Cohort formation has been described in detail elsewhere.24,25 In brief, we
identified patients with a primary or secondary diagnosis of AF according
to the International Classification of Diseases, 9th/10th revision
codes (427.3, 427.31, or 427.32 / I48) with the use of the following hospital
discharge abstract databases in Quebec and Ontario: Maintenance
et Exploitation des Données pour l’Étude de la Clientèle Hospitalière
and the Canadian Institute for Health Information Discharge Abstract
Database, respectively. The primary (principal) diagnosis code is the
main condition treated or investigated during the admission. However,
up to 7 diagnosis codes may be recorded by the hospital. The remaining
diagnoses (secondary) are considered to be the subsidiary diagnoses.
For patients with >1 eligible admission with an AF diagnosis, the date of
the first admission with an AF diagnosis was considered the index date
of entry into the study cohort. We determined patients’ baseline characteristics,
outcome data, and drug prescriptions from linkage between
hospital discharge, physician claims, prescription drug claims, and vital
status databases in Quebec and Ontario (Table I in the online-only Data
Supplement). For the stroke and bleeding outcomes, we used data from
emergency department visits in addition to the information from the
hospital discharge databases. We used validated database codes (whenever
possible) to determine stroke and bleeding outcomes.26–30
We used the physician claims databases maintained by la Régie
de l’assurance maladie du Québec and the Ontario Health Insurance
Plan, which contain information on in- and outpatient diagnostic and
therapeutic procedures. We also used the la Régie de l’assurance
maladie du Québec and the Ontario Drug Benefit Plan drug claims
databases, which contain information on dispensed outpatient medications
for patients aged ≥65 years. Drug prescriptions were identified
from these databases by using drug identification numbers. These
prescription claims databases provide highly accurate information on
dispensed outpatient medications.31–33
We grouped the selected AF cohort into dialysis and nondialysis
patients according to the presence of ≥3 dialysis procedural codes
(same or different codes for hemodialysis and peritoneal dialysis)
within the 12 months preceding AF hospitalization (database codes
in Table II in the online-only Data Supplement). Our 3-code rule
attempted to select patients undergoing maintenance dialysis. For all
patients, we assessed demographic characteristics and comorbidities
at and within 1 year preceding AF hospitalization by using validated