Objective To decrease interruptions around a centrally-located, centralised, open paediatric medication station.
Methods Several established human factors methodologies were used to study paediatric medication administration, including cases with ‘walk through’ and verbal protocols; semi-structured interviews, including critical incident analysis; hierarchical task analysis; and observation.
Results Inexpensive barriers were constructed that protected the tasks likely to lead to errors if interrupted. Meanwhile, sight lines were maintained preserving a family-friendly sense of accessibility of nurses, staff situation awareness and collegiality. Interruptions were significantly reduced and staff attitudes towards the station were significantly improved.
Discussion Targeted barriers may prove useful in other interruptive and chaotic hospital workspaces. They do not require costly training, can be achieved inexpensively and may reduce distractions and interruptions during tasks vulnerable to error. Additionally, the human factors methodologies employed can be applied to other safety improvement projects.
- Human factors
- medication safety
- quality improvement
- patient safety
Statistics from Altmetric.com
The Institute of Medicine has called for better design of work processes to improve patient care and safety. ‘Poor designs set up the workforce to fail, regardless of how hard they try.’1 In 2005, the National Academies of Medicine partnered with the Institute of Medicine to explore engineering applications that could significantly improve healthcare. They indicate that systems engineering, and its subspecialty of human factors, provides analytic tools that can clarify work processes and suggest improvements.2 Human factors analyses seek to explain interactions among humans and other elements of a system, and theoretical principles, data and methods are applied in order to optimise human well-being and overall system performance.3 This study applies human factors techniques to address the problematic issue of interruptions during preparation of paediatric medications through the design of an environmental intervention.
Interruptions and distractions in healthcare have been studied across various hospital settings.4–8 It is well recognised that hospital environments are busy and highly interruptive.9 ,10 Interruptions during task performance lead to errors, particularly errors of omission.11 Westbrook et al studied medication administration and found an increase in procedural and clinical errors associated with the number of interruptions.12 Paediatric medication errors are particularly concerning because they are estimated to be three times more likely to result in significant harm than adult medication errors.13
This study was undertaken in a paediatric acute ward in an academic referral Children's Hospital. Nurses are responsible for checking, preparing and administering their assigned patients' medications. Information necessary for medication preparation and checking is paper based despite the use of a computerised practitioner order entry (CPOE) system by the hospital. At the start of each shift, paper orders for each patient are printed and placed in a counter-top filing system (Kardex (Westbrook, ME, USA)). The Kardex provides access to paper documentation related to a patient's care but may only be used by one person at a time. When the Kardex is in use, and another nurse needs to prepare medications for administration, they sometimes take the paper record from the Kardex to another counter in the medication station. Medications are kept in three places within the medication station depending on the type of drug: in drawers below the Kardex, in a refrigerator or in an electronic dispensing machine (Pyxis (San Diego, CA, USA)) (see figure 1).
The medication station is a centrally-located, centralised open design (see figures 1 and 2), a common design that offers many benefits. Centrally-located medication stations are intended to minimise the amount of walking required of nurses, thus maximising time spent with patients.14 They also offer convenience for pharmacy delivery and tracking of controlled substance usage. Open stations, not enclosed by walls, allow continual observation of patients and monitors. However, a major disadvantage of this design is that it leaves nurses open to frequent interruptions.
Several incident reports at the hospital under study cited interruptions during medication administration as a root cause of medication errors or near misses. Because medication preparation is considered a critical activity, it seemed appropriate to apply a sterile cockpit rule: an aviation regulation that specifically prohibits crewmembers from non-essential activities (eg, chatting) while performing critical activities (eg, take-off and landing).15 Other institutions have attempted to reduce disruptions during medication administration by having nurses don accessories, such as a specially coloured vest, to ward off interruptions:16 this was deemed impractical at our institution given the dynamic and rapid nature of nursing work. Signage has also been used,17 ,18 but this requires staff to read and heed a sign that may not be visible from every angle. Further, such ‘policy’ interventions are inevitably of low impact. They are often ignored, forgotten, or unknown and require frequent retraining.19 They may be coupled with sanctions for non-compliance.
The nurses requested that preparation of medications be moved to an enclosed private room. This would be an extremely costly change, and it was unclear whether such a change might have negative consequences due to nurses being less ‘available’. A second request was that the centralised medication station be enclosed with walls but there was concern this would take away from the ‘family-friendly’ ethos of the ward, a core value that the staff did not want to diminish.
The objective was to design an inexpensive, environmental intervention that would naturally reduce interruptions without introducing new error-prone processes.
The research team consisted of two paediatric physicians, one paediatric nurse and a human factors expert. The research proceeded in three phases: (1) evaluation of the current medication station and the tasks required for safe medication preparation; (2) design of an intervention; and (3) collection of preintervention and postintervention data to determine the effectiveness of the intervention. Appropriate ethics approval was obtained. Participant nurses signed consents and their anonymity was protected.
Phase one: evaluation of tasks and current medication station
A number of human factors techniques were used to gain an understanding of how nurses administer medications on this paediatric ward.
First, a simulated medication preparation task was performed. Using actual medical records, the research team developed two use cases depicting medication orders for a straightforward patient and a complex patient. Two nurses piloted the cases to ensure their accuracy and representativeness. We asked two senior nurses, who had not previously reviewed the cases, to ‘walkthrough’ administration of medications for the two patients and to ‘talk aloud’ while doing so. As the nurses talked aloud, they verbalised (‘verbal protocols’) their thoughts in the form of a commentary about their actions and their immediate perceptions of the reasons behind them.20 We listened, observed and tracked their movements as they worked through the task.
Next, we recruited three senior and three junior nurses to participate in semistructured interviews. Questions during a private single session focused on their work habits during medication administration and their procedure for checking the 5Rs (right patient, right medication, right dose, right time, right route). Participants were also interviewed about critical incidents. Specifically, they were asked to recall a medication error or near miss, what the causes of the error were, what they thought at the time and how they reacted. The premise of critical incident technique is that such incidents will be memorable to participants. The technique specifically attempts to get at off-nominal and rare events.21
We then conducted several days' worth of observations of nurses preparing medications during the peak administration hours of 10:00–11:00. The researchers stood 6–8 feet away from the Kardex and had a clear view of activities. They took field notes in a notebook and neither interacted with the nurses nor examined patient records or medications prepared. This type of naturalistic observation allows the researcher to better understand the context of the work, see things that may escape the awareness of the subjects and discover things no one has really paid attention to.22
The interviews, case studies and observations enabled us to gain an appreciation of nurses' medication preparation activities and to represent them in the form of a hierarchical task analysis (HTA) (see figure 3). HTA results in a hierarchical diagram that organises human work by goal. High-level goals are achieved by carrying out a number of subgoals, and so dependencies are represented in the hierarchical structure. This type of representation differs from the sequential flow diagram, or process map, and prompts the researcher to establish the conditions required for the performance of subtasks that must be completed in order to fulfil a system's goal.23
Although nurses have developed different strategies to execute the subtasks of medication administration, four subtasks were consistent across nurses to meet the high-level goal of safe provision of paediatric medications (see figure 3).
This check is intended to identify errors in the medication as ordered. Using their clinical assessment and reasoning skills, nurses check the appropriateness of the medication ordered. They also check the dose as detailed on the paper record in the Kardex against accepted reference standards using the Pediatric Dosing Handbook (PDH). Correct weight adjustment is confirmed by using a calculator, the routing and compatibility of the medication against other medications the patient is taking is checked against the PDH, and the timing of the medication is checked against the PDH and the paper medication administration record.
Collection of medications
Nurses physically gather the medications from three storage locations within the medication station.
Preparation of medications
Doses dispensed in multi-dose containers require specific measurement and preparation before they can be given to the patient. For example, pills might need to be counted out or syrup might need to be measured into an oral syringe.
This step again checks the 5Rs. Each medication as collected and prepared is checked against the paper record in the Kardex. Errors in dispensing, collecting or preparing are intended to be caught at this time.
The anchoring role of the Kardex became apparent through motion tracking of the nurses as they performed the use case simulation. Almost every task that was performed during medication administration was followed by a return to the Kardex (see dotted lines in figure 1). The Kardex, as well as the tools in close proximity (a calculator and the PDH), were used by nurses to check patient information and to prepare and check medications. Nurses described needing to concentrate when they performed tasks using these tools.
The HTA identified subtasks and the conditions required for their successful execution. Consideration of the subtasks and the required conditions for successful execution of these subtasks supports identification of steps at greater risk of failure in the face of interruptions. For example, tasks at the Pyxis are less error prone because these tasks do not require any moving, or looking back and forth. If task performance is interrupted, the Pyxis provides visual cues to the nurse (ie, electronic prompts) that remind them what subtask they were in the midst of performing. Thus, they are less likely to miss or repeat a step.
On the other hand, the two checking tasks require concentration and require working memory as the nurse moves from calculator to Kardex to PDH, back to Kardex, etc. If nurses are interrupted during these tasks, it is difficult for them to remember what stage of checking they have completed. They may inadvertently omit an important step, or restart the entire check from the beginning, which is inefficient. Thus, when interrupted, the likelihood of a nurse making an error during these cognitively-loaded checking tasks is more likely than during other tasks that are supported by physical cues, like the electronic prompts during tasks performed at the Pyxis.
Moreover, analysis of each checking subtask's failure modes indicates that failures may be severe and less likely to be recovered. Interruptions in either checking phase (dose or final check) could allow an incorrect order to be missed, an incorrectly dispensed medication to be missed or a medication to be missed completely.
Although interviews suggested that nurses always concentrated during dose and final checking and that they blocked interruptions, observation indicated that they frequently engaged with interruptions during these checks. Thus, their conception of what happened reflected the normative state (what should happen) rather than the actual state.
Direct observation also indicated that the medication station was an important gathering place for staff. The natural congregation of nurses in the medication station supports beneficial collaboration, cross-checking and teaching, but also may lead to interruptions.
The nurses' ability to view the floor outside the medication station proved necessary. First, the nurses frequently glanced at the whiteboard that listed patient room and staff assignments. Interviews indicated that the nurses were using the whiteboard to prompt their prospective memory (remembering to perform an action in the future) as they worked at the Kardex. The importance of the whiteboard to clinical staff has been previously recognised.24–26 Nurses also had to have a clear view of patient monitors in order to respond to alarms.
Views from the ward to the medication station were also important. The medication station could be viewed from almost every patient's doorway. Consequently, parents and staff could always find a nurse quickly and have their needs addressed (usually interrupting the nurse in the medication station).
Finally, space in the medication station was at a premium and nurses would often wait to perform their medication preparation if they saw that the space was busy. This helped the nurses to efficiently manage their ‘stacked’ tasks. Stacking of tasks is when a nurse moves on to another task to prevent ‘down time’ when unable to complete the first task because of waiting for processes or an inability to access resources.27 ,28 In this case, instead of waiting in a queue for the Kardex, they would complete other stacked tasks until the Kardex was free.
It was also noted that occupancy of the medication station was reduced by a worker's ability to pass items into the station, over the wall, rather than entering the station. For example, the ward secretary could file papers in the Kardex without entering the station; this helped reduce crowding in the interior working space.
Phase two: design of the intervention
Results of the use cases and HTA identified that checking tasks (dose and final checks) required more cognitive work. Evaluation of failures of subtasks in the HTA indicated that errors in these steps could have serious consequences. Observation also indicated that despite frequent interruptions, there were benefits to the open design, such as nursing collaboration, communication with the Ward secretary, ability for nurses to view the whiteboard and alarm stations, and availability to parents. Thus, the existing suggestions of moving the medication station to a separate room or enclosing the current station with walls were predicted to cause new errors by preventing appropriate sight lines and communications. Due to the holistic view of the focused work required for dose and final checks and the medication station, it was determined that a less expensive, more targeted intervention would likely work (see figures 4 and 5).
Consequently, the design solution targeted placement of 24″ height walls around specific areas of vulnerability: the corner of the medication station surrounding the Kardex and the opposite counter used when the Kardex is busy. The end points of the barriers were dictated by the nurses' need to see the white board, monitors and ward secretary. Sight lines across the floor to patient rooms and staff workspaces were also preserved (see figure 1). A 3″ gap was left at the bottom of the barriers to enable full usage of the counter space, to accommodate the height of the back of the Kardex when in use, and to enable passing of papers and small items when necessary.
Tension between completely blocking the dose and final checks and allowing nurses to be seen by staff and families was resolved by specifying that the barriers be made with frosted glass. The nurses' view out of the medication station would be blunted at these locations (limiting distractions), but an outside observer would still be able to see that a nurse was present. It was predicted that the barrier around these places in the medication station would naturally reduce interruptions: it would signal that tasks being performed at these locations were important and would thus cause hesitation on the part of the interrupter without any need for training or signage.
Phase three: evaluation of intervention's effectiveness
In order to measure the effect of the intervention, two sets of data (nursing perceptions and frequency of interruptions) were collected approximately 6 months preintervention and 6 months postintervention.
Nursing perceptions of their work in the medication station
Nursing attitudes were measured using a paper questionnaire that was placed in staff mailboxes (see table 1 for the statements), accompanied by an email to notify nurses of the survey and request they complete it. The questionnaire gathered demographics (time worked on paediatric ward and how many hours per week) first and then asked for ratings of perceptions of their work in the medication station on a 7-point Likert scale of 1: strongly agree to 7: strongly disagree.
Frequency of interruptions and distribution of interruption type
Trained researchers watched, counted and classified interruptions of nurses preparing medications during the peak medication administration hours of 10:00–11:00. Days with patient occupancy below 90% and days with nursing students were excluded in order to reduce variability in nurses' activity level and behaviour. The medication station was also videotaped during these times. Half of the observations were made with two researchers when the academic calendar allowed. The other half of the observations were performed by a single researcher; the single or double observations were distributed equally across preintervention and postintervention observations. The purpose of periodically including a second researcher was to be able to report inter-rater agreement as a consistency check of findings.
A total of 20 h of observation data were collected preintervention and 20 h postintervention to determine rates of interruption. During each hour of observation, the length of time for each medication occurrence from the beginning to the end was collected (see table 2). During the occurrence, interruptions were counted. Interruptions that occurred when a nurse was in the medication station, but was not engaged in a medication administration occurrence, were excluded.
Content of the interruption was recorded in field notes for later coding. Immediately after each observation session, the researchers reviewed the videotape to confirm their field notes and to code each interruption as self versus other and direct versus indirect patient care (see table 2). There was 100% agreement between coding of self and other interruptions. There was 86% agreement between coding of direct and indirect patient care (κ=0.70). Disputes were decided by the senior researcher (LC). Occurrences in which the content of the interruptions could not be overheard were eliminated from content coding (eight preintervention and two postintervention) but were included in the number of interruptions.
Nursing perceptions of their work in the medication station
Twenty surveys were returned preintervention and postintervention (38% response rate, 20/52). The demographics of the groups were similar and likely included some overlap in respondents although anonymity precluded their matching.
The Mann–Whitney U test was used to analyse the difference in Likert scale results. Results showed a statistically significant shift towards favourable opinions of the medication station in all categories except efficiency and the sense that medication errors are unlikely (see table 1). The authors suspect that nurses are always efficient in managing their work and that the visibility of workers in that station both preintervention and postintervention supports productive task juggling. Interestingly, despite recognised problems with the original medication station, nurses thought they compensated and kept medication administration safe; previous research suggests this may be overconfidence.12–14
Frequency of interruptions
Observation data were collected preintervention and postintervention to determine rates of interruption during medication administration occurrence (measured from opening of the Kardex to leaving the medication station, see table 2). Data were obtained in single hour blocks for a total of 20 h preintervention and postintervention. Number of interruptions ranged from 0 to 12 per occurrence. Length of occurrences ranged from 20 s to 720 s (12 min). The mean interruption rate per minute of occurrence was significantly reduced from 1.4 preintervention to 0.27 postintervention (paired t test=5.7, df=98, p<0.01). The length of time spent from beginning to end of each occurrence was not significantly reduced (120 s vs 117 s preintervention and postintervention, respectively).
The distribution of types of interruptions (self vs other, patient care vs indirect patient care) among the total number of interruptions did not vary significantly preintervention versus postintervention as determined by χ2 tests for independence (see table 3).
Using human factors techniques, paediatric medication preparation in a centrally-located, centralised open medication station was analysed. Findings indicated that dose and final checking subtasks were more likely to lead to adverse events if the tasks were interrupted due to their cognitive nature. The importance of the open nature of the station in terms of necessary sight lines, collaboration between and availability of nurses to families precluded enclosure or moving of the entire station although these were the suggestions made by nursing staff prior to the more detailed analyses conducted here. Thus, barriers were targeted to specifically protect the location at which these tasks were performed while preserving important lines of sight. The design enabled family and staff to locate nurses, maintaining an open, ‘family-centred’ and ‘staff available’ feel to the ward. This solution had the additional benefit of being of a much lower cost than the suggested interventions prior to the human factors analysis. Data collected prior to the introduction and 6 months after the intervention was installed showed a significant reduction in the rate of interruptions, and nurses' perceptions of the medication station were significantly improved. No training was required to ‘teach’ nurses, other staff, patients and families about the importance of protecting nurses' concentration while preparing medications.
Although our work focused specifically on a medication station, targeted barriers could be used wherever healthcare workers experience high mental workload. The physical structure visually reminds both the worker and others that work done behind the barrier is important and should not be interrupted except for urgent direct patient care. Thus, the barriers alone constitute a sustainable and permanent interruption deterrent; there is no need for a costly education and training programme. The design of a ‘natural’ intervention such as this follows the principle of designing using the principle of ‘affordances’.30 For example, a countertop ‘affords’ the placement of items: it naturally indicates to humans that it is a sufficiently strong and flat surface to hold things. Similarly, see-through but frosted vertical walls naturally indicate ‘I am here, but I am busy’, and likely cause a moment's hesitation and a more conscious choice as to whether it is worth interrupting the concentrated work that is occurring there. Such a barrier also naturally cuts down on the noticeability of activities occurring on the other side, thus reducing distractions. Having a limited barrier also enables the worker to easily glance around the barrier when needed, enabling open communication and movement of objects within the environment (see figures 4 and 5).
There are several limitations to this study. First, participant behaviour during interviews and observation is subject to the Hawthorne effect: behaviour changes due to the act of being observed. However, we made observations over a long period of time (an hour on intermittent days over several months) with the aim of diminishing any sustained Hawthorne effect. Further, we performed similar observations preintervention and postintervention and only measured differences, so that any such effects would be equally likely to occur and thus accounted for in the analysis.
Second, our study focused on reducing interruptions during medication administration, but we cannot extrapolate this to an absolute reduction of medication error. Although previous research, root cause analyses and results from the critical incident technique suggest that interruptions can cause errors, reverse causality cannot be determined by this work. The only longitudinal measure of medication errors or near-misses at our institution is incident reports. Self-reported incident reports are recognised as under-representing actual errors.31 ,32 At our institution, frequency of such reports may also be confounded by changes in administrative attention to and support of self-reporting. Consequently, rates of reported errors are not the best proxy for actual errors.
Third, coding of the content of the interruptions was subjective (patient care/indirect patient care). We did not validate our coding decisions with the nurses as that would have amplified the Hawthorne effect. However, the coding definitions were consistently applied across preintervention and postintervention observation sessions. Further, as a validity check, for half of the observations, a second researcher was employed to enable cross checking of codes.
The results of the preintervention and postintervention questionnaires were limited by the incomplete matching of participants. Although we suspected most of the responders were the same, we could not assure this without breaking their anonymity.
This paper attempts to show the benefit of using human factors engineering to help design effective interventions. Quality improvement efforts often institute new procedures or policies that require staff to take extra steps (eg, ‘put on a special vest so that people know that you are busy and should not be interrupted’). Implementation of such policies requires expensive training, frequent reminders and sanctions for non-compliance. Consequently, overworked staff try to meet multiple competing objectives, thus setting themselves up for failure (‘Should I take extra time to don a vest or should I use that time to double-check my medications?’). In this study, a well-intentioned design proposed before a human factors evaluation, such as putting barriers up around the medication station, might have inadvertently introduced new forms of error. For example, the proposed design may have made it impossible to simultaneously perform essential concurrent tasks, such as monitor alarms or see the whiteboard. A reasoned design based on detailed analysis of the cognitive and physical requirements of the task led to solutions that avoided these unanticipated pitfalls. A human factors based solution was able to provide a design that improved nurses' perception of the workspace and reduced interruptions, without overt evidence of new problems. In this case, a straightforward application of human factors techniques led to a solution that is of a lower cost and which naturally affords ‘correct’ behaviour.
The authors would like to thank Gerard P Learmonth Sr., PhD, for his help with statistical analysis. The authors would like to thank the acute care paediatric nurses at the University of Virginia Children's Hospital for their voluntary participation and Molly Kampmann, Joseph Muething, Marcela Musgrove and Edward O'Leary for help with data collection. The authors thank Reecye Modny who provided technical support for the images and the Quality Committee at Children's Hospital who made this project possible.
Funding This work was generously supported by Grant Number T15LM009462 from the National Library of Medicine. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Library of Medicine, or the National Institutes of Health.
Competing interests None.
Ethics approval Ethics approval was provided by the Institutional Review Board, University of Virginia.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement Specific consent for data sharing was not obtained but the presented data are rendered anonymous and the risk of identification is low. The author may be contacted with requests for data sharing.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.