Precision Learning Across Neurology – Resident Education (PLAN-R): Mapping ICD-10 Categories to ABPN Content Areas to Develop a Timely, Individualized Learning System
Joshua Adams1, Adam Dziorny2, Brian Stamm1, Mark Mai3, Miya Bernson-Leung4, Robert Thompson-Stone2, Elizabeth Troy5, Kathryn Xixis6, Bittu Majmudar-Sheth7, Rachel Gottlieb-Smith8
1University of Michigan, 2University of Rochester, 3Children's Healthcare of Atlanta; Emory University School of Medicine, 4Boston Children's Hospital, 5University of Colorado, 6University of Virginia, 7Pediatric Neurology, Riley Children's Hospital, 8University of Michigan, Division of Pediatric Neurology
Objective:
To develop and validate a crosswalk from ICD-10 categories to American Board of Psychiatry and Neurology (ABPN) child neurology content areas as the foundation of an individualized, timely resident learning system.
Background:
In current state, child neurology resident education relies on delayed and indirect feedback including in-training and certification examinations and milestone assessments. We aim to develop an automated system that identifies meaningful resident-patient interactions using electronic health record features, and maps diagnostic codes from those encounters to relevant content domains. This system will allow for timely, data-driven insights to guide resident education. To construct this system, we mapped ICD-10 categories to ABPN content domains to support reliable measurement of resident topic exposure.
Design/Methods:
We expanded terminology from the ABPN child neurology content outline and text matched to ICD-10 categories. Matches were reviewed by 3 neurologists with mismatches reconciled by consensus discussion. Following this first step, unmatched ABPN content areas were manually mapped to ICD-10 categories. At one institution, we used an existing trainee-patient matching tool to record primary ICD-10 diagnoses seen by PGY3-5 child neurology residents between July 2024 and April 2025, and then measured the percent of these primary ICD-10 diagnoses captured in our crosswalk. We iteratively mapped unmatched ICD-10 categories until achieving >90% outpatient and >70% inpatient capture.
Results:
We mapped 385 unique ICD-10 categories to ABPN content domains, with 90.0% outpatient and 73.8% inpatient coverage across 3109 encounters. ICD-10 categories that did not map to a single content area, including hemiplegia/hemiparesis and shock, were intentionally excluded, limiting further coverage. The most frequent mapped diagnoses were epilepsy (outpatient) and cerebral infarction (inpatient).
Conclusions:
We developed a crosswalk between ICD-10 categories and ABPN content areas, with robust outpatient and moderate inpatient coverage. This provides a critical foundation for an automated dashboard to deliver timely trainee feedback regarding content exposure, guiding precision education.
Disclaimer: Abstracts were not reviewed by Neurology® and do not reflect the views of Neurology® editors or staff.