The number of Electronic Health Records (EHR) collected by healthcare providers is growing at an unprecedented pace. Clinicians often compose detailed clinic letters to record as much essential information during consultations as they can. This increases the workload of analyzing these letters, performing individual and collective analysis, and clinical decisionmaking. This paper presents a novel visualization tool, LetterVis, to support the analysis of clinic letters through advanced interactive visual designs and queries. We describe a letter-space that facilities the visual exploration of content and patterns inside a letter. Letters are processed using Natural Language Processing (NLP) techniques and explored in multiple linked interactive views providing different levels of abstraction. The tool includes customized visual designs and views for visualizing antiepileptic drugs (AEDs). We provide a range of filtering and selection options to assist pattern finding and outlier detection. We demonstrate LetterVis with three case studies using anonymized clinic letters, revealing insight that is normally either time-consuming or impossible to observe. Domain expert partners from EHR analysis review the software and are involved in every phase from the initial design to evaluation.