Ziming Gan, Wen Zhu, Weijing Tang, Sara Morini Sweet, Michele Morris, Yunqing Han, Chenyi Chen, Junwei Lu, Emily Song, Mohammed Moro, Shyam Visweswaran, Tianrun Cai, Tanuja Chitnis, Tianxi Cai, Zongqi Xia
Abstract
The multiple sclerosis (MS) therapeutic landscape has evolved over time. We conducted a knowledge graph-guided analysis of MS-specific disease-modifying therapy (DMT) prescription trends using longitudinal real-world clinical data. We analyzed registry-linked electronic health record (EHR) data from two large independent healthcare systems between 2004 and 2022, including both academic and community practices.
Introduction
Multiple Sclerosis (MS) is a chronic autoimmune disease characterized by inflammatory demyelination and progressive neurodegeneration in the central nervous system, leading to neurological impairments [1]. Affecting millions of people worldwide, MS diminishes individual quality of life and creates disproportionately high societal healthcare burdens [2], partly driven by the cost of disease-modifying therapies (DMTs). Given the wide range of DMT options and variability in clinical practices, robust methods for examining long-term DMT prescription trends are essential.
Methods
Ethics approval
The institutional review boards of the University of Pittsburgh (STUDY20070274 and STUDY21030127) approved the study protocols. The use of de-identified clinical data was deemed exempt.
Results
Patient characteristics
The study cohort included a total of 29,169 patients, comprising 10,301 from UPMC and 18,868 from MGB (Table 1). The overall median age at diagnosis was 45.9 years (IQR: 36.4–56.8), with a higher median age in the UPMC cohort (47.4 years) compared to MGB (44.8 years). The median disease duration was 12.4 years (IQR: 7.0–12.7), shorter in UPMC (11.8 years) and longer in MGB (12.8 years). The majority of the cohort were women (74.3%), with similar sex distributions across sites.
Discussion
In this study, we utilized a large, multicenter EHR dataset from two major healthcare systems (UPMC and MGB) spanning both academic and community care settings to evaluate MS diagnosis phenotyping performance using a novel algorithm (i.e., KOMAP) and to examine longitudinal trends in MS therapies through a temporal KG approach. Our findings demonstrate that KOMAP achieved strong performance in identifying MS patients across both cohorts, with higher accuracy observed when using combined codified and NLP-derived feature sets.
Conclusion
In summary, our study demonstrates the effectiveness of the advanced phenotyping algorithm KOMAP for accurately and efficiently identifying MS patients from EHR data across two large healthcare systems, and highlights the value of the temporal knowledge graph approach in capturing the complex real-world treatment dynamics of a chronic neurological condition.
Acknowledgments
The authors thank all research participants and the clinicians from UPMC and MGB.
Citation: Gan Z, Zhu W, Tang W, Sweet SM, Morris M, Han Y, et al. (2026) Knowledge graph-guided multiple sclerosis identification and therapeutic trend analysis: Real-world evidence from two large healthcare systems. PLOS Digit Health 5(8): e0001554. https://doi.org/10.1371/journal.pdig.0001554
Editor: Xiaoli Liu, Chinese PLA General Hospital, CHINA
Received: December 14, 2025; Accepted: June 19, 2026; Published: August 28, 2026
Copyright: © 2026 Gan et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: Anonymous summary-level registry data and electronic health record (EHR) data will be publicly available. Patient-level data will not be publicly available because patient-level clinical data, whether de-identified information or limited protected health information containing dates of clinical events, or even if anonymous due to concern for re-identification, are universally subject to the rules and regulations of each healthcare system, with which the authors may be affiliated but which may not be the same as their primary academic institutions. Patient-level data contain an individual’s protected health information (e.g., age, sex, date, clinical information such as treatment). Even when the dataset does not contain an individual’s name or medical record number, there is a real concern for re-identification. Questions regarding data request should be directed to the University of Pittsburgh Office of Sponsored Programs.
Funding: This study was funded by the National Institute of Neurological Disorders and Stroke of the National Institutes of Health under award number R01 NS098023 (ZX). The funder had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
