Aim: Pestilence owing to severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) virus in recent times and its ongoing resurgence poses an urgent need for the identification of a predictive marker for determining coronavirus disease 2019 (COVID-19) severity. Considering inflammation as the fundamental cause of morbidity following COVID-19 infections, our study targeted acute phase proteins (APPs), potential markers reflecting the inflammatory state of the body. Materials and methods: We retrospectively analyzed the serum levels of seven APPs, namely, α-1 antitrypsin (AAT), α-1 acid glycoprotein (AGP), ceruloplasmin (Cp) and C-reactive protein (CRP), transferrin (Tf), transthyretin (TTR), and human serum albumin (HSA) in 788 COVID-19 patients. Multivariate regression analysis and receiver operating characteristic (ROC) analysis were performed to develop a mathematical model for predicting COVID-19 severity. Results: Out of the seven APPs, levels of CRP (p < 0.0001) and HSA (p < 0.0001) were found to have significant associations with the incidence of intensive care unit (ICU) admission in COVID-19 patients, as analyzed primarily in 219 patients. A mathematical model was designed with ROC displaying sensitivity and specificity of 88.89 and 66.67% with 0.802 [95% confidence of interval (CI), 0.4408–0.6508] area under the curve (AUC). The logistic regression equation thus developed was validated in a different cohort of 569 COVID-19 patients where probabilities for ICU admission calculated on the 1st day of admission were higher (p < 0.0001) in patients actually admitted to ICU at any time point during their hospital stay. Conclusion: The model appears fit for predicting the severity of COVID-19 patients at its early stages. Utilization of this model is feasible and practical as severity is decided based on a few biochemical parameters.