Machine Learning Based Risk Assessment and Early Diagnostic Model of Inflammatory Bowel Disease Associated Arthropathy

Abstract

Inflammatory bowel disease (IBD) is a persistent idiopathic disorder responsible for intestinal inflammatory conditions. IBD exhibits several extraintestinal manifestations, the most common being arthropathy, that are important triggers and risk factors for adverse progression of the disease. Due to a lack of definite diagnostic criteria and treatment regime, a need arises for early diagnosis and management of IBD associated arthropathy to reduce its morbidity. The present study utilizes a machine learning approach for development and validation of a risk prediction and early diagnostic disease model for IBD associated arthropathy. A publically available IBD cases vs control dataset from University of Massachusetts Medical School’s institutional repository was taken and the data was filtered into 54 IBD individuals along with their demographic and clinical characteristics. The arthropathy characteristics were incorporated from literature in the IBD dataset. Data was randomly split into training (50%; n=27) and testing (50%; n=27) for development and validation of model using logistic regression. Out of 54 IBD cases, 21 had a higher risk of developing arthropathy. The area under the receiver operator curve for the validated model was 0.90 (95% CI 0.80–0.99; accuracy 96%). This disease model can aid in identification of high-risk individuals and for early diagnosis of arthropathy in IBD cases before reaching the imaging and invasive diagnostic stage. This model warrants for prospective case-control trials validation.