Developing and evaluating pediatric phecodes for high-throughput phenotyping using electronic health records

Abstract

There is a critical need for high-throughput electronic health record (EHR) phenotyping tools designed for use in pediatric populations. Previous advances in EHR phenotyping have involved the development of phecodes that aggregate ICD-9-CM, ICD-10-CM, and ICD-10 codes to better represent clinically meaningful diseases and traits. However, pediatric patients have different diseases and outcomes than adults, and existing phecodes, which were created using population-based diagnoses primarily from adult patients, do not capture the distinctive pediatric spectrum of disease. We developed specialized pediatric phecodes (Peds-Phecodes) for better high-throughput phenotyping of pediatric patients using EHRs. We applied a hybrid data- and knowledge-driven approach leveraging EHRs and genetic data from Vanderbilt University Medical Center to modify the existing phecodes to better capture pediatric phenotypes. First, we compared the prevalence of patient diagnoses in pediatric and adult populations to identify disease phenotypes differentially affecting children and adults. We then used clinical domain knowledge to remove phecodes representing phenotypes unlikely to affect pediatric patients and create new phecodes for phenotypes relevant to the pediatric population. The final Peds-Phecodes set consisted of 2,051 phecodes, providing a mapping for 15,533 ICD-9-CM codes and 82,949 ICD-10-CM codes to phecodes. We found that Peds-Phecodes replicated more known pediatric genotype-phenotype associations than phecodes (248 versus 192 out of 687 SNPs, p<0.001), and detected 348 potentially novel genotype-phenotype associations for pediatric conditions. Our results suggest that Peds-Phecodes capture higher-quality pediatric phenotypes and deliver superior outcomes in phenome-wide association studies compared to phecodes. We expect that Peds-Phecodes will facilitate efficient, large-scale phenotypic analyses in pediatric populations.