Abstract 1578: Comparison of Risk Stratification Schemes for Stroke in Atrial Fibrillation
Background Anticoagulation management of patients with atrial fibrillation (AF) should be tailored individually on the basis of ischemic stroke risk. The objective of this study was to compare the predictive ability of published stratification schemes for stroke risk in actual clinical practice in the UK. Fifteen schemes were identified, including CHADS2, Framingham, NICE, AFI, SPAF and ACC/AHA/ESC 2006
Method AF patients aged ≥18 years included in the General Practice Research Database (GPRD) containing computerized medical records. Each AF patient was matched to one control patient by age and sex. Each scheme was divided into low, intermediate, and high predicted risk categories except CHADS2, modified CHADS2 and Framingham which use actual scores. The c-statistic (to assess discrimination) was estimated for each scheme. Outcomes included death, stroke and transient ischemic attack (TIA), as recorded in GPRD. Hospitalization for stroke was obtained from the National Registry of Hospital Admissions in England (linked to a subset of GPRD).
Results The study cohort included 79,844 AF patients and 79,844 controls followed for an average of 4 years. All risk schemes had modest discriminating ability in AF patients, with c-statistics ranging from 0.54 to 0.63 for strokes (10,603 cases), 0.56 to 0.71 for TIAs (7091), and 0.52 to 0.65 for death (50,459) in GPRD; with values of 0.55 to 0.65 for strokes (any type) (2211) and 0.55 to 0.64 for ischemic strokes (1111) in the hospital registry. Higher c statistics were found with CHADS2, modified CHADS2, and Framingham, compared to the other schemes. C-statistics were generally better in control patients compared to AF patients. The proportion of patients assigned to individual risk categories varied widely across the schemes.
Conclusion Stroke risk is predicted modestly well in AF patients by the majority of published risk schemes. Development of a better risk stratification schema is warranted.