To the Editor:—
The use of “smart” intravenous infusion pumps incorporating microcomputer technology holds the promise of safer medication administration and is endorsed by ECRI (formerly the Emergency Care Research Institute).1A sophisticated feature of smart pumps is the medication library for particular patient types or care venues. Drugs in the library are given absolute (hard) or advisory (soft) preprogrammed dosing limits. The user selects the appropriate library, drug, and concentration, thereby invoking the limits for that medication. If a limit is breached, an alarm is both seen and heard. An “anesthesia mode” within each library allows prolonged pause, alarm management, and dose limits specific for the operating room.
After an intensive multidisciplinary study that included review of safety data, a return-on-investment analysis, a failure mode and effects analysis, and a usability trial, the University of Wisconsin Hospital and Clinics selected and implemented the Alaris Medley Medication Safety System intravenous pump (ALARIS Medical Systems, Inc., San Diego, CA) in October 2003. Before use in the operating room, training to highlight pump safety features, setup, programming, and capabilities was mandated for all anesthesia providers.
The failure mode and effects analysis team was aware, via an Internet discussion group and discussions with the manufacturer, of reports describing incorrect loading of the pumping segment of the Alaris intravenous tubing. Two types of misloads involving a hard plastic upper fitment were described. The first resulted from lifting the upper fitment as the pump door was closed, thereby stretching the silicone plastic pumping segment, typically causing an underinfusion. The second type of misload was less well understood and difficult to reproduce. It was thought to involve trapping the upper fitment in a tilted position as the pump door was closed. Because of these reports, preimplementation training specifically focused on correctly loading the upper fitment.
Three weeks after pump implementation, a 58-yr-old man presented for elective coronary revascularization as the first case of the day. Preoperative anesthesia equipment setup included Alaris intravenous pumps mounted at eye level to facilitate reading the programming screen. One tubing set was primed with nitroglycerin, the roller clamp was closed, the tubing was loaded, and the pump module door was closed and latched. The pump was turned on; the infusion was programmed and placed into prolonged pause as indicated by a yellow light at the top of the pumping module. The tubing was connected to a primed carrier fluid system that included a stopcock manifold (Merit Medical Systems, Inc. South Jordan, UT) and narrow bore extension tubing (Arrow International Inc., Reading, PA). Roller clamps and manifold stopcocks were opened to allow immediate initiation of therapy as necessary; however, the extension tubing that would eventually connect the manifold to the patient's central venous line remained clamped.
The patient was brought to the operating room. After induction of anesthesia and preparation for surgery, the extension tubing from the manifold was connected to the infusion port of the central venous catheter (MAC; Arrow International, Inc.). Later, the clamp on the extension tubing was opened to allow the slow infusion of carrier fluid via the manifold.
Almost immediately, and for no readily apparent reason, the patient's arterial pressure decreased and required repeated treatment with bolus administration of vasopressors by syringe. Transesophageal echocardiography revealed a marked decrease in left ventricular end-diastolic volume and function. The blood pressure recovered within minutes. Only then was the nitroglycerin bottle supplying the infusion pump found to be completely empty. A free-flow malfunction of the pump was suspected.
Close examination of the nitroglycerin pump module revealed a gap at the top of the door (fig. 1). The pump in question was removed and sequestered. Surgery proceeded without incident. Postoperatively, the patient was found to be neurologically at baseline and without untoward sequelae.
The sequestered pump was photographed, and the databases within the control and pump module were downloaded. Visual examination revealed that part of the intravenous tubing, the upper fitment, a molded hard plastic flange designed to be loaded from above into a recess, had been “front loaded” and held in place by the flange as the door was closed. The door closed sufficiently to latch and to open the flow-stop slide clamp below the pumping mechanism. The tubing flange held the door away from the reticulating finger pumping mechanism that normally sequentially occludes the tubing and controls flow. When the tubing is loaded as designed and the pump is off or paused, the fingers press the tubing against the door and completely prevent flow. In this event, the reticulating fingers could not reach the door, the tubing was not occluded, and free flow occurred.
Review of the downloaded pump databases revealed that during the setup, the pump alarmed twice before the audio and visual alarm indicators were cancelled when the pump was placed in prolonged pause. The alarm message displayed was “Fluid side occlusion.”
Three issues are particularly concerning. First, the failure mode and effects analysis conducted before the implementation of the pumps was lengthy and thorough but did not predict the failure mode causing the frank free flow we report. Second, the alarm message displayed during setup indicated an occlusion as opposed to a potential free flow, a message that did not alert the user to the fault. Finally, this event occurred despite intensive user training before implementation that emphasized correct upper fitment loading.
We believe other factors also contributed to this event. Because the pump was mounted at eye level, the door gap at the top was not visible. It is likely that time pressure, distraction from other setup activity in the operating room, and the practitioner's inexperience clinically with the new pump increased the likelihood of this event.
Clinical introduction of new products may result in unanticipated consequences despite preintroduction evaluation, institution-specific usability testing, and carefully planned user training.2Such training cannot be relied on to overcome design flaws in equipment.
This incident was reported through the US Food and Drug Administration reporting system. The manufacturer has since modified the pump module and error messages to reduce the risk of free flow from this cause.3
The authors thank Paul R. Malischke, B.S.E.E. (Anesthesia Equipment Manager, University of Wisconsin Hospital and Clinics, Madison, Wisconsin).
*University of Wisconsin School of Medicine and Public Health, Madison, Wisconsin. firstname.lastname@example.org