This paper presents a pilot study on the process of manually annotating natural language EHR questions with a formal meaning representation. This formal representation could then be used as a structured query as part of a natural language interface for electronic health records. This study analyzes the challenges of representing EHR questions as structured queries as well as the feasibility of creating a sufficiently large corpus of manually annotated structured queries for EHR questions. A set of 100 EHR questions, sampled from actual questions asked by ICU physicians, is used to perform the analysis. The ultimate goal of this research is to enable automatic methods for understanding EHR questions for use in a natural language EHR interface.