In Reply Dr Kurlansky raises a concern about the positive predictive value of diagnosis codes used to identify perioperative AF in our study of the long-term risk of stroke associated with this condition. Two separate issues should be considered in this regard.
The sensitivity of 88% and specificity of 86% that we cited concerned the present-on-admission status for a given diagnosis code, not the diagnosis code itself. In other words, among patients with a documented diagnosis of AF during a surgical hospitalization, the present-on-admission indicator would be expected to be 88% sensitive and 86% specific for distinguishing cases of AF that were diagnosed prior to the hospitalization compared with new-onset cases during the hospitalization. Given a true prevalence of approximately 67% for preexisting AF in the perioperative setting,1 the positive and negative predictive values for preexisting (as opposed to new-onset) cases of AF would be expected to be approximately 80% to 90%.
Gialdini G, Bhave PD, Kamel H. Stroke Risk Following Perioperative Atrial Fibrillation—Reply. JAMA. 2014;312(22):2410-2411. doi:10.1001/jama.2014.14505