History of decision support
This is a timeline of the early history of clinical decision support, from 1959 through 1993. Progress has been made since this time, particularly in the area of standards.
1959: Ledley and Lusted propose a mathematical model for diagnosis in their article "Reasoning foundations of medical diagnosis; symbolic logic, probability, and value theory aid our understanding of how physicians reason", published in Science. This article has been called the first work in medical informatics.
1961: Homer Warner, of the University of Utah, develops a mathematical model for diagnosing congenitial heart disease. Their approach uses a contingency table with diagnoses as rows and symptoms as columns. The system predicts the diagnosis with the highest conditional probability given a set of symptoms.
1964: Morris Collen of Kaiser develops a system for automated multiphasic diagnosis. Patients were given a stack of computer punch cards containing symptoms and questions, and they sorted them into Yes and No piles. A computer used these cards to develop a preliminary diagnosis.
1969: Howard Bleich develops a system to suggest therapy for acid-base disorders. It is the first decision support system which proposes a management plan in addition to a diagnosis.
1972: F.T. de Dombal builds a probabilistic model to diagnose abdominal complaints. The computer predicted the correct surgical diagnosis based on initial findings 91.8% of the time. On average, senior clinician predicted the diagnosis correctly 79.6% of the time.
1972: The HELP system, developed at the University of Utah is enhanced with HCOM, a compiler that allowed Bayesian decision support modules to be developed. The HELP system was expanded many times, and formed the basis of many research projects in clinical decision support, including the COMPAS ventilator management system by Dean Sittig and an antibiotic advisor by Scott Evans.
1975: Ted Shortliffe of Stanford develops MYCIN, an expert system for antibiotic dosing. MYCIN consisted of three modules: a consultation system which collects information, applies the rules in its knowledge base, and recommends therapy; an explainer system, which explains these recommendations; and a rule acquisition syhstem, which an expert uses to build rules for the knowledge base. In early evaluation, Mycin suggested acceptable therapy 75% of the time, but it got better as more rules were added.
1976: Clem McDonald publishes "Protocol-based computer reminders, the quality of care and the non-perfectability of man", a groundbreaking work in the history of clinical decision support, in the New England Journal of Medicine. The article looks at Alerts and Reminders in the Regenstrief Medical Record System. In an experimental trial, McDonald found that physicians responded to 51% of the alerts they received, but provided the corresponding care only 21% of the time when alerts were not provided.
1981: INTERNIST-I is developed by Randy Miller, Harry Pople and Jack Meyers. INTERNEST, a diagnostic decision support system, was the first decision support system to span all of internal medicine. Users would enter the findings for a case into INTERNIST, and the system would develop a differential diagnosis, and ask questions to improve this diffential. The system contained 500 disease profiles, and 3,550 clinical manifestations.
1983: Perry Miller develops the ATTENDING system for anesthesia management. ATTENDING is the first medical critiquing system.
1987: Octo Barnett develps DXPlain, a diagnostic decision support system similar to INTERNIST. A web version of DXPlain is still available today.
1993: Physician order entry and clinical decision support becomes available at the Brigham and Women's Hospital with the release of the Brigham Integrated Computer System (BICS), developed by Jonathan Teich. BICS is a model platform, and has been widely studied and evaluated.