Background: Limited data are available about the predictors and outcomes associated with prolonged SARS-CoV-2 RNA shedding (VS). Methods: A retrospective study including COVID-19 patients admitted to an Italian hospital between March 1 and July 1, 2020. Predictors of viral clearance (VC) and prolonged VS from the upper respiratory tract were assessed by Poisson regression and logistic regression analyses. The causal relation between VS and clinical outcomes was evaluated through an inverse probability weighted Cox model. Results: The study included 536 subjects. The median duration of VS from symptoms onset was 18 days. The estimated 30-day probability of VC was 70.2%. Patients with comorbidities, lymphopenia at hospital admission, or moderate/severe respiratory disease had a lower chance of VC. The development of moderate/severe respiratory failure, delayed hospital admission after symptoms onset, baseline comorbidities, or D-dimer >1000 ng/mL at admission independently predicted prolonged VS. The achievement of VC doubled the chance of clinical recovery and reduced the probability of death/mechanical ventilation. Conclusions: Respiratory disease severity, comorbidities, delayed hospital admission and inflammatory markers negatively predicted VC, which resulted to be associated with better clinical outcomes. These findings highlight the importance of prompt hospitalization of symptomatic patients, especially where signs of severity or comorbidities are present.