Introduction

In the last decades human factors research has been acknowledged as crucial in the development of high risk medical equipment [1–3]. Infusion pumps are among the most frequently used medical devices, and are used for the controlled intravenous (IV) administration of many infusion fluids and drug solutions. Erroneous use of infusion pumps may lead to (temporary) under- or overdosing of vital drugs, with potentially severe consequences. The usability of infusion pumps has often been identified as an important contributor to the incidence of such medication errors [4–8]. Various studies using heuristic analysis have already pointed to various design flaws in pumps [7,9,10]. Factors such as inadequate training, workflow interruptions or high workload can further increase the likelihood of such errors [7,11,12]. In some studies where new user interfaces were designed and tested, the focus was on the usability of individual pumps [13–15]. However, this is a situation not representative of a critical care setting where patients are often treated with over six pumps simultaneously. Also, medication errors are more frequent in the critical care arena than in any other hospital department, while patients who are most severely ill may be even more vulnerable to such errors [11,16–19]. Hence, in the current proof of concept study we focused on the IV delivery system as a whole instead of focusing on individual pumps.

We designed a new user interface for the centralized monitoring and control of multiple infusion pumps by ICU nurses (S1 Fig). Actions that would normally be performed directly on the infusion pump’s user interface, such as changing an infusion rate, can now be performed on a new central bedside user interface. The central user interface forwards the appropriate commands to each individual pump under its control. We hypothesized that such a more user-friendly interface would reduce the risk of user errors and would also improve the workflow and user satisfaction compared to the conventional control of multiple infusion pumps. The objective of this study was to develop and evaluate the possible usability benefits of a bedside central control interface for multiple infusion pumps compared to conventional multi-pump operation. For this purpose we followed the ISO 9241–11 definition of usability: “The extent to which a product can be used by specified users to achieve specified goals with effectiveness, efficiency and satisfaction in a specified context of use” [20].

Materials and methods

Experimental tasks

Participants performed several typical tasks related to intravenous therapy in a simulated ICU setting. Examples of such tasks were changing administration rates, replacing syringes, administering boluses and navigating through menus. A between-subject design was used where participants performed the experiment using either conventional pump control or the centralized control interface. Each participant performed a set of thirteen pump-related tasks. Some tasks could only be performed when another was finished. For example, the volume to be infused (VTBI) functionality could only be turned off after it had been set up in the first place. However, such interdependent tasks did not necessarily follow each other directly; any number of other tasks could be scheduled in between. An overview of all experimental assignments is displayed in Table 1. Step-by-step workflows of each task type are included in S1 File. Note that some task types occurred more often than others in the experiment (change rate; replace syringe, setup and stop VTBI) in order to reflect their frequency in the real-world ICU environment.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Description of experimental assignments.

https://doi.org/10.1371/journal.pone.0183104.t001

Results

Table 2 displays the mean execution times, clicks and total number of errors per assignment type. The overall mean execution time (conventional vs. central mean ± SD, 406±119 vs. 421±108 seconds, not significant) did not differ between interfaces, whereas the overall mean number of clicks was lower using the central control interface (73±20 vs. 40±3 clicks, p < 0.001). Overall fewer errors were made using the central control interface (2.9±2.3 vs. 0.9±1.0 errors, p < 0.05). A description of all recorded errors is available in Table 1 in S1 File.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Execution times, clicks and errors per assignment type.

https://doi.org/10.1371/journal.pone.0183104.t002

Median ratings on questionnaire statements and Mann-Whitney U test results can be found in Table 2 in S1 File. Ratings on statements regarding the aesthetics, clarity, intuitiveness and ease to discriminate between pumps with a user interface (statements 2–5) differed significantly between conditions in favor of the central control interface. The mean ± SD rating calculated over all questionnaire statements was higher for the central control interface than the conventional interface (4.6±0.3 vs. 4.1±0.5, p = 0.03).

Discussion

We investigated whether a new central control interface would improve the overall usability of infusion pump control in a multi-infusion setting. Task-based usability analysis indicated that both objectively and subjectively the central control interface improved usability. Means calculated over all assignments indicated that participants required fewer clicks to perform the experimental tasks in the central control condition and also made fewer errors, indicating a more efficient and effective interaction. Questionnaires indicated that participants preferred the central control interface over the conventional interface, indicating greater user satisfaction.

There was no overall difference in execution times. This is remarkable since all participants were expert users of the conventional interface and had no prior experience with the central control interface. Moreover, it is likely that with further training the central control interface will outperform the conventional interface in terms of task execution times [25]. Data on specific execution times revealed that participants were able to change syringes 30 seconds faster on average using conventional pumps than using the central control interface. A difference between the two conditions is that the central control interface automatically stops and restarts the pump when a new syringe is placed. Although this feature was intended to improve the workflow, a relatively slow data connection between the central control interface and the infusion pumps limited the number of pump control commands to roughly 1 command per second. Hence, the pump’s RS232 communication layer was slowing down the task. As the communication protocol for the infusion pumps did not allow the automated confirmation of the syringe, the syringes had to be confirmed by manually pressing the confirm button on the pump instead of on the central control interface, which also impaired workflow. In case of a follow up of this study, a low latency communication layer between the pumps and interface will be required to optimize the task of replacing syringes.

In setting up a volume to be infused (VTBI) infusion on the pump participants were faster in the central control condition. The central control interface automatically calculated the required administration rate based on a preconfigured drug concentration, a target dose or volume and a predefined timeframe. In the conventional condition, the pumps offered no calculation support in the VTBI menu. Therefore, participants had to calculate which rate was required, which is a slow and error-prone task. Roughly half of the participants indicated that they did not regularly use the VTBI menu in the conventional interface, which may also explain that significantly more errors occurred during the VTBI tasks using the conventional interface. The difference in overall error rate appears to be mostly driven by errors in the VTBI tasks. The participants mentioned that the menu structure of the conventional interface was too complicated for the VTBI tasks. We believe this illustrates how counterintuitive design choices for the VTBI menu have contributed to poor usability and the occurrence of errors. This issue may be specific to the brand and model of infusion pump used in this study. A revision of the workflow of this particular VTBI menu should be considered if this proves to be an issue in real hospital environments as well.

In this study we compared the usability of a newly developed central control interface to that of a conventional pump setup. The results of this study indicate that this central control interface has a better overall usability than the conventional interface. However, this does not imply that central control in general will have a better usability than a conventional setup of individual pumps. In order to test for such a difference the layout and workflow of the central control interface should mimic that of the separate pumps. Such a setup will have very limited added value for a central control interface as it does not add any new features to the system as a whole. Added value may be gained by implementing sophisticated multi-pump profiles, for example, a system that switches from one pump to another when a syringe is almost empty or more complex multi-drug profiles (e.g. oncology treatment) [26]. Another extension could be early detection of IV line occlusions using combined pressure sensor readings from multiple pumps, which may reduce alarm fatigue by reducing the number of false alarms [27]. We believe that a well-designed central interface is complementary to a set of well-designed individual infusion pumps in a multi-infusion setting.

As a future extension of this study we propose testing with larger number of participants as well as ensuring a low-latency connection with the pumps. Testing with nurses who are not familiar with the conventional interface may reveal performance differences in other tasks than just the VTBI tasks. Including a longer training period with the central control interface may reveal differences in execution times as well. Although the current study did not focus on alarms, a future challenge for central pump control will be the channeling of alarm signals as existing visual and audible cues will remain important for the swift localization of the affected pump.

Conclusion

In this proof of concept study, we have shown how the usability of infusion pumps can affect the occurrence of errors related to intravenous therapy. A user-friendly user interface to control and monitor multiple infusion pumps was developed and its usability was compared to that of the current method of manually operating multiple infusion pumps in a task-based usability analysis. Results suggest that the centralized control system has an overall better usability and reduces the number of errors.

Supporting information