Internal Audit of the Canadian Neonatal Network Data Collection System

Background Neonatal databases worldwide have become a prominent tool for benchmarking, evaluation of outcomes, and quality improvement initiatives. We aimed to assess the precision of the Canadian Neonatal Network (CNN) database by conducting an internal audit of data extraction. Methods An audit was conducted in all 31 neonatal units participating in the CNN. Ninety-five data items selected for reabstraction were classified into categories (critical, important, less important) based on predefined agreement rates. Five records were randomly selected at each site for reabstraction, including one short (3-7 days), two medium (8-12 days), and two long (18-22 days) stay cases. Agreement rates for each data item were calculated for individual units and across the network. Results A total of 155 cases and 14,725 data fields were reabstracted. The overall agreement rates for critical, important, and less important data items were 98.0, 96.1, and 96.3%, respectively. Individual site variation for discrepancies ranged between 0.2 and 12.8% for all collected data items. Conclusion Neonatal data extraction within the CNN database structure exhibited high precision; thereby, revealing the reliability of our data abstraction for neonatal demographic, processes of care, and outcomes information. An independent external audit of data extraction would be beneficial.