Supervised Models predicting M12 Mental Health and QoL Sample Clauses

Supervised Models predicting M12 Mental Health and QoL. Generalizability of risk prediction models across clinical sites (FORTH)‌ In this version of deliverable D4.4 we extended the assessment of the generalizability of the supervised learning algorithms and associated pipeline across the four clinical sites. Specifically, we focused on the models that address the most challenging clinical question, namely identification of patients who initially report minimal mental health symptoms and/or good QoL and later experience a significant decline in psychological well-being (Models F2 and F4 in D4.3b). In this framework, for each one of the two models (one for mental health and one for QoL) we considered each cohort separately as the testing/validation set of the training model. The training phase was implemented each time on three out of four datasets keeping the last cohort for testing purposes regarding the different combinations of the four available cohorts. This approach therefore validates the proposed ML-based pipeline when new unseen patient data are considered for predicting M12 outcomes given the patient characteristics at baseline (M0) and 3-months post diagnosis.