Capturing clinical reports in a large academic medical center: feeding a central patient data repository.

Clinical reports, notes, and other narratives are highly used components in the patient record. Unfortunately, the methods by which these reports are generated are as diverse as the fiscal autonomy of academic clinical departments in a university-based health science center. In this paper, we report on electronically capturing clinical reports, notes, and other text fragments from several hospital sources and many outpatient clinics. The purpose of the capture is to feed the ACIS (Advanced Clinical Information System) central patient data repository that is in use at the University of Utah Health Sciences Center (UUHSC). A survey conducted in early 1994 indicated that about 917,150 reports were generated per year at UUHSC representing about 1.2 million pieces of paper, occupying about 2.3 gigabytes of storage. The most crucial problem encountered in capturing the reports was linking them to the proper patient. Systems that had functioning and well-maintained admit-discharge-transfer (ADT) information performed well, but systems that relied on the human dictator to identify patients, produced patient linkage errors. In our open loop telephone dictation systems this error rate averaged between 6 and 10%. Subsequent to the wide-spread availability of clinical reports on ACIS, this error rate dropped to 3-5%, presumably due to increased demand for on-line availability of this information. From clinical secretaries who use their word processor to create the clinical reports, the linkage error rate was < 1% due to the use of our Advanced Text Upload (ATU) utility. The clinical text component in ACIS contributed significantly to the success of a JCAHO site visit in December 1995.