Neurodegenerative diseases are incurable diseases whose decline impairs the normal life of a human being. A timely diagnosis increases the chance of getting access to modern treatments with the aim of increasing the quality of life. This work presents the application of a novel decision-level fusion framework. This framework fuses classification results of different handwriting tasks with the aim of increase accuracy in detecting patients suffering of neurodegenerative disease in their early stage. The framework is structured in three layers: as first layer there are different classifier for each task performed by a patient, in the second layer there is an ensemble algorithm applied on decision fused scenario that come from the previous layer. The third layer, instead, computes accuracies. The results show that fusing decisions of a sub-group of the tasks which makes use of the Random Hybrid Stroke (RHS), outperforms the result of the model without the RHS achieving an Fl-Score of 0.7850.