Objectives
Relapse among multiple sclerosis (MS) patients is associated with disability progression and worsening outcomes. This study aims to identify characteristics of inpatient MS relapse using claims data and machine learning techniques.
Methods
MS patients with a new prescription or administration of a disease modifying therapy (DMT) were identified based on ICD-9/10 diagnosis codes in de-identified Optum® Clinformatics® Data Mart from 2000-2019. The first DMT date was the index date with >=2 MS diagnoses required in the preceding 6 months (baseline). Inpatient relapse was defined as an inpatient visit with a primary MS diagnosis code during the 12 months following the index date. Features included demographics, comorbidities, concomitant medications, healthcare resource utilization (HRU), DMT route of administration and proportion of days covered (PDC) for DMTs. Five-fold cross validation was used to tune and evaluate traditional and regularized logistic regression, XGBoost, support vector machine, random forest and feed-forward neural network models. The best model was selected using the area under the ROC curve (AUC) and accuracy, recall, precision, F1-score and specificity were assessed.
Results
Inpatient relapse was observed for 984 (5.2%) patients of 18,820 patients (mean age=44.2 years; females=75.8%). The XGBoost model had the best AUC (AUC=79.3%; accuracy=74.3%; recall=69.7%; precision=13.1%, F1=0.22, specificity=74.5%). Predictors of inpatient relapse included MS related HRU measures (previous IP or ER visit with MS diagnosis, number of MS related encounters, utilization of home care services and durable medical equipment), epilepsy/convulsions, paralysis, urinary tract infections, potential medication side effects (nausea and vomiting), use of muscle relaxants, anticonvulsants and antidepressants. Factors protective of relapse were increased PDC, older age, DMTs administered as infusion, Caucasian race and being female.
Conclusions
Our study identified demographic and clinical predictors of inpatient MS relapse with high predictive accuracy. Our findings can potentially be utilized to better manage patients at high risk of relapse.
Article info
Identification
Copyright
© 2021 Published by Elsevier Inc.
User license
Elsevier user license | How you can reuse 
Elsevier's open access license policy
Elsevier user license
Permitted
For non-commercial purposes:
- Read, print & download
- Text & data mine
- Translate the article
Not Permitted
- Reuse portions or extracts from the article in other works
- Redistribute or republish the final article
- Sell or re-use for commercial purposes
Elsevier's open access license policy
