Converting Paper Medical Records to Electronic Version To Support Breast Cancer Translational Research and Clinical Practice.

Abstract

Abstract Background: The healthcare system has extremely large volumes of patients' paper medical records (PMRs) scattered throughout various medical facilities. Currently the industry is transitioning to Electronic Medical Records (EMR). Although each source of information is equally important, a complete longitudinal health record is rarely available or currently attainable. The goal of this effort is to convert the existing PMRs into searchable electronic equivalents and merge them with existing EMRs to create a more complete longitudinal health record to support clinical care and biomedical research.Method: Using a subset of PMRs for subjects enrolled in the Clinical Breast Care Project at Walter Reed Army Medical Center, (WRAMC), we are developing an automated method to digitize and index the records, extract the biomedical information and prepare the data for delivery to clinicians and researchers. The electronic records were loaded into a database for immediate access by clinicians. To support translational research, MRs need to be de-identified, and information extracted and loaded into a research database; we used ten MRs to develop the operational method. Several methods were tested in the de-identification process: 1) manually, by striking out the protected health information (PHI) on paper before it was digitized and 2) “electronically”, by redacting the record electronically on the computer after it was digitized. We are in the process of testing automated data extraction tools and natural language processors to automatically de-identify and extract data from the EMR.Result: We quickly realized that only 10% of PMRs existed onsite at WRAMC; remaining MRs were held by the patients or other medical facilities. Approximately 300 PMRs, containing 66,600 pages, were scanned and digitized for this project. We have created a successful digitization process, which includes creating PDFs with hidden and searchable information. We have compared the effort and accuracy of various redaction methods. To de-identify 10 records, it took ∼10 hours manually and ∼20 hours electronically. It took longer electronically because of the preparations to ensure the removed information could not be retrieved. We expect that automated redaction tools will greatly reduce that effort. We also found that the electronic method had a 99% accuracy compared to 96% for the paper method. A portal prototype to allow access to medical records by clinicians and researchers is currently being tested and evaluated.Discussion: Conversion of non-searchable data into an explorable and computable format will enable clinicians to acquire needed information more conveniently in their clinical care including treatment plan development. Similarly, properly de-identified, complete MRs will serve as a rich source of clinical information to support translational research. Although the method we are developing will initially satisfy the CBCP need for clinical service and research, it will be further developed into a full solution to expand into other disease condition fields. Citation Information: Cancer Res 2009;69(24 Suppl):Abstract nr 5123.