PubMed1995
An analysis of paper-based charting was carried out at Sunnybrook Health Sciences Center as a prelude to developing a strategic plan to implement an electronic patient record. A relational model of the Sunnybrook paper chart was developed, describing each of its forms in terms of specific data fields. Three hundred and forty nine different forms are in current use at Sunnybrook, containing 64 types of data fields such as Patient Demographics, Vital Signs, and Doctor's Orders. The extent of data field duplication at the level of hospital forms was significant. A Patient Demographics field was present on all forms and on all pages of a patient's chart, as would be expected. Twenty seven other fields were duplicated on more than ten different forms, including Working Diagnosis which was present on 110 forms, History of Past Illness on 42 forms, and History of Present Illness on 32 forms. Current Medications were recorded on 32 forms and Allergy fields were present on 29 separate forms. Only five data fields of the total 64 were present on only one form. The duplication of data fields within complete patient charts was then examined to confirm that data field duplication was occurring within the actual healthcare delivery process. Using the relational model of the charting system, 143 acute care in-patient encounters were abstracted into the database. The charts were selected randomly from each of the hospital inpatient services. The numbers and types of forms within each chart were recorded, amounting to 18,654 physical pages and using 165 of the different forms. The average in-patient encounter within the model was 130.4 pages long, with a minimum of 27 pages, a maximum of 559 pages, and containing on average 25.8 different forms. The duplication of data fields within actual charts followed a pattern similar to the duplication found on the forms. Initial diagnosis was present on an average of 20.4 pages within the charts, with a minimum of 2 pages and a maximum of 152 pages containing this data field. Other frequently occurring data fields included History of Past Illness present on an average of 12.2 pages per chart, History of present illness on 10.2 pages per chart, and allergies on an average of 9.1 pages per chart. The results obtained through the examination of Sunnybrook charts should be generalizable to most paper-based systems. This study did not measure the workload associated with duplicating chart data. however, the magnitude of the duplication seen in this study leaves little doubt that significant amounts of time could be saved with appropriate modification of health care delivery processes. Multiple copies of the same data field within a database are referred to as aliases of each other. This study demonstrates that hospital charts contain many copies of the same information, such as medication lists, allergies etc. Due to manual replication of data fields, there is no mechanism to ensure that each copy of a data element within a chart actually contains the same information. This aliasing of data through manual duplication compromises the integrity of data within paper-based charts. Decisions and therapy based upon contradictory or inaccurate data are likely to lead to inefficient or erroneous care delivery; this has significant implications for hospital liability and quality of patient care. A growing recognition of the costs and risks of data replication within paper-based charting systems is driving the healthcare industry toward Electronic Patient Record systems. Better quality patient care data, provided by computerized records, is likely to be necessary but not sufficient to improving the efficiency and effectiveness of the healthcare system. Intelligent application of this information will be essential.