Background: Anticoagulant-based prophylaxis in atrial fibrillation patients at high-risk for stroke remains underutilized in clinical practice despite extensive literature supporting its benefits. The recent 2014 AHA/ACC/HRS Atrial Fibrillation Guidelines emphasize risk stratification to target anticoagulation to those who would derive the most benefit. Rapid guideline implementation could be achieved through patient registries and decision support that provide actionable information at the point of care. Our objective is to risk-stratify primary care patients with atrial fibrillation at Denver Health (DH), an integrated safety net health system, and to characterize patterns of anticoagulation use and non-use as the first phase of a quality improvement initiative that includes development of a disease registry. Methods: We identified patients diagnosed with atrial fibrillation who made one or more visits over a two-year period to a primary care site at DH. We utilized the CHA2DS2VASc score to stratify patients as low-risk (score = 0 or 1 if female gender was the sole risk factor), intermediate-risk (score = 1), or high-risk (score = 2+) of stroke. We examined the use of anticoagulant therapy within each stratum. For patients at high-risk, we compared characteristics between those who were on any anticoagulant and those who were not. We examined differences in age, gender, payer, HAS-BLED score, and modifiable factors that contribute to the HAS-BLED score. HAS-BLED score was calculated using ICD9 codes, vital signs, and laboratory values drawn from electronic medical record data. Results: The study population of 867 patients had a median age of 64 years (interquartile range: 56-73) and 373 (43%) were women. Of the 867 patients, 168 (19.4%) were uninsured, 227 (26.2%) were covered by Medicaid, and 425 (49%) by Medicare. CHA2DS2VASc score-based stratification indicated 59 (6.8%), 71 (8.2%), and 737 (85%) patients were at low, intermediate, and high-risk, respectively. Of the low-risk patients, 13 (22%) were on anticoagulants, despite guidelines recommending no therapy for these patients. Anticoagulation rates among the intermediate and high-risk strata were 60.6% (43 of 71) and 64.5% (475 of 737), respectively. Among high-risk patients, those not on anticoagulation were more likely to have uncontrolled hypertension, abnormal liver function, or were on medications that predispose to bleeding. There were no significant differences in the proportion of patients with a HAS-BLED score of 3+ among those high-risk patients receiving anticoagulation and those not. Conclusions: A baseline assessment of stroke prophylaxis among atrial fibrillation patients in a safety net health system demonstrates nonguideline-concordant anticoagulation use among low-risk patients and suboptimal anticoagulation use among high-risk patients, patterns that could not be explained by HAS-BLED score.