Abstract Archives of the RSNA, 2013
Adam Randolph Travis MD, Presenter: Nothing to Disclose
Paul J. Chang MD, Abstract Co-Author: Co-founder, Stentor/Koninklijke Philips Electronics NV
Technical Advisory Board, Amirsys, Inc
Medical Advisory Board, Koninklijke Philips Electronics NV
Medical Advisory Board, MModal Inc
Medical Advisory Board, lifeIMAGE
Medical Advisory Board, Merge Healthcare Incorporated
Yuechen Qian, Abstract Co-Author: Nothing to Disclose
Merlijn Sevenster PhD, Abstract Co-Author: Employee, Koninklijke Philips Electronics NV
Gabriel Mankovich BSC, Abstract Co-Author: Nothing to Disclose
Johannes Buurman PhD, Abstract Co-Author: Employee, Koninklijke Philips Electronics NV
Physicians do not always provide adequate histories when ordering imaging studies; this may be due to Computerized Physician Order Entry (CPOE) systems that allow limited codified indications (“drop down menus”) as input parameters. Lack of history may result in suboptimal or even incorrect interpretation by radiologists. We test the hypothesis that a PACS-integrated view of patient history automatically synthesized from prior radiology reports improves the quality of clinical history sections in radiology reports.
CCI functions as a PACS plugin that extracts pertinent information from prior radiology reports and displays it along three “axes” for each exam: history, acute indication, and follow-up recommendations. CCI uses natural language processing (NLP) to populate the history and follow-up axes by extracting and filtering unique sentences from relevant sections in prior reports. The acute indication axis is populated with the Reason For Exam (RFE) from the CPOE system.
Prospective evaluation was conducted by a team of senior residents in normal workflow. First, the reader viewed the CCI summary and dictated the history based only on this information. Then, the reader reviewed all pertinent patient data (e.g., pathology, labs) from the EHR and modified the dictated history, if necessary. Later, for each dictated study an attending radiologist compared the quality of the initial CCI-only history to the final dictated history and to the RFE, which were each used as baselines.
Preliminary data on 32 neuro CT cases shows that 34.4% of CCI-only histories were rated significantly more complete than RFE histories. However, CCI-only histories were significantly augmented with pathology (18.8%) and/or other data (9.4%) derived from the EHR.
Patient clinical context derived from CPOE exam indications alone were enhanced by the automated extraction and PACS-integrated presentation of information derived from prior radiology reports. However, additional important patient information was derived from the EHR. Therefore, automated PACS-integrated tools designed to present patient context should extract data from both prior radiology reports and the EHR.
Automated PACS-integrated tools designed to present patient context should extract data from both prior radiology reports and the EHR. These tools can enhance information provided by CPOE alone.
Travis, A,
Chang, P,
Qian, Y,
Sevenster, M,
Mankovich, G,
Buurman, J,
Enhancing Provided Patient Clinical Information by Automated Review of Prior Radiology Reports Using the Clinical Context Indicator (CCI): A NLP Based Data Extraction and Presentation PACS-integrated Tool. Radiological Society of North America 2013 Scientific Assembly and Annual Meeting, December 1 - December 6, 2013 ,Chicago IL.
http://archive.rsna.org/2013/13023785.html