Background: chest CT is increasingly used for COVID-19 screening in healthcare systems with limited SARS-CoV-2 PCR capacity. Its diagnostic value was supported by studies with methodological concerns and its use is controversial. Here we investigated its potential to diagnose COVID-19 in symptomatic patients and to screen asymptomatic patients in a prospective study with minimal selection bias. Methods: From March 19, 2020 to April 20, 2020 we performed parallel SARS-CoV-2 PCR and CT with categorization of COVID-19 suspicion by CO-RADS, in 859 patients with COVID-19 symptoms and 1138 controls admitted to the hospital for COVID-19 unrelated medical urgencies. CT-CORADS was categorized on a 5-point scale from 1 (very low suspicion) to 5 (very high suspicion). AUC under ROC curve were calculated in symptomatic versus asymptomatic patients to predict positive SARS-CoV-2 positive PCR and likelihood ratios for each CO-RADS score were used for rational selection of diagnostic thresholds. Findings: CT-CORADS had significant (P<0.0001) diagnostic power in both symptomatic (AUC=0.891) and asymptomatic (AUC=0.700) patients hospitalized during SARS-CoV-2 peak prevalence. In symptomatic patients (41.7% PCR+), CO-RADS ≥ 3 detected positive PCR with high sensitivity (89.1%) and 72.5% specificity. In asymptomatic patients (5.3% PCR+), a CO-RADS score ≥ 3 detected SARS-CoV-2 infection with low sensitivity (45.0%) but high specificity (88.8%). Interpretation: CT-CORADS has meaningful diagnostic power in symptomatic patients, supporting its application for time-sensitive triage. Sensitivity in asymptomatic patients is insufficient to justify its use as screening approach. Incidental detection of CO-RADS ≥ 3 in asymptomatic patients should trigger reflex testing for respiratory pathogens.