Restating clinical impression of severity index for Parkinson's disease using just non-motor criteria

Clinical impression of severity index (CISI) for Parkinson's disease (PD) is an index to assess PD severity based on four clinical domains. Since this is a continuous value, a categorization policy is proposed by solving an optimization problem. Using this encoding, different comparisons between CISI and other non-motor indexes or items are addressed using wrapper item subset selection and estimation of distribution algorithms. Results show how some of the non-motor items are very relevant, achieving good classification performances when used to predict the CISI severity index