Validation of ICD-10 codes for the identification of paediatric out-of-hospital cardiac arrest patients

Aim: There is a need for large-scale epidemiological studies of paediatric out-of-hospital cardiac arrest (POHCA). To enable this, we developed and validated international classification of disease (ICD-10) search algorithms for the identification of POHCA patients from health administrative data.

Methods: We validated the algorithms with a registry of POHCA (CanRoc) as the reference standard. The reference standard included all atraumatic POHCA in Middlesex-London region for January 2012-June 2020. All algorithms included 1 day to <18-year-old patients transported to emergency department (ED) by ambulance and excluded trauma. We tested three algorithms, which were applied to the National Ambulatory Care Reporting System and Discharge Abstract Database. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), positive likelihood ratio (PLR) and negative likelihood ration (NLR) were calculated for each algorithm.

Results: During the study period, 17,688 children presented to the ED by ambulance. The reference standard included 51 POHCA patients. The algorithm using only ICD-10 code for cardiac arrest had a sensitivity of 65.5% and PPV of 90%. The algorithm with the highest sensitivity of 87.3% added sudden infant death syndrome, drowning or asphyxiation with CPR in addition to the cardiac arrest codes for inpatient and ED records. This algorithm had a specificity of 99.9%, PPV of 81.4% and NPV of ∼100.0%.

Conclusion: It is important that algorithms used for cohort identification are validated prior to use. The ICD-10 code for cardiac arrest alone misses many POHCA cases but the use of additional codes can improve the sensitivity while maintaining specificity.