Abstract Background and aims Patients with severely reduced kidney function have been excluded from randomized controlled trials and data on the safety and efficacy of direct oral anticoagulants (DOACs) according to kidney function remain sparse. The aim was to evaluate the safety and efficacy of the DOACs across subgroups of kidney function. Methods Using multiple Danish nationwide registers and laboratory databases, we included patients initiated on oral anticoagulants (OACs) with atrial fibrillation and available creatinine level and followed patients for 2 years to evaluate occurrence of stroke/thromboembolism (TE) and major bleeding. Results Among 26 686 included patients, 3667 (13.7%) had an estimated glomerular filtration rate (eGFR) of 30–49 mL/min/1.73 m2 and 596 (2.2%) had an eGFR below 30 mL/min/1.73 m2. We found no evidence of differences regarding the risk of stroke/TE between the OACs (P-value interaction >0.05 for all). Apixaban was associated with a lower 2-year risk of major bleeding compared to vitamin K antagonists (VKA) [hazard ratio 0.79, 95% confidence interval (CI) 0.67–0.93], and the risk difference was significantly larger among patients with reduced kidney function (P-value interaction 0.018). Rivaroxaban was associated with a higher risk of bleeding compared to apixaban (hazard ratio 1.78, 95%CI 1.32–2.39) among patients with eGFR 30–49 mL/min/1.73 m2. Conclusions Overall, we found no differences regarding the risk of stroke/TE, but apixaban was associated with a 21% lower relative risk of major bleeding compared to VKA. This risk reduction was even greater when comparing apixaban to VKA among patients with eGFR 15–30 mL/min/1.73 m2, and when comparing apixaban to dabigatran and rivaroxaban among patients with eGFR 30–49 mL/min/1.73 m2.