Since 1988, Scott & White Hospital and Clinic has been actively involved in medical informatics in general, and in developing an electronic medical record system in particular. Dr. Dennis Myers and Dr. Kim Culp, two researchers at the Texan hospital, have designed a comprehensive data model, describing the creation and flow of information in the Scott & White integrated health care delivery system. They linked the model to two key elements of any record system: the user "interface", and the data repository. In the meanwhile, the database has grown to about six million documents. Plans have been made to expand the system to all of Scott & White's 18 Regional Clinics, as well as to make the system accessible, under the strict condition of proper authorization and security, to users from home via dial-up.
The starting point for launching the Scott and White Electronic Medical Record (EMR) Project was the aim to build a suitable database from scratch to perform large aggregate clinical studies, and to fully support the information needs of clinicians when seeing individual patients. Dr. Myers, from the Department of Pulmonology, and Dr. Culp, working in the Division of Gastroentorology, had become extremely frustrated with the limitations of paper medical records in providing useful patient data. Requirements for a newly developed integrated EMR database would have to contain information about patient care, administration, and billing to accommodate both patient care and aggregate analysis needs. Therefore, the back-end data system should be tightly linked in its logic to a front-end interface.
The first interface had to address the interests of individual clinicians when caring for patients and has been labelled PIE, the Physician Interface Engine. The second would serve the needs of "interrogators" of the data, such as administrators and researchers, and thus be installed as ARIES, the Administrative/Research Interface Engine System. Initial work on interface design modelling was undertaken under contract by the Advanced Interface Design Team of Los Alamos National Laboratory. In 1991, the final choice fell on an object-based textual database as the core component of the back end database system. It would link to a object-relational database to allow structured data, like a master patient index, to be integrated with textual data under the comprehensive data model.
By 1993, a prototype Unix X-Windows based PIE was developed as well as a Unix-based back-end database system, called Text Machine and delivered by Synthesys Technologies. Its suite of Unix-based generic document browsers were planned to be used as a transitional approach to a true PIE and ARIES. Next, ASCII-based automated electronic data interfaces to the Text Machine database were constructed for the Dictaphone transcription system and the report-writer output of Lab Force, the Laboratory Information System. The ASCII document inputs to the EMR were marked up in an automated fashion under SGML constructs by the Synthesys Technologies database software. In July 1993, the document accumulation started in a production mode. The Department of Psychiatry became the deployment prototype unit for further testing and refining of the system.
In early 1994, an embryonic but functioning document management system was implemented. The next year, a functional production PIE-based system was ready to automatically "pull" electronic patient documents, based on data from the appointment system. The documents were built into the PIE's graphical representation of a medical record, allowing users to work quickly through the information present in a patient's electronic record. The interface supported virtual views of data and reconstruction of the views into customized, individualized versions. In autumn 1995, the hospital embarked on a two year project to replace the home grown systems with commercial or standard systems. The core technology present in the original Text Machine back-end text database management software was built into a suite of health care data applications called EMRx, which were Windows-based on the front-end, rather than Unix-based.
In the meanwhile, work had been continued in parallel on improving the PIE interface design, and turning all the EMRx links to other systems, modular and based on information interchange standards, such as HL-7. To support aggregate clinical data analysis on a permanent and comprehensive basis, a Clinical Data Analysis Laboratory (CDAL) was created and staffed with data analysts, statisticians, and biometrists in 1996. After eight years of labour, including three years of production level testing, the Scott and White EMR Project was formally put into large-scale production. About 25 users were assigned passwords to the data browsers. By June 1998, there were more than 1500 password-registered clinical users who "pulled" over 20.000 electronic medical records via the system monthly. The database has grown to nearly six million documents, occupying about thirty gigabytes of storage space.
All central transcription at Scott and White is now captured and appears in the system for use no later than one hour after it is transcribed. All x-ray reports are captured in the same time window. Lab reports are batch-loaded into the system electronically on a daily basis. Certain memos relating to patient care are equally integrated into the system. The EMR database is also used extensively by Medical Records and Billing departments as the primary data source for documentation and compliance issues. Information extracted by the CDAL methodology from the source document databases is ported to relational databases for further analysis. As a result, secondary databases to support specific analysis needs can be automatically generated. This same methodology/process is currently being used as a framework to establish the disease-specific care management repositories/registries for hepatitis C and peripheral vascular disease; colorectal cancer screening; childhood infectious diseases; and monitoring radiographic reports to enhance quality of care.
Plans are underway to convert the Scott and White Hospital network to an Asynchronous Transfer Mode (ATM) architecture. The EMRx has led to the development of superior retrospective research and provides ready access to many on-line reference materials, including Scientific American Textbook of Medicine, all of the MicroMedix series on pharmacology and poisonings, and many other references. The ability to search electronically for patients who meet specific study criteria has increased the variety of answerable questions on diagnosis, treatment, and patient outcome. The electronic data collection is much faster than manual reviews of paper medical records. Over the past year, investigations included potential drug interactions, patient compliance, costs of pharmacological regimens, and physician prescribing habits.
Most recently, the clinical data collected through the EMRx has begun to be paired with the financial data. This will enable better reporting of charge and costs related to specific health events for research and quality investigations. The system also supports an innovative physician's workstation. An offshoot of the project has been the development of a patient-friendly fully automated, touchscreen-based multimedia interactive workstation, which allows direct patient input of answers to questionnaires. More information on the EMRx system is to be found at the Web site of the Scott & White Hospital and Clinic.