SP09C - Factors Associated with Initiation of Oral vs. Injectable Disease Modifying Therapies Among Multiple Sclerosis Patients: A Machine Learning Approach
Background: The number of oral disease modifying therapies (DMTs) for multiple sclerosis (MS) has increased sharply in the last decade. In the absence of clinical guidelines regarding DMT selection, we sought to understand the factors associated with DMT choice.
Objectives: This study assessed factors associated with oral DMTs as first-line therapy among MS patients using a diverse U.S. EHR database.
Methods: MS patients over 18 years with a new prescription of an oral or injectable DMT were identified in de-identified Optum® Clinformatics® Data Mart from 2000-2019. The first DMT date was the index date with ≥2 MS diagnoses and continuous enrollment required in the preceding 12 months (baseline). The study measure was the DMT route of administration. A high dimensional set of covariates from the baseline was defined with patient demographics, history of comorbidities, medications and procedures, measures of healthcare resource utilization (HRU) and specialties of prescribing and attending providers. Elastic net (EN) models were trained and evaluated using a 60%/40% data split to select a sparse subset of correlates of oral DMT selection. The area under the ROC curve (AUC) was used as the evaluation metric for tuning and recall, precision, specificity and Brier score were assessed.
Results: Among 7,000 MS patients initiating DMTs, the median age was 46 (IQR, 36-55), 75.5% were female and 41.2% initiated an oral DMT. The EN model had an AUC of 76.3% (recall=72.4%; precision=58.0%, specificity=63.2%, Brier score=0.19) and identified a subset of 60 from 204 potential correlates. Positive correlates of oral DMTs included having a neurologist as the index DMT prescriber, an index date on and after 2014, history of skin ulcers, male gender and Medicare Advantage low income subsidy membership. Additionally, oral DMT patients had higher rates of tests for varicella-zoster antibodies and tuberculosis antigens, testing for hepatic function and complete blood counts, electrocardiograms, ophthalmologist visits and CNS stimulant use. Patients 65 years or older, with a baseline IP or ER visit, using durable medical equipment, history of bacterial infection, liver diseases, encephalitis, hypertension, heart valve disorders, headaches, GI and esophageal disorders, asthma were less likely to start on an oral DMT.
Conclusions: This study demonstrates several associations between patient and prescriber characteristics and the selection of oral DMTs including an increase in uptake in more recent years, access to specialists, younger age, being male, screening for and absence of contraindicated conditions and less severe disease status.