A retrospective study analyzed 151 adult patients with severe COVID-19 treated with glucocorticoids between December 2022 and August 2023 to develop a laboratory nomogram to predict clinical outcomes. Non-response was defined as in-hospital mortality, need for mechanical ventilation, or persistent organ dysfunction. Using LASSO and logistic regression, key biomarkers were identified: ferritin > 970.7 ng/ml and IL-10 > 4.79 pg/ml, which predicted glucocorticoid resistance with an AUC of around 0.78. The nomogram also included the presence of diabetes and showed good calibration (Hosmer-Lemeshow p = 0.84) and discrimination (sensitivity 71.4%, specificity 70.0%). Diabetic patients had higher inflammatory responses and poorer outcomes, exacerbated by glucocorticoid-induced hyperglycemia. This model provides a tool for risk stratification and personalized management of patients with severe COVID-19 undergoing glucocorticoid therapy and requires further validation in larger multicenter studies.