Aims: In multiple sclerosis (MS), due to the high complexity and using the mean squared error (MSE), log-likelihood, and Harrell’s concordance statistic (C-Index). Results: The results of the analysis revealed that predictive models provide robust and accurate predictions and generalize to new patients and clinical sites. The output of PHREND is an independent recommendation for the therapy that is statistically most likely to succeed for each individual patient, presented in a transparent and easy-to-understand way. Conclusion: Applying personalized predictive models in RRMS patients is still new territory that is rapidly evolving and has many challenges. The proposed framework addresses the following challenges: robustness and accuracy of the predictions, generalizability to new patients and clinical sites, and comparability of the predicted effectiveness of different therapies. Nevertheless, we present the PHREND App already implemented in German doctors’ offices and we plan to expand our model to several other neurological disorders. uncertainty of disease progression, it currently takes long time to find an appropriate therapy for an MS patient and currently the treatment decision depends on a significant portion of intuition. Our solution to support relapsing remitting multiple sclerosis (RRMS) patients and their doctors during this difficult journey is to distil the large amounts of available data into meaningful and relevant decision-making information as efficiently as possible. Methods: This study employed clinical real-world data recorded in the NTD MS registry, a Germany-wide network of physicians in the fields of neurology and psychiatry founded in 2008. The registry includes about 25.000 patients with MS, which represents about 15% of all MS patients in Germany, with an average of observation period of 8,7 years. The PHREND App contains a data-driven mathematical regression model to predict MS disorder progression for individuals based on different therapies, taking into account two indicators of therapy effectiveness: number of relapses, and confirmed disability progression. We employed hierarchical generalized linear models (GLM) for both indicators of treatment response, with model parameters depending on patient’s profile and treatment. Additionally we implemented cross validation and quality of predictions assessment.