Abstract
Background/Aims
Chart notes and other text blocks in Epic’s electronic medical record system generally appear to be free text, requiring natural language processing (NLP) for analysis. In fact, though, many notes contain boilerplate text that can be traced back to structured data created using various Epic “SmartTools.” We investigated Clarity, Epic’s reporting database, to learn how to locate these structured data.
Methods
Through trial and error we developed a data map to guide searches for structured data in note text. Starting with a known block of text, we found the underlying boilerplate, traced the sources of the individual data elements, and finally located the values recorded for those elements in the encounter.
Results
Smart phrases, lists, links and forms are a few of the commonly used Smart elements in Epic. Each of these elements may be composed of other elements, increasing the difficulty of charting a straight path. So a SmartPhrase may include the option to choose from a SmartList, and those choices may, in turn, reference SmartLinks. However, if one can locate, for example, a definitive statement of pathology results within the structured Smart data of a results letter to the patient, it may be possible to avoid much of the expense and uncertainty involved in using NLP to parse the same information. Similarly, if some note text is known to be a direct copy of structured data from another part of the chart, one can safely exclude that text from further examination. Even if ones does not analyze the structured data directly, discovering which text is “Smart” and the expected values for the included data elements can greatly facilitate the development of NLP algorithms. It is important to work with clinicians at this first level of discovery, to learn the mnemonics or “dot-phrases” they use to invoke SmartTools and the text that results.
Conclusions
Epic’s SmartTools hold great promise for teasing structured data out of apparently unstructured text. However, unravelling the complex interrelations among SmartTools requires a good deal of motivation and ingenuity. Further investigation is needed to determine the most appropriate and effective uses of these data.
Keywords: Clarity, Structured data