The Predictive Value of Billing Codes and Treatment Data to Identify Relapse Hospitalizations among Patients with Neuromyelitis Optica Spectrum Disorder (NMOSD)
Mattia Wruble1, Philippe-Antoine Bilodeau1, Sathya Narasimhan2, Danielle Kei Pua3, Michael Levy1, Shamik Bhattacharyya1
1Neurology, Mass General Brigham / Harvard Medical School, 2Neurology, Baylor College of Medicine, 3Neurology, Westchester Medical Center
Objective:
To identify the combination of hospitalization billing codes and treatments that best distinguishes relapse from non-relapse hospitalizations among patients with NMOSD.
Background:
Patients with NMOSD may be hospitalized for relapse or non-relapse (e.g., treatment-related complications, unrelated comorbidities) events. There are no real-world studies validating hospitalization billing codes to predict relapse-associated hospitalizations among patients with NMOSD.
Design/Methods:
Patients with NMOSD hospitalized within a healthcare system between January 2016 and March 2025 were included. Each hospitalization was adjudicated as relapse (admission for a core demyelinating event) or non-relapse (clear alternate etiology and/or no new clinical or imaging evidence of relapse) by a neuroimmunologist. For each hospitalization, ICD-9 and 10 billing codes associated with neuroinflammatory events (NMOSD, optic neuritis, transverse myelitis, other neuroinflammatory disorders [ONID]) and common infections were examined, and administration of acute immunotherapy was reviewed. The sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) for predicting relapse hospitalization were calculated for 40 combinations of billing codes and treatments.
Results:
53 patients -- median age of 53, 77% female, 84% seropositive -- with 133 hospitalizations (52 [39.1%] relapses, 81 [60.9%] non-relapses) were included. The NMOSD billing code alone (G26.0 or 341.0) showed a sensitivity, specificity, PPV, and NPV of 75.0%, 70.4%, 61.9%, and 81.4%, respectively. The combination of NMOSD or ON or TM or ONID codes with intravenous (IV) steroid treatment (without consideration of infection) yielded the best predictive value (sensitivity 96.5%, specificity 77.8%, PPV 77.8%, NPV 90.0%). In general, the addition of other immunotherapies and the absence of infection increased specificity but decreased sensitivity.
Conclusions:
The NMOSD billing code alone only moderately predicts relapse hospitalization in NMOSD patients, while the combination of NMOSD or ON or TM or ONID billing codes with IV steroid use best predicts relapse hospitalizations. This variability underscores the need to validate billing codes to facilitate accurate real-world NMOSD research.
Disclaimer: Abstracts were not reviewed by Neurology® and do not reflect the views of Neurology® editors or staff.