Key Points

Introduction

“Smart” infusion pumps, computerized provider order entry (CPOE), and bar coding of medication administration (BCMA) constitute the troika of medication safety-related health care information technology. Although CPOE is intended to optimize safe medication ordering and BCMA is intended to ensure that the drug and dose of medication match the intended patient, smart infusion pumps are intended to ensure that infused medications are delivered within a safe range of doses (if not the exact prescribed dose). Unfortunately, all three of these safety technologies merit the same criticism of that given an underachieving student: “They are definitely smart but not living up to their potential.” This article helps readers understand the realized and unrealized potential of smart infusion pumps, while exploring limitations and barriers of this technology in the larger scope of medication delivery systems.

Introduction

“Smart” infusion pumps, computerized provider order entry (CPOE), and bar coding of medication administration (BCMA) constitute the troika of medication safety-related health care information technology. Although CPOE is intended to optimize safe medication ordering and BCMA is intended to ensure that the drug and dose of medication match the intended patient, smart infusion pumps are intended to ensure that infused medications are delivered within a safe range of doses (if not the exact prescribed dose). Unfortunately, all three of these safety technologies merit the same criticism of that given an underachieving student: “They are definitely smart but not living up to their potential.” This article helps readers understand the realized and unrealized potential of smart infusion pumps, while exploring limitations and barriers of this technology in the larger scope of medication delivery systems.

What makes smart pumps so smart?

A skeptical reader might instead ask why these infusion pumps are even considered smart. In contrast to traditional infusion pumps, smart pumps are infusion devices that have a computer embedded within the device. These computers run vendor-specific software that is intended to reduce dosing errors through the use of drug-specific limits to allowable doses. Particularly in the context of pediatrics, these software libraries of medications and their specific doses are not provided as “ready to implement.” Instead, a hospital must create organization-specific (and often unit-specific) libraries, which list the different medications that might be infused, and the associated dosing limits. Additionally, organizations must decide if crossing a limit results in a warning or alert that can be overridden by the programmer (a “soft limit”) or absolutely prevents the programmer from entering the dose outside the predetermined limit (a “hard limit”).

Understanding smart pumps in the context of systems

Safe patient care increasingly is known to be an emergent property of systems. Whether patient care is safe or not results from the interactions and properties of individual systems elements and not simply the elements themselves. One model for thinking about patient safety is the Systems Engineering Initiative for Patient Safety model.

This model consists of five interacting components ( Fig. 1 ). The first component is people: this component includes patients, their families, clinicians, and other support staff working in the environment of interest. Importantly, each person brings to the system his or her own culture including culture of safety. The people use the second component, tools and technologies. These tools and technologies can range from pens and paper documents to CPOE and smart pumps. The people use the tools and technology to perform tasks, the third component of a system. The tasks include gathering data, assessing patients, prescribing medications, dispensing medications, and programming of an infusion pump. These three components occur and interact within an environment, the fourth component, which has its own impact in the form of noise, temperature, lighting, and physical layout. Finally, all of the previously components interact within the larger context of an organization. The component of an organization includes leadership behaviors, policies and procedures, purchasing decisions, and the organization’s culture of safety. The presence or absence of safety is an emergent property of the system: it is more than the sum of the individual components. As a result, any change to a system, such as the introduction of new infusion pump technology, inevitably impacts every other component and the resultant safety of the system.

Adaptation of Systems Engineering Initiative for Patient Safety model to smart pump technology illustrating the systems nature of medication delivery with pumps and the implications for patient safety.

When smart pumps are framed using this model, their strengths and limitations become more transparent. Although the pumps have the potential to limit programming errors, they depend on how people interact with them. However, how people interact with the pumps is subject to other system components. These include other technologies used for patient care; other people involved in the care of patients (directly or indirectly); the culture of the individuals and the organization in which they work; environmental factors (eg, noise and lighting); and organizational decisions, such as how the pumps are implemented, training, and decisions on the drug libraries.

Just how smart are these pumps?

The principle goal of smart pumps is to prevent dosing errors when administering medications. Fanikos and colleagues analyzed alerts from smart pumps used for delivering anticoagulants to adult patients and found an association between the use of smart pumps and fewer infusion rate errors. Although the authors perhaps overstated the association as being a causative relationship, particularly because not all anticoagulant infusions necessarily used the libraries and thus may have missed inclusion, this study is notable for showing a potential reduction in errors as a result of this technology.

Another study of 3 days of smart pump data from a Canadian adult heart hospital revealed that the pumps intercepted 14 infusion errors. An additional 67 overridden alerts were captured in the same time period. Perhaps most intriguing was the observation that of 1306 infusions started during this time period, 973 (74.5%) of the medications were infused in a mode that bypassed the smart functionality. This is consistent with findings by Malashock and colleagues, who studied alerts in smart pumps from three subspecialty units and found evidence of 157 alerts that led to reprogramming of the pumps. The authors provided no denominator of number of unique infusions. An additional 696 alerts were reported as overridden and, without clear evidence for the conclusion, attributed this to medication boluses or loads.

Two other publications draw unsubstantiated conclusions about the impact of smart pumps. In a 2003 newsletter of the Anesthesia Patient Safety Foundation, Reves cites unpublished data to state “smart pump” technology reduces errors. One source of Reves’ data was an award winning presentation by Kinneally. The Kinneally data reflected a drop in the number of voluntarily reported administration errors from the pre–smart pump to post–smart pump time period. Unfortunately, Reves ⁶ later observes, “While the full impact of the new “smart pumps” on the reduction of intravenous medication errors has yet to be measured, the preliminary data are very encouraging.”

Similarly, Rosenkoetter and colleagues reach several disturbingly unsupported conclusions about smart pump use. Drawing from surveys of nurses and pharmacists, the authors state in their abstract that “the pump offers an effective approach to the reduction of intravenous medication errors,” and that “use of smart pumps had no effect on use of pharmacy staff and did not negatively affect staff nursing work load.” The authors, however, neither measured intravenous medication errors nor pharmacy or staff workload. Instead, they cited survey data about perceptions. The sole evidence supportive of error reduction was a reference to work by Wilson and Sullivan that measured 1 prevented error among 80 analyzed infusions.

Finally, Larsen and coworkers reported a drop in “reported errors” after three interventions: (1) the implementation of standard drug concentrations, (2) smart pump technology, and (3) user-friendly medication labels. This was despite the absence of any direct error measurement and the fact that the reduction in reported errors might be explained by the possibility that nurses were too busy responding to these three system changes to report errors that occurred.

In summary, the available evidence shows minimal to moderate error reduction as a result of smart pump technology and no measured evidence of a reduction in patient harm.

Are smart pumps living up to their potential?

The literature on smart infusion pumps is particularly enlightening on the aforementioned charge of “not living up to their potential.” A 2006 study observed the infusions of 426 medications and identified 389 different errors. Of these, only one error would have been prevented by smart pump technology without the use of additional interface technology that was not available at that time. This finding is consistent with that of Nuckols and colleagues, who found that of 100 identified intravenous adverse drug events that caused harm, only 4% were potentially addressed by available smart pump technology.

Using high-fidelity simulation, Trbovich and colleagues observed that the use of soft limits with smart pumps “had no significant effect in preventing dosing errors.” This simulated study reached similar conclusions to a controlled trial of smart pumps used with adult cardiac patients, which found that although errors occurred frequently, the smart pumps had no measurable effect on the rate of serious medication errors.

Finally, Schroeder and colleagues published a chilling case report of a tubing misload resulting in free-flow administration of nitroglycerin administered with a smart pump. Ironically, the same technology had undergone proactive risk assessment using a failure modes and effects analysis. Although the manufacture has addressed the design defect that resulted in the event, the documentation of preventable harm (by a pump manufacturer) in the absence of error prevention or harm reduction data is concerning.

What is the cost of these pumps?

In 2005, smart pumps were described as attractive because of the cost and speed of installation in a facility. At that time, implementation was reported to be as quick as 90 days for a facility with a cost from $2 to $3.5 million for 1000 pumps. These cited costs are strictly that for hardware and software. There are additional hidden costs related to training staff, server and networking, maintenance, and support. Pediatric facilities are faced with even greater hidden costs associated with the time required to develop context-appropriate drug libraries for installation. Whether or not the technology is viewed as attractive depends on the actual cost of all implementation and ongoing maintenance.

Even then, a published analysis of alerts during the first 4 months of smart pump implementation in a pediatric intensive care unit identified numerous problems that necessitated retraining of nursing staff, and changes to previously selected limits because of clinical alert irrelevance and errors in library programming. Readers should not underestimate the cost associated with changes to a smart pump library. In the absence of additional network technology, which is not standard on smart pumps, changing the library (be it additions, corrections, or deletions) requires manual interaction with every pump in an organization. This is no small task. In one free-standing children’s hospital, each manual upgrade was associated with an operational cost estimated at $25,000 and requiring 3 days (Brian Jacobs, personal communication, 2012). Additionally, infusion pumps that are used to provide life-sustaining vasopressors and inotropes cannot be updated until the pump is no longer in active use with a patient. After a center implements pump networking, the pumps can be upgraded as soon as the pumps are powered on, as long as the end-user accepts the update and the pump is not in active use.

Thus, the true cost of ownership (implementation and maintenance) is difficult to quantify and arguably greater at a per-pump level for pediatric centers than adult centers, which can standardize libraries and have fewer weight-based infusion types.