Jose Luis Morell

November A&R Clinical Connections (1 of 2) Machine Learning to Predict Remission in Rheumatoid Arthritis Key points • In external validation, the full machine learning model achieved an area under the curve of 0.797 and a negative predictive value of 0.902. • The machine learning model was more useful for identifying patients unlikely to achieve remission than for confirming which patients would achieve remission. • The most influential predictors of remission included patient global assessment, tender joint count, previous biologic or targeted synthetic DMARD exposure, HAQ-DI, and physician global assessment. • A simplified 10-variable model and conventional logistic regression performed similarly to the full machine learning model. acrjournals.onlinelibrary.wi…