Venous thromboembolism (VTE) in patients with COVID-19 in intensive care units (ICU) is frequent, but risk factors (RF) remain unidentified. In this meta-analysis (CRD42020188764) we searched for observational studies from ICUs reporting the association between VTE and RF in Medline/Embase up to 15 April 2021. Reviewers independently extracted data in duplicate and assessed the certainty of the evidence using the GRADE approach. Analyses were conducted using the random-effects model and produced a non-adjusted odds ratio (OR). We analysed 83 RF from 21 studies (5296 patients). We found moderate-certainty evidence for an association between VTE and the D-dimer peak (OR 5.83, 95%CI 3.18–10.70), and length of hospitalization (OR 7.09, 95%CI 3.41–14.73) and intubation (OR 2.61, 95%CI 1.94–3.51). We identified low-certainty evidence for an association between VTE and CRP (OR 1.83, 95% CI 1.32–2.53), D-dimer (OR 4.58, 95% CI 2.52–8.50), troponin T (OR 8.64, 95% CI 3.25–22.97), and the requirement for inotropic drugs (OR 1.67, 95% CI 1.15–2.43). Traditional VTE RF (i.e., history of cancer, previous VTE events, obesity) were not found to be associated to VTE in COVID-19. Anticoagulation was not associated with a decreased VTE risk. VTE RF in severe COVID-19 correspond to individual illness severity, and inflammatory and coagulation parameters.