Clinical decision support systems: the fascination with closed-loop control.

To the engineer a closed-loop control system refers to using a controller with a feedback loop to control states or outputs of a dynamic system. That dynamic system might be the cruise (speed) control on your automobile, maintaining a patient’s paO 2 with a ventilator, controlling depth of anesthesia with an anesthesia machine, or controlling arterial blood pressure with medications from an IV pump. To a cowboy from the Wild West a closed “loop” might be considered a lariat or lasso rope, used to rope cattle. Similarly, for some in medical informatics, closed-loop control means surrounding a medical problem area to better manage it. I suggest that, in fact, there is a broad range of Clinical Decision Support (CDS) systems and that closed-loop controllers represent only one specif ic type. CDS systems include diagnostic systems, intelligent Computerized Physician Order Entry (CPOE), alarming and alerting systems, care advice systems, etc. These CDS systems can be used to help with tasks such as optimizing blood ordering, medication selection and dosing, and using protocols to improve and optimize patient care. Since the 1950’s, when physicians began to understand control system theory, there has been a fascination with having these control systems be closedloop without the need for any human intervention. Bickford at the Mayo Clinic was one of the f irst to suggest closed-loop anesthesia, using the electroencephalogram (EEG) as the measure [1]. In 1968 Sheppard and his colleagues, at the University of Alabama, showed that the blood pressure of post operative open-heart patient’s blood pressures could be controlled with closed-loop strategies [2]. Since those early times, when the use of closed-loop systems was the ultimate decision support strategy, there have been a myriad of other decision support methods developed. However, there still seems to be some magic in being able to “close the loop.” Just to give an update on how we are doing some 50 to 60 years later with implementing closed-loop systems, articles in this Yearbook have addressed the topic [3, 4]. Recently, considerable literature on the clinical implementation of closed-loop control has begun to appear. Two examples: a) the accuracy and clinical feasibility closed-loop control of the hypnotic component of anesthesia was recently reported [5]; b) in addition, news reports indicate that closed-loop control of insulin administration for diabetics, the so-called “artif icial pancreas,” is coming closer to reality [6]. As I was preparing this Keynote manuscript, I began to ask myself why hadn’t our group at LDS Hospital in Salt Lake City developed and published ANY work on operational closed-loop control systems. After all, our entire group has hundreds of publications, and most of them are in the f ield of decision support and Medical Informatics. So, I took the opportunity to refresh my memory on the almost 50 years of Medical Informatics history that I have lived through. It quickly became apparent that as a consequence of working at LDS Hospital with our mentor Homer R. Warner [7] and colleagues T. Allan Pryor, R. Scott Evans, Paul D. Clayton, multiple graduate students and other professional colleagues, we