Clinical Decision Support Systems (CDS) are built to make biomedical literature efficiently accessible to clinicians. The typical queries submitted to these systems are medical cases which describe various aspects of a patient health including symptoms, lab tests, and previous treatments. Because these queries are expected to be long and may contain some ambiguous words, performing some reformulation tasks to those queries (e.g., removing and re-weighting some of the query terms) could have a substantial beneficial effect on the performance of CDS systems. In this paper we describe several such reformulations and use them to achieve a major improvement in retrieval effectiveness.