Epidemiology and Patient Journey in Inclusion Body Myositis (IBM): A Machine Learning (ML) Methodology Applied to Claims Data in the United States (US)
Devon Grochowski1, Dave Bradley1, Raymond Mankoski2, Karen Tubridy3
1Real Chemistry, 2The NemetzGroup, 3Abcuro, Inc.
Objective:
Evaluate epidemiology and diagnostic journey of US patients with IBM.
Background:
IBM, a rare, progressive disease characterized by invasion of healthy muscle by highly differentiated cytotoxic CD8+ T cells, is associated with loss of grip strength, difficulty walking, and/or dysphagia. Published global epidemiologic prevalence estimates range from 1.3−50.5/million. We sought to estimate the number of symptomatic undiagnosed IBM patients and define the patient journey.
Design/Methods:
Patients with ≥2 IBM International Classification of Diseases, Tenth Revision (ICD-10) diagnoses between 8/1/2019 and 7/30/2022 (Real Chemistry claims database) were compared with a matched cohort without IBM. Demographics, time to diagnosis, IBM-related symptoms, and resource utilization were assessed. Separately, we used machine learning (ML) to synthesize an “Ideal Patient Population” (IPP) consisting of individuals with ≥2 IBM ICD-10 diagnoses between 1/1/2020 and 4/15/2024 and ≥1 muscle biopsy or myositis antibody test to develop algorithms that identified symptomatic but undiagnosed patients.
Results:
Patient journey: 2435 patients were qualified. 61% were male; 71% were aged ≥65 years at diagnosis. Pain and weakness were predominant pre-diagnosis symptoms, followed by (vs controls): mobility conditions (23% vs 4%), falling (16% vs 4%), and difficulty swallowing (15% vs 2%). Median time between first IBM-related symptom and first IBM ICD-10 diagnosis claim was 4.4 years (3.3 years when pain was excluded as a symptom). In the year before diagnosis, 30% presented to the emergency department (controls, 17%); 20% who were admitted stayed an average of 1 week. Epidemiology: The IPP comprised 1633 patients; over 80% had a mobility condition, and 65% suffered a fall. The US diagnosed prevalence estimate was 26.45/million. Prevalence of symptomatic but undiagnosed patients ranged from 18.72 (highly specific algorithm) to 123.93/million (highly sensitive algorithm).
Conclusions:
There are at least as many symptomatic undiagnosed patients with IBM as there are diagnosed patients. Earlier recognition of IBM is needed to provide appropriate care.
Disclaimer: Abstracts were not reviewed by Neurology® and do not reflect the views of Neurology® editors or staff.