Background Patients with atrial fibrillation (AF) are frequently nonadherent to oral anticoagulants (OACs) prescribed for prevention of stroke and systemic embolism (SSE). The effects of OAC adherence on clinical outcomes are uncertain because previous studies used short time periods, treated adherence as binary, and/or excluded VKA users. Our objectives were to quantify the relationship between adherence to OACs as a continuous variable and AF-related clinical outcomes, and to compare the consequences of nonadherence between OAC classes and individual OACs. Methods This retrospective, observational cohort study included incident cases of AF from population-based administrative data of 5 million British Columbians from 1996 to 2020. The exposure of interest was proportion of days covered (PDC) for incident OAC prescriptions during 90 days before a clinical event or end of follow-up. Multivariable Cox proportional hazard models were used to evaluate time to first composite outcome of SSE, transient ischemic attack (TIA), or death and SSE and several secondary outcomes. Models were also stratified by the OAC patients were receiving within 90 days of the outcome event. Results The study cohort included 34,946 patients (mean age 70.1, 45% female) with median follow-up of 6.7 years. Each 10% absolute decrease in PDC was associated with a 4.3% (95%CI 2.9-5.6) and 10.3% (95%CI 7.5-13.0) increase in hazard of SSE, TIA, or death for VKA and direct OAC (DOAC) users, respectively. For SSE, hazard increases per 10% PDC reduction were 22.5% (95%CI 19.6-25.4) and 34.2% (95%CI 28.2-40.6) for VKA and DOAC, respectively. Similar significant effects were seen for all secondary efficacy outcomes. Differences between VKA and DOAC were statistically significant for all outcomes (p<0.001 all contrast tests) except major bleeding. Conclusions Even small reductions in OAC adherence in patients with AF are associated with significant increases in risk of SSE and death, and these effects are significantly greater with DOAC nonadherence than with VKA nonadherence. These results suggest that DOAC recipients are more vulnerable than VKA recipients to increased risk of stroke and death even with small reductions in adherence. The worsening efficacy outcomes associated with decreasing adherence occur without any benefit in terms of major bleeding reduction.