This study established a prediction model for the severity of COVID-19 by analyzing NKG2A ⁺ NK cells incorporating laboratory indicators in moderate and severe COVID-19 patients. By classifying patients based on disease severity, we identified nine key laboratory indicators significantly associated with disease progression: WBC, N, L, NLR, PDW, D-D, CRP, LDH, and IL-6. Through correlation matrix and random forest models, CRP, IL-6, LDH, NLR, L, and D-D were identified as key indicators for the baseline prediction model. Age-stratified analysis revealed that patients over 80 years old have higher inflammation and thrombotic risk. Further stratification by comorbidities and APACHE-II scores confirmed the significance of these six indicators in different patient subgroups. In addition, analysis showed that severe patients had higher levels of NKG2A ⁺ NK cells, correlating with decreased lymphocytes and increased NLR, D-D, LDH, CRP, and IL-6. Adding NKG2A ⁺ NK cells to the prediction model improved its AUC value. This was validated in patients with the SARS-CoV-2 Wuhan strain, where blocking NKG2A enhanced NKG2D ⁺ NK cell activity, In summary, NKG2A ⁺ NK cells incorporating laboratory indicators as a key predictive indicator highlight its potential for improving patient stratification and targeted therapeutic strategies.