POS0468 PREDICTION OF RHEUMATOID ARTHRITIS DISEASE ACTIVITY BY AN ADAPTIVE DEEP NEURAL NETWORK: BETTER RESULTS IN SEROPOSITIVE PATIENTS WITH LONGER DISEASE DURATION

Background: Deep neural networks learn from former experiences on a large scale and can be used to predict future disease activity as potential clinical decision support. AdaptiveNet is a novel adaptive recurrent neural network optimized to deal with missing clinical data. In rheumatoid arthritis (RA) it is unknown how disease characteristics influence the predictability by deep learning in terms of classification (e.g. active disease yes/no) or regression (numeric values such as DAS28). Objectives: To investigate in which clinical RA subtypes AdaptiveNet achieves the best results for the prediction of individual disease activity Methods: Demographic and disease characteristics from over 9500 patients and 65.000 visits from the Swiss Quality Management (SCQM) database were used to train and evaluate the network. Patient characteristics, clinical and patient reported outcomes, laboratory values and medication were used as input features. DAS28-BSR served as a target to predict active RA and future numeric individual disease activity by classification and regression. Feature importance was determined by a Random Forest to define the relative importance of variables for disease prediction. Results: AdaptiveNet predicted active disease defined as DAS28-BSR >2.6 at the next visit with an overall accuracy of 75.6% (SD +- 0.7%) and a sensitivity and specificity of 84.2% (SD +- 1.6%) and 61.5% (SD +- 3.6%), respectively. The performance of the prediction for correct disease status was significantly higher in patients with a disease duration >3 years and positive rheumatoid factor. Regression allowed forecasting individual DAS28-BSR values with a Mean Squared Error (MSE) of 0.9 (SD +- 0.05). Compared to Linear Regression, Random Forests and Support Vector Machines, AdaptiveNet showed an increased performance of 7% in MSE. MSE was significantly lower in patients with disease duration > 3 years and with positive anti-CCP antibodies. Feature importance identified number of painful joints, longer disease duration and age as most relevant factors for prediction of remission, whereas medication played a smaller role. Conclusion: Predictability of disease activity in RA by this deep neural network was stronger in patients with a longer disease history and a positive auto-antibody status, potentially due to a more stable disease course. Generally, AdaptiveNet had a superior capacity to predict numeric RA disease activity compared to classical machine learning architectures, however all investigated models had limitations in low specificity. References: [1]Hugle M, Kalweit G, Hugle T, Boedecker J. Dynamic Deep Neural Network For Multimodal Clinical Data Analysis. Stud Comput Intell: Springer Verl. 2020. Acknowledgements: We thank all rheumatologists and their patients for participation to SCQM.The entire SCQM staff was instrumental for data management and support. A list of rheumatology practices and hospitals that are contributing to the SCQM registries can be found on http://www.scqm.ch/institutions. Disclosure of Interests: None declared