PS model accounts for age, gender, CHA2DS2 VASc score, year of treatment initiation and the following diagnoses and treatments in baseline: non-major bleeding events, anaemia, diabetes, hypertension, heart failure, osteoporosis/hip fracture, malignant neoplasm, acute kidney injury, chronic kidney disease, asthma/COPD, dementia, aspirin, antiplatelets other than aspirin, warfarin, antimeric preparations, NSAIDs, opioids, SSRIs, antidepressants other than SSRIs, antiepileptics, antipsychotics, benzodiazepines, lipid lowering drugs, insulin, antihyperglycemics other than insulins, antihypertensives, antiarrhythmics, nitrates cardiac vasodilators, cardiac stimulants, gastrointestinal protective agents, bisphosphonates and other agents affecting bone structure, systemic corticosteroids, antineoplastics, systemic antibiotics, systemic antivirals, vaccines/immunoglobulins.

DOACs comprised apixaban, rivaroxaban, edoxaban, dabigatran.

In analysis of stroke, MI, TIA, major bleeding events, and AMS, patients were followed until occurrence of outcome, death, end of patient registration or end of study period (December 2020).

*In analysis of all-cause mortality, patients were followed until occurrence of outcome (death), end of study period (December/2020) or the later date of end of patient registration and any recorded death within 90 days of end of patient registration. Propensity score matched sample size for analysis of all-cause mortality differs from sample size in analysis of other outcomes because of differences in censoring criteria, which impact a small number of patients' eligibility for inclusion in analysis at the start of follow-up.

aHR, adjusted HR; PS, propensity score; PY, person years.