Healio
March 1, 2023
Robert Herpen
Key takeaways:
- A machine-learning model produced 46,644 estimated scores on a disability scale from 3,404 scores.
- Application could increase study volunteers and create potential for a tool to tailor treatment plans.
WEST PALM BEACH, Fla. — Application of a machine learning model to databases of patients with multiple sclerosis expanded understanding of disease progression, according to a poster at ACTRIMS 2024.
“The [Expanded Disability Status Score] is a critical endpoint in clinical trials, but in the real world it’s not routinely collected,” Carl D. Marci, MD, managing director of mental health and neurosciences at OM1 Inc., a Boston-based health care technology company, told Healio. “We used an AI machine learning model to read neurologists’ notes and essentially generate a score where there wasn’t one before.”
Marci and colleagues attempted to assess feasibility of a machine-learning model in reducing the number of missing EDSS scores for individuals with MS, to characterize and simplify understanding of disability progression from relapsing-remitting forms of the condition to the secondary progressive form. Application of the model would, in turn, increase the number of potential volunteers for future studies concerning disease progression and degree of outcomes, while creating the potential for a bedside decision-support tool for neurologists to accurately tailor treatment plans.
Their model was applied to OMI Inc.’s MS PremiOM Dataset, which contains deidentified, real-world clinical and administrative data on more than 17,000 individuals with MS in the United States between 2013 and 2021. Analysis identified 4,366 individuals (3,568 with relapsing-remitting MS (RRMS); 556 with primary-progressive MS (PPMS); 242 transitioning between diagnoses)
A total of 3,404 clinician-administered EDSS scores were documented. When the machine-learning model was applied, an additional 46,644 estimated EDSS scores became available, allowing for more specific definitions of disability progression in the analyzed cohort. Subsequently, those in transition from RRMS to PPMS were confirmed to have increasing disability via an increased average EDSS score by month from the date of documented transition.
“Now, we can look at the journey of every patient over time,” Marci told Healio. “And what we’re doing with our clients and our partners is doing things like looking at health economics and different outcome measures to see if we can personalize interventions and therapeutics for multiple sclerosis.”