Introduction: Atrial fibrillation (AF) is a major cause of ischemic stroke. Individuals with undiagnosed AF lack the stroke risk reduction afforded by oral anticoagulant therapy. In 1983 Wolf documented that 24% (95% CI 14-37) of AF-related strokes had AF first diagnosed at the time of stroke. Given increased medical and lay attention to AF-stroke, we sought to determine whether this percentage had decreased in contemporary care.
Hypothesis: Less than 20% of patients with AF-related stroke have their AF first diagnosed at the time of stroke.
Methods: We identified patients admitted to Massachusetts General Hospital from 01-01-2010 to 12-31-2013 with a new ischemic stroke and either previously or newly diagnosed AF by searching comprehensive hospital databases for stroke and AF ICD-9 codes in conjunction with a hospital stroke registry. Physician reviewers screened 1037 potentially eligible patients, categorized AF as previously known or newly diagnosed, and performed a structured chart review of the stroke event. To confirm the diagnosis of AF was new, we conducted automated searches for AF terms in the patients’ electronic medical records (EMRs) prior to the stroke admission.
Results: We validated 856 cases (83%) as AF and ischemic stroke. AF was considered newly diagnosed in 156/856 (18%; 95%CI: 16-21). In the newly diagnosed group, no patient was on oral anticoagulants and the strokes were consequential (median NIHSS=12; 60% with mRankin of ≥3 at discharge, including 15% deaths). Pre-stroke CHA2DS2-VASc score was ≥2 in 89%. About half (76/156) had a prior medical encounter in the EMR. Evidence of pre-stroke AF was found in 8/76 records, often peri-procedural, but the AF diagnosis was not carried forward.
Conclusions: In this large, contemporary cohort, nearly one in five AF-related strokes occurred in patients who did not carry a pre-stroke AF diagnosis, similar to Wolf’s 1983 finding. The vast majority would have been at high enough pre-stroke risk to merit anticoagulation. Our findings support screening for AF in patients before they have strokes. Further, patients with past transient AF identified via automated EMR searches might merit more intensive screening.