO₂/FiO₂ <200 provided strong incremental prognostic information. Cystatin C was the first biomarker retained, followed by BNP and hs-cTnI; galectin-3 did not add significant information after these biomarkers. Apparent AUCs were 0.9560 for Model 5A and 0.9656 for Model 5B; optimism-corrected AUCs were 0.9315 and 0.9457, respectively. Corrected Brier scores were 0.0838 and 0.0735. The correlated ROC comparison did not demonstrate a statistically significant difference (ΔAUC=0.0096; p=0.102).
Conclusions: In this single-centre CCP-treated cohort, integrated models combining baseline clinical characteristics, respiratory severity and selected renal and cardiovascular biomarkers showed strong discrimination after internal optimism correction. The models represent candidate prognostic frameworks rather than validated bedside tools and require external validation in independent and contemporary populations.
Keywords: COVID-19, SARS-CoV-2, convalescent plasma, mortality, prognostic model, Cystatin C, BNP, cardiac troponin, firth regression, decision-curve analysis