Derivation of a multi-biomarker model for predicting mortality in hospitalised COVID-19 patients

Abstract

Introduction: This study aimed to derive and assess the performance of a multi-biomarker model from a combination of basic laboratory biomarkers in predicting mortality of hospitalized COVID-19 patients. Methods: This was a cross-sectional study conducted in a university-affiliated hospital in Malaysia. Data of confirmed COVID-19 patients who were admitted from January 2020 to August 2021 were retrieved including their admission C-reactive protein (CRP), lactate dehydrogenase (LDH), and neutrophil-lymphocyte ratio (NLR). Patients were classified as non-survivors or survivors according to their hospital mortality status. Multi-variable logistic regression analysis was used to derive the multi-biomarker model. Results: A total of 188 confirmed COVID-19 patients were analysed, of which 46 (23%) died in the hospital. Their mean age was 52 (SD 17) years, 104 (52%) were males, 114 (57%) had severe COVID-19 pneumonia, with mean APACHE II score of 14 (SD 10). On admission, those who died had higher median levels of CRP 96.0 (IQR 39.8–182.0) vs 23.0 (IQR 0–67.0 mg/L, p < 0.001), of LDH 973.0 (IQR 706.5–1520.0) vs 515.1 (408.8–738.8 IU/L, p < 0.001), and of NLR 10.1 (IQR 5.5–23.6) vs 2.8 (IQR 1.5–5.9, p < 0.001). The multi-biomarker model had a higher area under the curve (0.866, 95% CI 807-0.925) compared to its constituent individual biomarkers. At its optimal cutoff, this model had 78.9% sensitivity and 76.5% specificity for mortality prediction. Conclusion: A multi-biomarker model of CRP, LDH, and NLR predicted in-hospital mortality with a very good performance in our hospitalised COVID-19 patients.