A research article entitled "Design and Evaluation of an Automated Pediatric Acute Lymphoblastic Leukemia Registry from Clinical Data Warehouses" has been published in BMC Medical Informatics and Decision Making (2026). BMC Medical Informatics and Decision Making is a leading international journal that publishes high-impact research on the development, implementation, and evaluation of medical informatics, clinical decision-making systems, and health information technologies. Our study presents a comprehensive framework for constructing a high-quality, scalable pediatric acute lymphoblastic leukemia (ALL) registry through automated extraction of electronic medical record (EMR)-based real-world data (RWD) from clinical data warehouses (CDWs).

The study developed PeARL (Pediatric ALL Registry using eLectronic medical record-based RWD) by integrating standardized mapping, multivariate transformations, and rule-based natural language processing (NLP) to capture clinically relevant variables from two university-affiliated tertiary-care hospitals in South Korea: Seoul National University Hospital (SNUH) and the Catholic Medical Center (CMC). A total of 1,609 pediatric patients with ALL (663 from SNUH and 946 from CMC) were included. Data quality was systematically evaluated using 228 rules across five dimensions (completeness, validity, accuracy, uniqueness, consistency).

The results showed that the automated data extraction proportion reached 89.7% at SNUH and 75.0% at CMC, with most elements processed through single-field transformations. After applying the quality management process, the overall error rate decreased from 1.858% to 0.001% at SNUH and from 0.129% to 0.001% at CMC, corresponding to an estimated reliability of 99%. Moreover, the registry demonstrated robust multicenter applicability through standardized table specifications and cross-site review, despite structural differences between institutional CDWs.

Overall, the study highlights the potential of a clinically guided, standardized framework as a reproducible and scalable solution for registry construction. By combining structured mapping, multivariate transformation, and rule-based NLP, the proposed approach offers a practical pathway for multicenter research and regulatory-grade real-world evidence generation.