Objectives: COVID-19 remains a significant healthcare burden. Leveraging the combined power of clinical trial data and big data from the real world, this study elucidated baseline factors predictive of subsequent outcomes relating to severe COVID-19 disease (SD) and the effect of nirmatrelvir/ritonavir (Tx), a protease inhibitor, on disease progression. Methods: We retrospectively analyzed data from the Evaluation of Protease Inhibition for COVID-19 in High-Risk Patients (EPIC-HR) clinical trial (NCT04960202) to discern observational associations between baseline factors and subsequent SD outcome. Baseline factors, including demographics, clinical laboratory results, symptoms, medical history, vital signs, and electrocardiogram features, were studied using machine learning (ML) for their importance in predicting hospitalization or death through Day 28, with Tx effects analyzed statistically. Generalizability of results was evaluated using real-world data (RWD) Optum Electronic Health Records. Results: Modeling indicated Tx was the greatest predictor of whether a patient progressed to SD. The most important baseline factors associated with increased risk of SD were elevated baseline (1) viral load (VL; > ~4 log10 copies/mL), (2) hsCRP (> ~1 mg/dL), (3) ferritin (> ~280 ug/L), (4) haptoglobin (> ~210 mg/dL), and (5) increased age (> ~48 years). Tx reduced VL and abnormally high hsCRP and haptoglobin to greater extents than placebo at the measured time points. RWD validation supported findings on increased risk with elevated hsCRP and ferritin and increased age (no data were available on VL and haptoglobin). Conclusion: ML analysis identified critical baseline factors immediately before or at the beginning of COVID-19 infection predictive of progression to SD in adults that are common to a heterogeneous population. This study provides insights on multivariate signatures of COVID-19 disease progression and Tx effects, which may aid future studies and inform treatment decision making.