Regenstrief Medical Record System (RMRS)
The Regenstrief Medical Record System (RMRS) is a comprehensive homegrown data model for electronic health records (EHR) developed by the Regenstrief Institute of Indiana University, as well as an associated database-management system (DBMS), communications system, and a suite of clinician-facing client-side applications called the Medical Gopher. The system pioneered Indiana Health Information Exchange—one of the first health-information exchanges—as well as one of the first computer-assisted order-entry workstations.
Between 1972 and 2016, RMRS stored and managed patient records across multiple institutions of the greater Indianapolis area; at its peak, it contained more than 200 million coded observations and 3.25 million narrative reports, for more than 1.5 million patients, across at least fifteen independent healthcare organizations. However, its usage was largely retired in 2016. It is succeeded today by the Indiana Network for Patient Care (INPC), the Indiana Health Information Exchange, the Regenstrief Teaching EMR, and Arden syntax—all of which were originally based upon the RMRS data model.
1972 was a year from the age of the 16-bit minicomputer. That year, Clement McDonald and Charles Clark, two physicians at Sidney & Lois Eskenazi Hospital (then called Marion County General Hospital) used a PDP-11 minicomputer to create a computer-stored medical record for 35 patients with diabetes, with the intent of automating diagnosis and management after manually capturing patient data, using a hardcoded but generic data structure. Their goals included the following:
- Eliminating logistic problems of paper records, such as availability and indecipherability.
- Reducing the work of clinical bookkeeping that is incidental and nonessential to patient management: “no more missed diagnoses when laboratory evidence shouts its existence, no more forgetting about required preventive care”.
- Making informational “gold” in medical records available to clinical, epidemiological, and managerial research.
After encountering difficulty with capturing comprehensive data for the pilot population of 35 patients at their diabetes clinic, McDonald, Clark, and other colleagues would build several systems for capturing data at laboratory and pharmaceutical sources.
Data model and RDBMS
Over 1972–1974, it was found that maintaining a hard-coded data model was not sustainable, due to having to change multiple programs whenever any change was made to their data model’s file structure; the decision was therefore made to create a relational-database management system (RDBMS), due to no commercial DBMS being available for the PDP-11 microcomputer being used.
RDBMS, by 1989, ran on several VAX superminicomputer workstations and stored the RMRS relational data model in several large files. Records were physically sorted by patient ID, observation ID, and date/time (reverse chronologically). A single term dictionary controlled the system’s concept vocabulary, including all diagnoses, findings, variables (e.g., Glasgow coma score, diastolic blood pressure), and sets of terms. Questionnaires were defined as structured lists of questions and variables stored in the RMRS; these questionnaires were flexible and accepted codes from many different code systems (e.g., ICD-9, CPT-4, local Regenstrief codes) that were associated with each variable’s term. The physical RDBMS and its data model resembled the following:
- Dictionary/master files containing:
- Terms, each with a name, synonyms, external references (e.g., ICD-9 and SNOMED), etc.
- Structured-observation types, e.g., “microbiology”.
- Numeric-observation types, each with a scale factor, reference ranges, etc.
- Coded-observation types, each with ordinals, nominals, etc.
- Doctors, each with a name, a license, a speciality, etc.
- Relationships between these relations:
- Structured-observation types, numeric-observation types, and coded-observation types were subclasses of terms.
- Each term could be an aggregate of at least two structured-observation types as its “elements”.
- Each term could have any number of coded-observation types as its value set.
- Each term could have one numeric-observation type as its unit.
- Operational data containing:
- Encounters, each with a date, billing data, etc.
- Orders, each with a date/time, a priority, notes, etc.
- Patients, each with a name, address, etc.
- Medical-record items, each with a date/time, etc.
- Narrative reports, each with a header, text, etc.
- Numeric observations, each with a value, etc.
- Binary data, e.g., images, voice recordings, electrocardiograms, etc.
- Coded observations.
- Relationships between these relations:
- Each encounter was associated with one patient.
- Each order was associated with one encounter.
- Each medical-record item was the result of one order.
- Each medical-record item was associated with one patient.
- Each medical-record item had one term as its parameter.
- Narrative reports, numeric observations, binary data, and coded observations were subclasses of medical-record items.
- Each coded observation could have any number of terms for its value(s).
Data capture, research, and display
By 1984, comprehensive multisite data capture by RMRS had improved, and new data standards were being developed to reduce the necessity of ancillary service systems. And even as early as the 1980s, several retrospective clinical studies would be performed using the RMRS dataset.
In addition, computerized physician order entry (CPOE) in the outpatient setting was initiated in 1984, which was later extended to the inpatient setting in 1990. These CPOE systems would evolve into elaborate “Medical Gopher” client-side workstations. Although RMRS’s developers noted initial physician unhappiness in reaction to computerized order entry, this unhappiness would later improve as the Gopher system would improve.
Even at these early stages, RMRS generated much research about clinical decision support in the outpatient setting, especially regarding rule-based computer reminders. Furthermore, in the 1980s, RMRS’s reminder-rule systems and Health Evaluation through Logical Processing (HELP) would give rise in 1989 to Arden syntax, a markup language for representing shared medical knowledge, which would later become an HL7 standard in 1999 and an American National Standards Institute (ANSI) standard in 2002..
By the 1990s, RMRS continued to grow and RMRS was gradually expanded from the outpatient clinics to inpatient settings. It would facilitate the care of more than one million patients while generating dozens of studies. At its peak circa 1999, it contained more than 200 million coded observations, 3.25 million narrative reports, 15 million prescriptions, and 212,000 electrocardiograms, for more than 1.5 million patients—across at least fifteen independent healthcare organizations, with access by more than 2500 employees.
Core system: three-legged stool
RMRS’s flagship institutions continued to be Sidney & Lois Eskenazi Hospital (which by this period had been renamed from Marion County General Hospital to Wishard Hospital), as well as the three hospitals of Indiana University Health (formerly Clarian Health)—together with Regenstrief, these together were characterized as RMRS’s “three-legged stool”.
Its homegrown central relational-database management system (RDBMS) ran on DEC Alpha servers with an ancillary Novell personal-microcomputer network, both hosted by Eskenazi Hospital. By 1999, RMRS also managed the medical records of all outreach/homeless clinics in the Indianapolis area, as well as mental-health clinics. These would connect with the central Eskenazi hub system, via ethernet cables, Transmission System 1 (T1) lines, and dedicated telephone lines, largely with HL7 messages over Internet Protocol (IP).
In this core system, RMRS captured clinical data from more than two dozen disparate systems, such as direct capture from bedside measuring devices, direct links to electrocardiography carts, direct links to financial systems, regular technician entry from public health records, nurse radio/computerized dictation and voice recording, clerk-entered questionnaire/form results, and computerized physician entry via “Medical Gopher” client-side microcomputer workstations. Data standards used as of 1999 included HL7 messages, DICOM, NCPDP, LOINC, ICD-9, CPT-4, and ACR radiology codes; UMLS and SNOMED were being explored but not supported at this time.
Physician order entry and clinical-data charts were initially displayed in text-based interfaces from “Medical Gopher” computer terminals. Eventually, graphical interfaces in web pages were also supported; hard copies were also printed on paper—in general, these were designed to yield specialty-specific, single-page, information-dense overviews of each patient, akin to hospital “scut” cards.
Informational feedback such as alerts, e.g., for contraindicated medications, occurred during order entry; research was extended from the outpatient setting to the inpatient setting during the 1990s. Order suggestions were also given depending on rules expressed in a pair of the homegrown rule languages, CARE and G-CARE; these included including “corollary orders”, as well as formulary restrictions from several health-maintenance organizations (HMOs); for example, entering DVT in a problem list would trigger order sets for heparin and warfarin. Physicians, however, were “never blocked from completing an order” but were rather made “aware of the potential problems that an order may cause” .
Statewide health-information exchange
Outside of the four-hospital core system, the RMRS data model was further developed during this period to span multiple healthcare systems: separate medical repositories (data “vaults”) were created for each healthcare systems, but these vaults would share a single standardized term dictionary and a “global patient index”. A personal-name matching algorithm would determine which patient records in different healthcare systems referred to the same patient, assigning them the same “global patient ID”. This “cross-enterprise” normalization occurred at a low level in the data model, and “virtual medical records” from multiple healthcare systems were able to be automatically displayed to clinicians with minimal modification to client applications.
By 1996, the now-“cross-enterprise” RMRS had evolve into one of the first statewide health-information exchanges (HIE): the Indiana Network for Patient Care and Research (INPCR) database. This would later evolve during the late 1990s into the Indiana Network for Patient Care (INPC) database.
A 1999 review of RMRS characterized its clinician users across its several core system as “happy”; it identified several challenges for the future, including further accommodating dictation-based workflows, further developing its then-new web-based graphical user interface, exploring wireless and remote clinician access, and maintaining secure authentication with biometric technology.
However, by 2014, the three hospitals of Indiana University Health (formerly Clarian Health) had switched from to Cerner’s EHR suite, and Sidney & Lois Eskenazi Hospital remained the only major healthcare institution directly using RMRS and its Gopher system for order entry.
In the meantime, the Regenstrief Institute in 2004 gave control of INPC to a new nonprofit corporation, the Indiana Health Information Exchange (IHIE), which was formed from an alliance between thirteen Indiana healthcare institutions and incorporated in 2004, with the goal to create sustainable business models and provide commercial-level support for INPC, while the Regenstrief Institute would continue managing research access to its information.
RMRS’s Gopher system continued to serve Sidney & Lois Eskenazi Hospital’s patients until 2016, when Eskenazi Hospital switched from RMRS to Epic. This was a part of a greater trend of the replacement of homegrown EHRs in the United States by vended products that were government-certified under meaningful-use programs.
As of 2021, the Indiana Health Information Exchange (IHIE) and the Regenstrief continue to manage INPC and provide statewide health-information exchange via several client-side applications, which were originally derived from RMRS’s homegrown data model from the 1970s, and which and continue to use the cross-enterprise global patient index that RMRS pioneered in the 1990s.
Regenstrief, the Indiana School of Medicine, Sidney & Lois Eskenazi Hospital, and the American Medical Association jointly created the Regenstrief Teaching EMR (tEMR) as a “copy” of the RMRS data model and other technology. The Regenstrief tEMR has been successfully deployed in at twelve health-professional schools with over 11,000 unique student users as of 2021.
RMRS’s clinical decision support in the 1980s, along with Health Evaluation through Logical Processing (HELP), would give rise in 1989 to Arden syntax, a markup language for representing shared medical knowledge. These continue on as HL7 and American National Standards Institute (ANSI) standards..
- McDonald CJ, Overhage JM, Tierney WM, et al. The Regenstrief Medical Record System: a quarter century experience. Int J Med Inform. 1999 Jun;54(3):225-53. doi: 10.1016/s1386-5056(99)00009-x.
- Colicchio TK, Cimino JJ. Twilighted Homegrown Systems: The Experience of Six Traditional Electronic Health Record Developers in the Post-Meaningful Use Era. Appl Clin Inform. 2020 Mar;11(2):356-365. doi: 10.1055/s-0040-1710310. Epub 2020 May 20.
- Apathy N, Vest J, Menachemi N, Harle C. Changes in User Perceptions of Their Electronic Health Record Following the Replacement of a Legacy System. AcademyHealth Annual Research Meeting. 2017. Retrieved 2021 Oct 24. https://academyhealth.confex.com/academyhealth/2017arm/meetingapp.cgi/Paper/16422
- McDonald CJ. The medical gopher—a microcomputer based physician work station. Proc Annu Symp Comput Appl Med Care 1984:453–459.
- McDonald CJ, Tierney WM, Overhage JM, Dexter P, Takesue BY, Abernathy G. The three-legged stool: Regenstrief Institute for Health Care. Proc 3rd Annual N.E. Davis CPR Recognition Symp. 1997; 101–147.
- Bob Supnik. Simulators: Virtual Machines of the Past (and Future). 2004. ACM Queue. 2(5).
- Eskenazi Health. History. 2020. Retrieved 2021 Oct 24. https://www.eskenazihealth.edu/about/history
- McDonald CJ. Computer applications to ambulatory care. Proceedings of the IEEE Conference on Systems, Man, and Cybernetics; Boston, Massachusetts, November 5–6, 1973.
- McDonald CJ, Bhargava B. Tree systems for medical information processing. Proceedings of the Eleventh Annual Rocky Mountain Bioengineering Symposium, USAF Academy, Colorado, March 1974.
- McDonald CJ. Action-Oriented Decisions in Ambulatory Medicine. Yearbook Medical Publishers, Chicago, 1981.
- McDonald CJ, Martin G. A model lab information system. Proceedings of the Fifth Annual Pittsburgh Conference on Modeling and Simulation, April 1974.
- Carlstedt B, Jeris DW, Kramer W, Griefenhage R, McDonald CJ. A computer-based pharmacy system for ambulatory patient care. The Indiana Pharmacist 1977; 58:92–98.
- McDonald CJ, Bhargava B, Jeris DW. A clinical information system (CIS) for ambulatory care. Proceedings of the AFIPS National Computer Conference, Anaheim, California 1975.
- McDonald CJ, Day Z, Martin DK, Tierney WM. The Regenstrief medical record: 1989, a campus-wide system. Proceedings of the Symposium on Computer Applications in Medical Care, Washington, DC 1989.
- McDonald CJ, Blevins L, Tierney WM, Martin DK. The Regenstrief Medical Records. MD Comput 1988; 5:34–47.
- McDonald CJ, Wiederhold G, Simborg DW. A discussion of the draft proposal for data exchange standards for clinical laboratory results. Proceedings of Symposium on Computer Applications in Medical Care 1984; 406413.
- McDonald CJ, Hui SL, Tierney WM. Diuretic-induced laboratory abnormalities that predict ventricular ectopy. J Chron Dis 1986; 39:127–135.
- Psaty BM, Tierney WM, Martin DK, McDonald CJ. The value of serum iron studies as a test for iron-deficiency anemia in a county hospital. J Gen Intern Med 1987; 2:160–7.
- Tierney WM, Martin DK, Hui SL, McDonald CJ. Using clinical data to predict abnormal serum electrolytes and blood cell profiles. J Gen Intern Med 1989; 4:375–383.
- Tierney WM, McDonald CJ, Luft FC. Renal disease in hypertensive adults: Effect of race and type-II diabetes mellitus. Am J Kid Dis 1989; XIII:485–493.
- McDonald CJ, Tierney WM, Overhage JM. H.I.S. and the physician: Direct inpatient order entry by physicians through medical gopher workstationsproblems and promises. Proceedings of the IMIA Working Conference on Trends in Modern Hospital Information Systems. (Gottingen, Germany, September 7–11, 1991). North-Holland: Elsevier Science Publishers B.V., 1992 IMIA.
- McDonald CJ. Use of a computer to detect and respond to clinical events: its effect on clinical behavior. Ann Intern Med 1976; 84:162–167. McDonald J. Computer reminders, the quality of care and the nonperfectability of man. N Engl J Med 1976; 295: 1351–1355.
- Prokosch HU, McDonald CJ. The effect of computer reminders on the quality of care and resource use. Chapter in hospital information systems; design and development characteristics; impact and future architecture. Edited by Prokosch HU, Dudeck J. 1995; III(13):221–239. Elsevier Publishing Co. Amsterdam.
- McDonald CJ, Hui SL, Smith DM, Tierney WM, Cohen SJ, Weinberger M. Reminders to physicians from an introspective computer medical record. Annals of Internal Medicine 1984; 27, 100: 130–138.
- Hripcsak G, Wigertz OB, Clayton P. Origins of the Arden Syntax. Artif Intell Med. 2018 Nov;92:7–9. doi: 10.1016/j.artmed.2015.05.006. Epub 2015 Jul 2.
- Indiana University Media Relations. Clarian Health to become Indiana University Health in spring 2011. 2010 May 5. Retrieved 2021 Oct 24. https://newsinfo.iu.edu/news/page/normal/14377.html
- McPhee SJ, Bird JA, Fordham D, Rodnick JE, Osborn EH. Promoting cancer prevention activities by primary care physicians: results of a randomized, controlled trial. JAMA 1991; 66:538–544.
- Rind DM, Safran C, Phillips RS, Wang Q, et al. Effect of computer-based alerts on the treatment and outcomes of hospitalized patients. Arch Intern Med 1994; 154:1511–1517.
- Pestotnik SL, Classen DC, Evans RS, Burke JP. Implementing antibiotic practice guidelines through computer-assisted decision support: clinical and financial outcomes. Ann Intern Med 1996; 124:884–890.
- Overhage JM, Tierney WM, McDonald CJ. Computer reminders to implement preventive care guidelines for hospital inpatients. Arch Internal Medicine 1996; 156:1551–1556.
- Overhage JM, Tierney WM, Zhou XH, McDonald CJ. A Randomized Trial of “Corollary Orders” to Prevent Errors of Omission. JAMIA 1997;4:364–375.
- Overhage JM, Tierney WM, Mamlin B, Warvel JA, Warvel JS, McDonald CJ. A tool for provider interaction during patient care: G-CARE. JAMIA 1996; 19th SCAMC Proceedings: 178–182.
- Mamlin BW, Overhage JM, Tierney W, Dexter P, McDonald CJ. Clinical decision support within the Regenstrief Medical Record System. Clinical decision support systems: Theory and practice, Second edition. 2007: 190–214.
- Overhage JM, Tierney WM, McDonald CJ. Design and implementation of the Indianapolis network for patient care and research. Bulletin of the Medical Library Association 1995; 83:48–56.
- Sideli RV, Friedman C. Validating patient names in an integrated clinical information system. SCAMC Proc 1992; 588–592.
- Park S. The Buck Stops Here. Cerner Perspectives. 2018 Sep 19. Retrieved 2021 Oct 24. https://www.cerner.com/perspectives/the-buck-stops-here
- Overhage JM. Indiana Health Information Exchange, Nationwide Health Information Network (NHIN) Trial Implementations, Task 9 Deliverable: Jurisdiction-Specific Business Plan, Contract No. HHSP23320074102EC. 2009 Jan 12. Retrieved 2021 Oct 24. Retrieved 2021 Oct 24. https://www.healthit.gov/sites/default/files/indiana_business_plan_nhin_final.pdf
- Regenstrief Institute Data Services. How we bring the data to you. 2020. Retrieved 2021 Oct 24. https://www.regenstrief.org/rds/data/
- Indiana Health Information Exchange. 2020. https://www.ihie.org/
- Takesue BY, Tierney WM, Embí PJ, Mamlin BW, Warvel JA, Litzelman DK. Regenstrief Teaching Electronic Medical Record (tEMR) platform: a novel tool for teaching and evaluating applied health information technology. 2021 Mar 17;4(1):ooab010. doi: 10.1093/jamiaopen/ooab010. eCollection 2021 Jan.
Submitted by J. S. Choi, MD.