Abstract 30: Association Between Hospital-ascertained Atrial Fibrillation And Central Retinal Artery Occlusion: A Study Of 12 Million Patients
Lusk, JB; Song, A; Unnithan, S; Al-Khalidi, HR; Piccini, JP; Xian, Y; O'Brien, EC; Mac Grory, BC
Published in: Stroke
Atrial fibrillation (AF) is a major risk factor for cerebral ischemic stroke. However, it is not known whether AF predicts the development of central retinal artery occlusion (CRAO), a form of ischemic stroke affecting the retina.
A retrospective cohort study was undertaken using data from the California Healthcare Cost and Utilization Project (HCUP) State Inpatient and State Emergency Department Datasets (SID/SEDDs). Patients 18 years and older discharged from non-federal hospitals between 2005 and 2011 who did not have a history of CRAO were analyzed. The exposure variable was AF and the endpoint was CRAO, each identified using validated ICD-9-CM diagnosis codes. Association between AF and CRAO was modeled using a Fine-Gray method with death as a competing risk with adjustment for age, biological sex, race, and vascular co-morbidities.
A total of 12,181,778 patients were included, among these 806,397 with AF and 11,375,381 without AF. In total, 309 patients had CRAO. In an unadjusted analysis, there was a higher risk of CRAO in patients with versus without AF (HR 2.24 (95% CI: 1.51 to 3.32)). After adjustment for pre-specified covariates, there appeared to be a lower hazard of CRAO in patients with AF (aHR 0.61 (95% CI: 0.45 to 0.98)). Further analyses including cerebral ischemic stroke (aHR 1.16 (95% CI: 1.14 to 1.18)) and specifically embolic stroke (aHR 4.29 (95% CI 4.10-4.48)) as positive controls argued against overadjustment bias. We present sensitivity analyses including CRAO identified in any position of the discharge ICD list, using different ascertainment windows for AF and using broader categories of retinal ischemia.
The incidence of CRAO was higher in patients with AF than those without AF, but the hazard of CRAO was not higher for patients with AF after adjustment for measured covariates. Endpoint and exposure ascertainment may have been limited by inclusion only of inpatient and emergency department encounters.
