Background Long-term shedding of viral RNA in COVID-19 prevents timely discharge from the hospital or de-escalation of infection prevention and control practices. Key questions are the duration and determinants of infectious virus shedding. We assessed these questions using virus cultures of respiratory tract samples from hospitalized COVID-19 patients as a proxy for infectious virus shedding. Methods Clinical and virological data were obtained from 129 hospitalized COVID-19 patients (89 intensive care, 40 medium care). Generalized estimating equations were used to identify if viral RNA load, detection of viral subgenomic RNA, serum neutralizing antibody response, duration of symptoms, or immunocompromised status were predictive for a positive virus culture. Findings Infectious virus shedding was detected in 23 of the 129 patients (17,8%). The median duration of shedding was 8 days post onset of symptoms (IQR 5 – 11) and the probability of detecting infectious virus dropped below 5% after 15,2 days post onset of symptoms (95% confidence interval (CI) 13,4 – 17,2). Multivariate analyses identified viral loads above 7 log 10 RNA copies/mL (odds ratio [OR]; CI 14,7 (3,57-58,1; p<0,001) as independently associated with isolation of infectious SARS-CoV-2 from the respiratory tract. A serum neutralizing antibody titre of at least 1:20 (OR of 0,01 (CI 0,003-0,08; p<0,001) was independently associated with non-infectious SARS-CoV-2. Interpretation Infection prevention and control guidelines should take into account that patients with severe or critical COVID-19 may shed infectious virus for longer periods of time compared to what has been reported for in patients with mild COVID-19. Infectious virus shedding drops to undetectable levels below a viral RNA load threshold and once serum neutralizing antibodies are present, which warrants the use of quantitative viral RNA load assays and serological assays in test-based strategies to discontinue or de-escalate infection prevention and control precautions. Research in context Evidence before this study We searched PubMed, bioRxiv, and medRxiv for articles that reported on shedding of infectious virus in COVID-19 patients using the search terms (“coronavirus” OR “SARS” OR “SARS-CoV-2” OR “COVID-19”) AND (“shedding” OR “infectivity” OR “infectious” OR “virus culture”) with no language or time restrictions. A detailed study on nine patients with mild COVID-19 reported that infectious virus could not be isolated after more than eight days of symptoms. The probability of isolating infectious virus was less than 5% when viral loads dropped below 6,51 Log 10 RNA copies/mL. Similar results were obtained with a larger diagnostic sample set, but that study did not report on clinical parameters such as disease severity. Finally there is a report of a single patient shedding infectious virus up to 18 days after onset of symptoms. No published works were found on the shedding of infectious virus in patients with severe or critical COVID-19, and no published works were found on factors independently associated with shedding of infectious virus. Added value of this study We assessed the duration and determinants of infectious virus shedding in 129 patients with severe or critical COVID-19. The duration of infectious virus shedding ranged from 0 to 20 days post onset of symptoms (median 8 days, IQR 5 – 11). The probability of detecting infectious virus dropped below 5% after 15,2 days post onset of symptoms (95% confidence interval (CI) 13,4 – 17,2). Viral loads above 7 log 10 RNA copies/mL were independently associated with detection of infectious SARS-CoV-2 from the respiratory tract (odds ratio [OR]; CI 14,7 (3,57-58,1; p<0,001). A serum neutralizing antibody titre of at least 1:20 (OR of 0,01 (CI 0,003-0,08; p<0,001) was independently associated with non-infectious SARS-CoV-2. Implications of all the available evidence Infection prevention and control guidelines should take into account that patients with severe or critical COVID-19 may shed infectious virus for longer periods of time compared to what has been reported for in patients with mild COVID-19. Quantitative viral RNA load assays and serological assays should be used for test-based strategies to discontinue or de-escalate infection prevention and control precautions.