An analysis of clinical queries in an electronic health record search utility. Academic Article uri icon



  • New York

MeSH Major

  • Data Mining
  • Medical Records Systems, Computerized
  • Search Engine


  • While search engines have become nearly ubiquitous on the Web, electronic health records (EHRs) generally lack search functionality; furthermore, there is no knowledge on how and what healthcare providers search while using an EHR-based search utility. In this study, we sought to understand user needs as captured by their search queries. This post-implementation study analyzed user search log files for 6 months from an EHR-based, free-text search utility at our large academic institution. The search logs were de-identified and then analyzed in two steps. First, two investigators classified all the unique queries as navigational, transactional, or informational searches. Second, three physician reviewers categorized a random sample of 357 informational searches into high-level semantic types derived from the Unified Medical Language System (UMLS). The reviewers were given overlapping data sets, such that two physicians reviewed each query. We analyzed 2207 queries performed by 436 unique users over a 6-month period. Of the 2207 queries, 980 were unique queries. Users of the search utility included clinicians, researchers and administrative staff. Across the whole user population, approximately 14.5% of the user searches were navigational searches and 85.1% were informational. Within informational searches, we found that users predominantly searched for laboratory results and specific diseases. A variety of user types, ranging from clinicians to administrative staff, took advantage of the EHR-based search utility. Though these users' search behavior differed, they predominantly performed informational searches related to laboratory results and specific diseases. Additionally, a number of queries were part of words, implying the need for a free-text module to be included in any future concept-based search algorithm. 2010 Elsevier Ireland Ltd. All rights reserved.

publication date

  • July 2010

has subject area

  • Data Mining
  • Medical Records Systems, Computerized
  • New York
  • Search Engine



  • Journal Article



  • eng

PubMed Central ID

  • PMC2881186

Digital Object Identifier (DOI)

  • 10.1016/j.ijmedinf.2010.03.004

PubMed ID

  • 20418155

Additional Document Info

start page

  • 515

end page

  • 522


  • 79


  • 7