(P149) VALIDATING ICD-10 CODES FOR IDENTIFYING OUT-OF-HOSPITAL CARDIAC ARRESTS IN AN ADMINISTRATIVE DATABASE

Background: Out-of-hospital cardiac arrest (OHCA) affects roughly 60,000 Canadians yearly. One method of retrospectively identifying OHCAs among those surviving to hospital admittance is searching administrative datasets for hospital records containing ICD-10 (International Classification of Disease, 10th revision) cardiac arrest codes. However, the sensitivity of these codes for identifying OHCA cases admitted to hospital is not well established, and relies upon cases being classified as OHCAs in hospital record databases. We aim to estimate the sensitivity of searching ICD-10 cardiac arrest codes in identifying cases of OHCA in British Columbia (BC).

METHODS AND RESULTS: We used EMS-treated OHCA cases from the British Columbia Cardiac Arrest Registry as the gold standard. This registry prospectively identifies and manually abstracts data on all OHCA cases treated by EMS, within four metropolitan areas in BC (including 3.3 million people, approximately 75% of BC’s population). Cases are identified using triage codes, defibrillator downloads, vital sign data, paramedic impression codes, prehospital disposition data, and a mandatory question on all paramedic clinical records asking if the patient experienced a cardiac arrest. We included presumed cardiac OHCAs among individuals aged 1 to 85, occurring between January 1, 2012 and December 31, 2016. EMS records were linked to their corresponding in-hospital records in the Discharge Abstract Database (DAD). DAD records were searched for select codes relating to cardiac arrest occurring before hospital admission including I46.0 (cardiac arrest with successful resuscitation), I46.9 (cardiac arrest, unspecified), and I49.0 (ventricular fibrillation). To estimate sensitivity, we calculated the proportion of EMS-identified OHCAs containing one or more of these codes. Among 6,009 OHCAs identified, 2,032 (33.8%) patients survived to hospital admission (Figure 1). Cardiac arrest-specific ICD-10 codes were recorded for 1,357/2,032 of the OHCAs that survived to hospital admission (sensitivity 66.8%, 95% CI 64.7, 68.8). Of these patients, the most common ICD-10 codes used were I46.0 (888/2,032; 43.7%), I49.0 (375/2,032; 18.5%), and I46.9 (344/2,032; 16.9%) (Table 1). Among the records not identified by the cardiac arrest ICD-10 codes (675/2,032; 33.2%), the most common codes were I25.1 (atherosclerotic heart disease; 29.6%), G93.1 (anoxic brain damage; 23.3%), and I210 (acute transmural myocardial infarction of anterior wall; 12.1%).

Conclusion: Almost a third of patients suffering OHCA who were admitted to hospital were not coded as having cardiac arrest in the DAD. Most often, these cases were coded as having primarily myocardial ischemic or cerebral events. Research in cardiac arrest relying on administrative databases can result in underestimation of cardiac arrest incidence.