Houston, we (still) have a problem.
Every surgical patient and nearly every hospitalized patient receives medications as well as fluids via the intravenous (IV) route. Infusions are prone to error. For example, in a 2005 study of hospitalized Australian patients receiving continuous IV infusions, there was an 18% incidence of error.
I have been practicing anesthesiology for more than 30 years. In this time, I have watched the evolution of the technology to control the delivery of IV fluids and medications from passive mechanical devices (some may remember, often fondly, the Dial-a-Flow device) to simple electromechanical infusion devices, to today’s so-called “smart” infusion pumps. The pumps are called “smart” primarily because they incorporate “dose error reduction software” (or DERS) intended to identify drug dosages that fall outside individual facility defined best-practices ranges. The pump presents a warning and other information on its user interface (UI) screen when the clinician attempts to program a drug dose that falls outside the limits delineated in each hospital’s unique drug library.
Despite three decades of effort, infusion pump technology has not attained the level of safety expected or required of this life-critical technology. In 2012, the ECRI Institute identified infusion pumps as one of the most dangerous and hazard-prone technologies in healthcare today. A string of recalls by the Food and Drug Administration (FDA), affecting nearly every major infusion pump company and receiving ample public attention, further suggests that current pump technology does not meet the safety and usability we should expect of high-risk medical devices.
In a 2010 study, 73% of cases involving use of an IV pump had at least one dosing error. The FDA obtains more than 10,000 infusion device-related adverse event (AE) reports annually, yet we know that most events go unreported. In a 2011 study by Marla Husch, seven more pump-related infusion errors were observed in a single day than were reported over two years in that hospital’s incident reporting system. An ongoing study led by David Bates, MD, at Brigham and Women’s Hospital in Boston, is also looking the issue of smart pumps and medication errors.
Reported AEs vary in severity, but many include significant patient injury and even death. There are untold case reports in which clinicians made a simple usability error (e.g., misprogramming a pump to deliver a drug at ten times the intended rate) that caused irreparable patient harm. Unfortunately, the clinician is often blamed for these device-related errors. Yet, how can so many well-intended, well-trained clinicians make so many rueful errors, undeterred by the provision of technology (DERS) specifically designed to prevent them? The reality is that in addition to the technological (i.e., feature) deficiencies described in more detail below, the most prominent contributors to pump-related use errors are poor UI design in part related to a failure of vendors to employ a user-centered design (UCD) approach that incorporates a systems perspective and a thorough consideration of the myriad complex factors inherent in both routine and non-routine patient care.
What we have here is a failure to communicate.
In a study we did for the FDA more than a decade ago, we categorized IV medication errors reported by 49 inpatient pharmacists and critical care nurses during eight focus groups. For the 47 events described, the full implementation of “smart” pumps alone could only have prevented about one-quarter of the events. In another study of 4,600 patient-pump encounters, 100 preventable adverse drug events occurred, 29 of which involved overdoses. Only 4% of all events were intercepted by the “smart” pumps. A more recent study in a simulated inpatient unit performed by Trbovich, Easty and colleagues at the University of Toronto, similarly reported that the ability to remedy “wrong drug” errors between smart and traditional pump types was insignificant. Soft limit alerts had no impact on the performance of nurses to affect overdose errors. These findings strongly suggest that only a comprehensive solution involving the full integration of improved infusion technology with each facility’s medication management systems will be able to prevent most (but not all) medication errors. Medication management software includes computerized provider order entry (CPOE), pharmacy management, clinical medication list and medication reconciliation, and point-of-care administration review and documentation (e.g., bar-code medication administration (BCMA)) software.
Further, safety opportunities are lost because infusion technology does not reliably and contemporaneously ‘talk’ with continuous monitoring technology thereby allowing, for example, an infusion of blood pressure lowering medication to continue even after the patient’s blood pressure is dangerously low. (Some have argued that we do not have commercially available “closed-loop control” infusion technologies because: 1) the FDA is too risk-averse to approve them; and 2) the vendors are so risk-averse to develop them—both because of anticipated impediments to FDA approval and potential liability.)
While not a focus here, I will note that appreciable design improvements also will be required for medication management software (regardless of vendor) to attain the usability required to achieve high levels of safety.
It’s not a people problem, but a pump problem.
A 2013 survey of 400 nurses, interviewed one year after their facility implemented smart pump technology, found that 32% did not trust their new pumps any more than they had trusted their traditional “dumb” pumps. One reason for this lack of trust may be that, besides not living up to the hype, the new technology has created new types of errors that are more insidious in their presentation and in the ability of front-line clinicians to detect and mitigate them. Two examples: errors in the hospital’s drug library and excessive alarms leading to inefficiency and alarm fatigue.
Institutions are now forced to buy smart pumps because older simpler (and less expensive) technologies pumps are no longer available and a new standard of care stipulates the inclusion of safety software. Yet, in one hospital survey, most purchase decisions were unrelated to these safety features, and these hospitals did not use best practices during the implementation process. Further, smart pumps are still nearly universally regarded by clinicians as being difficult and inconvenient to use, and an impediment to, as often as a facilitator of, high-quality patient care. These issues, and especially their insufficient support of clinician workflow and workload leading to work inefficiencies, are at the core of this technology not being user centered. A nearly completely unaddressed aspect of clinician work related to infusion therapy, and fraught with inefficiency and potential error (not to mention air), is the management of the fluid “plumbing” going from the drug/fluid sources through the pumping mechanism and on to the patient.
Note that we are not just talking about nurses and anesthesiologists as pump users. Pharmacists will tell you that it requires excessive effort to create and maintain drug libraries while healthcare technology managers will expound on the undue effort required to locate, upgrade, and maintain the pump fleet.
In writing this piece, it dawned on me that the word “smart” has three distinct definitions. The people in marketing at the pump companies intended for it to mean intelligent. I guess their industrial designers might have thought of “smart” as referring to fashionable and elegant. However, for all too many clinicians, current “smart” pumps will occasionally sabotage our efforts to deliver safe effective care, thereby producing a sharp stinging pain.
Only when the issues delineated herein have been addressed will infusion pumps take on the intended meaning of “smart.” This will require a deliberate and rigorous application of UCD approaches that have been refined in the field of human factors over the last half-century. Then, not only can infusion pumps become the exemplar for other devices, but they also have the opportunity to become the definitive source of medication safety at the bedside.
Matthew B. Weinger, MD, MS, is vice president for research on the AAMI Board of Directors. He is the Norman Ty Smith chair in patient safety and medical simulation, and a professor of anesthesiology, biomedical informatics, and medical education at Vanderbilt University School of Medicine in Nashville, TN. Dr. Weinger has an equity stake in and consults for Ivenix Corp., a private start-up company that is building a next-generation infusion pump.