Most critically ill patients experience malnutrition, resulting in a poor prognosis. This study aimed to evaluate the association of prealbumin (PAB) with the prognosis for severely and critically ill coronavirus disease 2019 (COVID-19) patients and explore factors related to this association. Patients with laboratory-confirmed COVID-19 from West Campus of Union Hospital in Wuhan from January 29, 2020 to March 31, 2020 were enrolled in this study. Patients were classified into the PAB1 (150-400 mg/L; N = 183) and PAB2 (< 150 mg/L; N = 225) groups. Data collection was performed using the hospital's electronic medical records system. The predictive value of PAB was evaluated by measuring the area under the receiver-operating characteristic (AUROC) curve. Patients were defined as severely or critically ill based on the Guidance for COVID-19 (7th edition) by the National Health Commission of China. During this analysis, 316 patients had severe cases and 65 had critical cases. A reduced PAB level was associated with a higher risk of mortality and a longer hospital stay. The AUROC curve for the prognosis based on the PAB level was 0.93, with sensitivity of 97.2% and specificity of 77.6%. For severe cases, a lower level of PAB was associated with a higher risk of malnutrition, higher NK cell counts, and lower B lymphocyte counts; these factors were not significant in critical cases. C-reactive protein and nutritional status mediated the association between PAB and prognosis. This retrospective analysis suggests that the PAB level on admission is an indicator of the prognosis for COVID-19.