Safe Use of Electronic Medication Administration Records
During Medication Preparation and Administration:
A Socio-Technical System Perspective
Laura Keemink, BSc.
11
Master Medical Informatics, Amsterdam University Medical Centers, location
AMC, University of Amsterdam
Master Thesis
Safe Use of Electronic Medication Administration Records During Medication Preparation and Administration: A Socio-Technical System Perspective
Author
L.V. Keemink (Laura), BSc.
Department Medical Informatics, University of Amsterdam Meibergdreef 9, 1105 AZ Amsterdam
Supervisor
L.W.P. Dusseljee-Peute (Linda), PhD
Department Medical Informatics, University of Amsterdam Meibergdreef 9, 1105 AZ Amsterdam
Mentor
L. Spadon (Leandra), MSc. Furore Informatica B.V.
Bos en Lommerplein 280, 1055 RW Amsterdam SRP Duration:
December 2019 – June 2020 SRP Location:
Furore Informatica B.V.
Bos en Lommerplein 280, 1055 RW Amsterdam
Acknowledgements
This thesis is my final assignment for the Master Medical Informatics at the University of Amster-dam. I worked on my thesis at Furore in Amsterdam and at the Amsterdam University Medical Center (location Amsterdam Medical Center). Looking back on the past months, I am proud of the process and results of my scientific research project. I was and still am very interested in my subject and hope to see new insights and improvements in this field in the future.
There are a few persons that I would like to thank in particular for making this possible. I would like to sincerely thank my supervisor Linda Dusseljee-Peute for her guidance, expert knowledge and support. She did not only teach me how to perform proper research but also motivated me with her enthusiasm.
Second, I would like to especially thank my mentor Leandra Spadon for sharing her experiences, knowledge and advice with me. Even in busy periods, she was always willing to help me. Further-more, I would like to thank Marieke de Visscher for sharing her knowledge of medication systems and new interesting technologies with me. I also would like to thank Carly van Bussel for being my second researcher, questioning my findings and extensively discussing them with me. Finally, I would like to thank all my colleagues at Furore, everyone was willing to help me and I had a great time working with them.
Furthermore, I would like to thank everyone who participated in this research, especially the nurses who all made time for me and did not show any resistance in being observed.
At last, I would like to thank my family and friends for their support and interest in my study.
Abstract
IntroductionSafe medication administration is fundamental for patient safety, because medication errors are directly associated with mortality and morbidity. Nurses have an important role in the medication management process, as they form the last chain in the delivery of medication to the patient. They can detect mistakes made earlier in the process but their actions are also determinative for the safety of the patient. The Electronic Medication Administration Record (eMAR) can support nurses during medication preparation and administration but has the unintended effect of increas-ing the number of Medication Administration Errors (MAE) if not correctly designed or used. This study focuses on human factor instigated medication errors in the use of eMARs and the influence of socio-technical factors on these errors in order to provide recommendations on the improvement of safe eMAR use.
Methods
Open interviews with medication system experts provided insight into experienced difficulties in eMAR design and use. In a mixed-methods approach, a literature review combined with an ob-servational study provided insight into human factor instigated medication errors in the use of eMARs and how these applied to three non-teaching Dutch hospitals. A SEIPS-based analysis provided insight into the influence of socio-technical system aspects on these human factor errors. In addition, a literature review followed by a focus group resulted in recommendations for the improvement of safe eMAR use.
Results
In the literature study, 23 articles were included. An in-depth analysis of these articles resulted in a taxonomy based on 70 individual errors, which could be classified in 21 subclasses and led to 5 main classes for human factor instigated medication errors in the use of eMARs. In the obser-vational study, 881 medication administration actions were observed in a three-week time span in which human factor errors were observed 266 times. 3 out of 21 classes from the taxonomy were not applicable in the three observed hospitals, 17 out of the remaining 18 classes were observed in at least one of the hospitals and 13 out of 18 classes were observed in all hospitals. The SEIPS-based analysis indicated that workarounds by nurses in the use of eMARs and poor communication between doctors and nurses within the system formed the main problems in making safe use of eMARs. The literature review and focus group resulted in 7 recommendations and indicate the overall need for support within the eMAR system and interventions to gain insight into the prob-lems nurses experience on a daily base regarding eMAR use.
Discussion
This study indicates that in the use of eMARs, nurses experience a lack of support by the system leading to workarounds and insufficient communication between doctors and nurses. Although medication errors leading to patient harm were not observed in this study, the high number of hu-man factor errors detected might lead to high workload and frustration for nurses (and clinicians). Future research is needed to identify how these errors are associated with high workload and how work process and communication of doctors and nurses in eMARs can be improved.
Keywords—Electronic Medication Administration Systems, Human Factor Errors, Socio-Technical System, Medication Administration Errors, Patient Safety
Samenvatting
IntroductieVeilige medicatietoediening is fundamenteel voor patiënten, omdat medicatiefouten direct in ver-band staan met sterfte en ziektecijfers. Verpleegkundigen hebben een belangrijke rol in het medi-catie management proces, omdat ze de laatste keten vormen in het leveren van medimedi-catie aan de patiënt. Zij kunnen eerder gemaakte fouten in het proces opsporen, maar hun acties zijn ook bepal-end voor de veiligheid van de patiënt. De toedienregistratie kan verpleegkundigen ondersteunen tijdens het voorbereiden en toedienen van medicatie, maar kan onbedoeld het aantal medicatie toedienfouten verhogen wanneer het niet goed ontworpen of gebruikt wordt. Deze studie focust op medicatiefouten veroorzaakt door menselijke factoren in het gebruik van toedienregistraties en de invloed van socio-technische factoren op deze fouten om aanbevelingen op het verbeteren van veilig gebruik van de toedienregistratie te kunnen geven.
Methode
Open interviews met medicatiesysteem experts gaven inzicht in de ervaren problemen in het on-twerp en gebruik van toedienregistraties. Een mixed-methods studie van een literatuuronderzoek en een observationeel onderzoek gaven inzicht in medicatiefouten veroorzaakt door menselijke fac-toren in het gebruik van toedienregistraties en hoe deze voorkomen in drie perifere Nederlandse ziekenhuizen. Een op SEIPS gebaseerde analyse gaf inzicht in de invloed van socio-technologische systeem aspecten op deze menselijke fouten. De combinatie van een literatuuronderzoek en een fo-cusgroep resulteerde in aanbevelingen op het verbeteren van veilig gebruik van de toedienregistratie. Resultaten
In het literatuuronderzoek werden 23 artikelen geïncludeerd. Een diepgaande analyse van deze ar-tikelen resulteerde in een taxonomie gebaseerd op 70 individuele fouten welke konden geclassificeerd in 21 subklassen die leidden tot 5 hoofdklassen van medicatiefouten veroorzaakt door menselijke factoren in het gebruik van toedienregistraties. In het observationele onderzoek werden in drie weken tijd 881 medicatie toedienacties geobserveerd waarbij 266 keer een menselijke fouten werden geobserveerd. 3 van de 21 klassen van de taxonomie waren niet toepasbaar in de geobserveerde ziekenhuizen, 17 van de resterende 18 klassen werden in minstens één van de ziekenhuizen geob-serveerd en 13 van de 18 klassen werden in alle ziekenhuizen geobgeob-serveerd. De SEIPS gebaseerde analyse indiceerde dat workarounds door verpleegkundigen in het gebruik van toedienregistraties en slechte communicatie tussen artsen en verpleegkundigen binnen het systeem de belangrijkste prob-lemen vormden in veilig gebruik van toedienregistraties. Het literatuuronderzoek en focusgroep resulteerden in 7 aanbevelingen en laten zien dat de algehele noodzaak voor ondersteuning binnen de toedienregistratie en interventies om inzicht te krijgen in de problemen die verpleegkundigen dagelijks ervaren met betrekking tot het gebruik van de toedienregistratie.
Discussie
Deze studie wijst erop dat verpleegkundigen een tekort aan ondersteuning ervaren in het gebruik van de toedienregistratie, wat leidt tot workarounds en onvoldoende communicatie tussen artsen en verpleegkundigen. Ondanks dat er geen medicatiefouten zijn geobserveerd die hebben geleid tot schade aan de patiënt in deze studie, zou het hoge aantal gedetecteerde menselijke fouten kunnen zorgen voor een hogere werkdruk en frustratie van verpleegkundigen (en clinici). Verder onder-zoek is nodig om te identificeren hoe deze fouten van invloed zijn op hoge werkdruk en hoe het werkproces en de communicatie tussen artsen en verpleegkundigen in toedienregistraties verbeterd kan worden.
Steekwoorden—Elektronische toedienregistraties, Menselijke Fouten, Socio-Technologisch Systeem, Medicatiefouten, Patiëntveiligheid
Contents
Acknowledgements 2 Abstract 3 Samenvatting 4 List of Abbreviations 7 General Introduction 8 1.1 Research aim . . . 9 1.2 Thesis outline . . . 9 Theoretical Framework 11 2.1 SEIPS model . . . 112.2 Swiss Cheese model . . . 12
Classification of human factor instigated medication errors in eMAR use 13 Abstract . . . 13 3.1 Introduction . . . 14 3.2 Methods . . . 14 3.2.1 Open Interviews . . . 14 3.2.2 Literature Review . . . 15 3.3 Results . . . 16 3.3.1 Open Interviews . . . 16 3.3.2 Literature Review . . . 16 3.4 Discussion . . . 20 3.5 Conclusion . . . 20
Identification of socio-technical barriers in safe eMAR use in practice 21 Abstract . . . 21 4.1 Introduction . . . 22 4.2 Methods . . . 23 4.2.1 Observational Study . . . 23 4.2.2 SEIPS-based Analysis . . . 24 4.3 Results . . . 25 4.3.1 Observational Study . . . 25 4.3.2 SEIPS-based Analysis . . . 27 4.4 Discussion . . . 29 4.5 Conclusion . . . 30
Recommendations on improving the use of eMARs in relation to medication safety 31 Abstract . . . 31 5.1 Introduction . . . 32 5.2 Methods . . . 33 5.2.1 Literature Review . . . 33 5.2.2 Focus Group . . . 33 5.3 Results . . . 34 5.3.1 Literature Review . . . 34 5.3.2 Focus Group . . . 34 5.4 Discussion . . . 38 5.5 Conclusion . . . 38 Overall discussion 39 6.1 Principal findings . . . 39
6.2 Implications for the field . . . 40
6.3 Strengths and Limitations . . . 41
6.4 Future research . . . 41
6.5 Overall conclusion . . . 41
References 42
Appendices
A Search string on medication errors 50
B Open interviews with medication experts 51
C Medication errors found from literature review 53
D Classification of human factor instigated medication errors 55 E SEIPS-based analysis of medication preparation and
administration activities 57
F Analysis of barriers in the medication process and work system elements
influ-encing them 60
G Search strings on experience with medication safety systems 63 H Overview of articles on advantages and disadvantages of CLMM systems 64
List of Abbreviations
ADE Adverse Drug EventAMD Automated Medication Dispensing BCMA Barcode Medication Administration CLMM Closed Loop Medication Management CDSS Clinical Decision Support Systems COW Computer on Wheels
CPOE Computerized Physician Order Entry EHR Electronic Health Record
eMAR Electronic Medication Administration Record HFE Human Factors Engineering
HIT Health Information Technology IT Information Technology
IV Intravenous
MAE Medication Administration Error
SEIPS Systems Engineering Initiative for Patient Safety
VTGM ’Voor Toediening Gereed Maken’ (Integrated Preparation Protocols)
General Introduction
Since the report ‘To err is human’ was published in 1999, the attention for patient safety in-creased [1]. While medication errors are directly associated with mortality and morbidity, safe medication administration is fundamental for patient safety [2]. Nurses have an important role in the medication management process as they form the last chain in the delivery of medication to the patient [3]. Therefore, they can detect mistakes that were made earlier in the process, but their actions are also determinative for the safety of the patient [4]. A system that can support the nurse at the moment of medication preparation and administration is the electronic Medication Administration Record (eMAR) [5]. Despite its potential benefits, eMARs have the unintended effect of increasing the number of Medication Administration Errors (MAE) if not correctly de-signed or used [6]. A study of Carayon, et al. concluded that 34% of the medication errors made in hospitals were eMAR related [7]. A lot of research is conducted on detecting errors in systems used for medication prescription, but it remains unclear how eMARs form a problem for their users at the moment of medication preparation and administration [8].
‘To err is human’ by the Institute of Medicine showed that around 98,000 Americans die in any given year from preventable medical errors that occur in hospitals [1]. It broke the silence that surrounded medical errors and their consequence and pleaded for reducing medical errors and improving patient safety. Not by pointing fingers at caregivers who make unintended mistakes, but through the design of a safer health system. The idea that making errors is human, en-couraged the implementation of Health Information Technologies (HIT) to reduce errors in health care [9,10]. Since then, several studies reported on the ability of Clinical Decision Support Systems (CDSS), Computerized Physician Order Entry (CPOE) systems, and EMARs to reduce medical error rates [6,11,12]. While HIT had the potential to improve patient safety, its implementation and use led to unintended errors, called technology-induced errors [13,14]. Even though research began to show that HIT could introduce this new type of error years ago, these errors are still reported on. To gain insight into medication errors in order to find their causes, error reporting by caregivers is critical [3]. In a survey among nurses, a third of them experienced barriers in reporting medication errors due to fear of liability or lawsuits, fear of being blamed and/or fear of disciplinary action [15]. This shows that person-based thinking in healthcare, where a single person is held responsible, forms a barrier for reporting errors and refrains improvement of medication and patient safety. Although the need for error reporting and root cause analysis of these errors within the medication process is currently acknowledged, the role of the IT system within these analyses is often forgotten and remains unknown [16,17].
The idea that the flaw is in the entire process is visualized in the Swiss Cheese model designed by James Reason, which is presented in Figure1 [18]. Every block of cheese represents a defence against failure. Their function is to protect patients from hazards. The holes in every layer present possible mistakes. A single mistake does not necessarily lead to an incident, while it can be recovered by other barriers. However, when the holes in many layers line up, leading to a trajectory of mistakes, an accident can occur. This model shows that when an MAE occurs, a trajectory of mistakes within the entire medication process was made, leading to that error.
Figure 1: Swiss Cheese model [18]
A theory that has been proven to understand, improve, and redesigning processes for safer care is Human Factors Engineering (HFE) [19]. The theory applies especially to complex interacting socio-technical systems, such as hospitals. An important principle of HFE is to go beyond improv-ing simprov-ingle system elements and rather to analyse and improve the entire system. HFE focuses on both the elements of the system and their interactions. Although it is known that human-machine interactions play a role in medication errors, evidence is lacking on the extent of their contribution to the occurrence of these errors [20]. Patient safety events are typically underreported, which makes it hard to identify usability and safety challenges in system use [21,22]. Despite this lack
of knowledge, it is known that not only interactions between human and technology play a role in medication errors, but also social factors contribute to the usability of technical systems [23,24]. Therefore, the design of the eMAR must be based on the socio-technical factors of the organisation and its users to be used successful [25]. When the system is not adapted to these socio-technical factors, nurses will find workarounds to the system. Because nurses must frequently adapt to changing conditions, workarounds are a tool to manage their daily work. When a nurse is faced with a system that does not work as well as it should, the solution to improve this is likely to deviate from the standard operating procedure [26].
From the HFE perspective, models and frameworks have been developed for technology evaluation. For example, the Unified Theory of Acceptance and Usage of Technology (UTAUT) model and the Human, Organization, and Technology-Fit (HOT-Fit) model [27,28]. These models have been applied in healthcare settings, but focus on evaluating HIT and their adoption by users. A model that focuses on patient safety is the Systems Engineering Initiative for Patient Safety (SEIPS) model [29]. This model provides a framework to analyse work system design and its impact on processes and outcomes. A study by K.H. Frith showed that the SEIPS model is a useful framework for identifying the causes of errors, contributing factors related to errors, and identifying interventions to reduce errors [30]. By combining the SEIPS model with the Swiss Cheese model, understanding may be gained on how socio-technical factors influence the trajectory of errors within the process. The SEIPS model has served as a framework for understanding the impact of CPOE systems on end users [31]. However, reports on the applications of the SEIPS model to eMAR system use in hospitals are not reported on in literature. Therefore, by combining the SEIPS model with the Swiss Cheese model from a socio-technical point of view, this study has the intention to gain insight into how eMARs form a problem for their users and to propose solutions based on these problems within the socio-technical system.
1.1
Research aim
The aim of this study is to provide insight on how to improve safe eMAR use during medication preparation and administration from a socio-technical system point of view.
1.2
Thesis outline
In figure2, the design of this study is presented. The main research question that will be answered in this thesis is: "What human factor instigated medication errors in eMAR use during medica-tion preparamedica-tion and administramedica-tion should be optimised in order to improve patient safety from a socio-technical perspective?".
To be able to answer the main research question, this thesis was divided into three study compo-nents. Chapter 2 contains the theoretical framework of this thesis, defining the models and theories used. In Chapter 3, open interviews and a literature study are conducted to identify and classify possible human factor instigated medication errors in eMAR use. In Chapter 4 an observational study is performed to identify if and how nurses experience issues in the use of the eMAR. These issues are analysed using the SEIPS model, providing insight into how socio-technical system el-ements can cause medication errors. In Chapter 5, solutions which could help to improve the system, reducing the chance of errors to occur, are extracted from literature. These solutions are discussed in a focus group. The overall results of this study are discussed in Chapter 6.
Figure 2: Study design
Theoretical Framework
Two models that will be used in this study are the SEIPS model and the Swiss Cheese model. These models were shortly introduced in the general introduction and will be further explained in this chapter.
2.1
SEIPS model
In Figure3, the SEIPS model designed by Carayon, et al. is presented [29]. This model provides a framework for understanding the work system, processes, outcomes, and their interrelationships in healthcare. The work system exists of five components; a person performs a range of tasks using various tools and technologies. The performance of these tasks occurs within a certain physical environment and under specific organisational conditions. These five components (person, tasks, technology and tools, environment, and organisation) interact with each other and influence each other. The entire work system needs to be well designed for optimal performance. For example, a nurse who has excellent skills and knowledge may not give the highest quality and safety of care to the patient if its medication is not available in the department at the right time. This is an example of a poor balance in the work system, where one element creates a barrier for optimal performance. In the SEIPS model, processes include both care processes and other processes, such as maintenance, housekeeping, and supply chain management, which are influenced by the work system design. Processes subsequently influence outcomes, which consist of both patient outcomes and employee and organisational outcomes. Patient outcomes and employee/organisation outcomes also influence each other. For example, nurses’ frustration with the eMAR system can influence medication safety, which affects the patient’s outcomes. The SEIPS model specifies feedback loops from processes and outcomes to the work system. While poor processes and outcomes can be a reason to redesign, these feedback loops represent pathways to redesign the work system. Negative work system components that affect processes and outcomes need to be identified. Improving these components will lead to better quality and safety of care, as well as improved employee and organisational outcomes.
Figure 3: Seips model [29]
The SEIPS model as described above is known to help identify bottlenecks in the work system, which affect the process and outcomes. A study by Wooldridge et al., proposed a new socio-technical process modelling method to describe and evaluate processes, using the SEIPS model as a conceptual framework [32]. This study states that process mapping, often used in the HFE approach to improve care delivery and outcomes, should be expanded to represent the complex, socio-technical aspects of healthcare processes. By applying this method work system barriers and their contributing system components can be identified, showing where improvement is needed. In this study, the approach by Wooldridge et al. will be applied in order to detect barriers and their contributing components in the medication preparation and administration process. Wooldridge made use of the SEIPS 1.0 model, which will therefore also be used in this study. To prevent confusion, while the SEIPS 2.0 and 3.0 exist; when mentioning the SEIPS model the SEIPS 1.0 as presented in Figure3 is referred to.
2.2
Swiss Cheese model
The Swiss Cheese model is a model used in the system-based approach, in which the assumption is made that although the human condition cannot be changed, the conditions under which humans work can [18]. As proposed in the introduction, this study will follow a system-based approach by researching how to optimise the system instead of blaming the healthcare providers using it. A central idea in this approach is that of system defences. All hazardous technologies have barriers and safeguards and if an adverse event occurs, the important issue is not who blundered, but how and why the defences failed. Examples of defences in the medication process are decision support implemented into the system, systems monitoring the patient, and alerts or alarms. In the Swiss Cheese model, these defences would form the blocks of cheese. The holes within the cheese represent latent errors, an opportunity for error, a weakness in defences against error, or an unsafe act [33]. The presence of holes in the slices does not necessarily cause a bad outcome [18]. This can happen only when the holes in many layers temporarily line up, permitting a trajectory of accident opportunity, bringing hazards into damaging contact with patients. As presented in Figure4, in an ideal situation all the defensive layers would be intact, allowing no penetration by possible accidents [34]. However, in reality each layer has weaknesses as presented on the right side of the figure. Although the defensive layers and their holes are shown as being fixed, in reality they are in constant flux in response to external factors. Defence layers can move in and out of the picture and the holes within each layer are shifting around, coming and going, shrinking and expanding.
Figure 4: Ideal versus Reality defences [34]
The holes in the defence layers arise for two reasons: active failures and latent conditions [18]. Nearly all adverse events involve a combination of these two factors. Active failures are the unsafe acts committed by people who are in contact with the patient or system and have a direct, short impact on the safety of the defences. Latent conditions are the inevitable errors within the system. They arise from decisions made by designers, engineers, procedure writers, and top-level man-agement. Latent conditions may lie silently within the system for many years before they create an opportunity for accidents. Unlike active failures, latent conditions can be identified before an adverse event occurs and resolved proactive rather than reactive. An analogy of James Reason, the founder of the Swiss Cheese model, represents the need to focus on preventing latent conditions over active failures, and thus choose a system-based approach over a person-based approach: “Active failures are like mosquitoes. They can be swatted one by one, but they still keep coming. The best remedies are to create more effective defences and to drain the swamps in which they breed.” [18].
When improving safety, the goal is to either plug the holes, preventing an error to occur, to add another slice, which blocks the repetition of errors and works as a safety net or to shift the blocks, so that the errors are present in the system, but not arising in the process [33]. In this study, the Swiss Cheese model will be used as a metaphor, where the blocks of cheese will symbolise different systems that were introduced into the medication process to improve patient safety. The idea that a negative event occurs when errors in each block are aligned will also be used. Therefore, the goal is to find these holes and fill them, shift them, or block them, avoiding risks for the patient and improving patient safety.
Classification of human factor instigated medication errors in
eMAR use
Abstract
Introduction: Medication errors cause at least one death every day and injure approximately 1.3 million people annually in the United States of America alone. Other countries are estimated to have similar rates. Not every medication error necessarily leads to an injury, but medication errors are still an important cause of patient morbidity and mortality. Therefore, all medication errors should be treated as potentially harmful and attempts to improve medication safety should be encouraged. Wrong use of medication systems can cause medication errors. Therefore, this study focuses on medication errors that follow from human factors in the use of medication systems. Methods: To gain insight into experienced problems with the studied eMAR, open interviews with medication system experts were conducted. Furthermore, a literature study was performed to classify human factor instigated medication errors in the use of eMARs.
Results: The main conclusion from the open interviews was that the studied eMAR system was not fully adjusted to the work process of its user. This was mainly caused by the lack of a feedback loop, leading to a lack of knowledge on the needs of nurses. The literature study included 23 articles mentioning human factor errors in the use of eMAR and BCMA systems. A total of 70 individual errors were used to design a classification model, consisting of 21 subclasses and 5 main error classes.
Discussion: The results from open interviews imply that the eMAR system should either be adjusted to the work process of its users or the user should be educated on how to make optimal use of the system. A large body of literature showed that human factors play an important role in the emergence of medication errors during eMAR and BCMA use. Remarkably, literature reported on the same type of errors for over ten years. This implies that no sufficient improvement has been reached in this field during this period of time. However, it may also show how complex this problem and healthcare process improvement in general is.
3.1
Introduction
The World Health Organization reported that medication errors cause at least one death every day and injure approximately 1.3 million people annually in the United States of America alone [35]. Other countries are estimated to have similar rates of medication-related adverse events. Medica-tion errors can have many reasons, such as overcrowding, staff shortages, and poor training. As stated earlier in this thesis, wrong use of medication systems is one of these reasons. Any one of these reasons, or a combination of them, can affect the prescription, dispense, consumption, and monitoring of medications, which can result in severe harm, disability, and even death [35]. Not every medication error necessarily leads to an injury but when this does happen, it is called an Adverse Drug Event (ADE). Although only around 8-10% of the medication errors lead to ADEs, medication errors are still an important cause of patient morbidity and mortality [36–38]. Therefore, all medication errors should be treated as potentially harmful and attempts to improve medication safety should be encouraged.
Many definitions of a medication error have been proposed. In this study, the definition as pro-posed by the World Health Organization will be used: “A medication error is any preventable event that may cause or lead to inappropriate medication use or patient harm while the medication is in the control of the healthcare professional, patient, or consumer. Such events may be related to professional practice, healthcare products, procedures, and systems, including prescribing, order communication, product labelling, packaging, and nomenclature, compounding, dispensing, distri-bution, administration, education, monitoring, and use” [39]. So, even if the event could potentially harm the patient it is defined as a medication error.
The definition of HFE that will be used in this study is as defined in the Wiley Encyclopedia of Biomedical Engineering: "The discipline of engineering concerned with the analysis, design, and development of human technological systems in which the primary emphasis is to improve or optimize system performance by considering the human’s capabilities and limitations in the sys-tem" [40]. In this research, human factor instigated medication errors are modelled. These are medication errors that follow from human actions when using the system, in this case the eMAR. In this chapter, the focus lays on discovering medication errors following from human factors in the use of eMARs. Interviews with medication system experts will be conducted to find out which issues with the eMAR system are already known. Furthermore, a literature study will be per-formed, searching for studies reporting on medication errors in the use of eMAR systems at the moment of medication preparation and administration. These errors will then be used to design a classification model of human factor errors in the medication process.
3.2
Methods
3.2.1 Open Interviews
Open interviews were conducted to gain insight into the experienced problems with the studied eMAR system. Two medication system experts working at Furore participated in these inter-views [41]. Both experts had experience as system administrators in hospitals, leaded medication workgroups, and taught nurses on the use of the eMAR. The experts were asked which modules of the systems were used for medication preparation and administration, by whom and if errors and dissatisfaction with the system were already known. A total of three interviews were conducted to receive sufficient information, interviewing only one expert at a time. The interviews were transcribed and coded by one reviewer using ATLAS.ti [42]. These codes were induced to smaller categories until theories on the problems experienced with eMAR were extracted.
3.2.2 Literature Review Inclusion Criteria
For the literature study, the medical databases PubMed and EMBASE were searched until January 2020 [43,44]. PubMed was chosen, while it was a credible database in the field of Medical Infor-matics and the researcher was most common with this database. EMBASE was chosen, while it covered the same subjects as PubMed with, among other things, an additional focus on medical de-vices [45]. These databases were searched for publications studying human factor errors in the use of eMARs and Barcode Medication Administration (BCMA) systems. BCMA systems are eMARs supplemented with barcode scanners for patient and medication verification [46]. BCMA systems were also searched for, while in some countries eMARs are always used with barcode scanners and therefore only referred to as BCMAs. Not including them could lead to missing information. The definition used for a ‘human factor error’ was as described in the introduction. An eMAR was referred to as a (part of a) system in which pharmacists and nurses can record the preparation or administration of medication and/or an infusion [47]. A BCMA system was defined as an eMAR which lets the user scan medication and/or the patient’s barcode [46].
Validation Set
To be able to validate the search strategy, a validation set of relevant articles was composed. The articles of the validation set were found by open search terms in PubMed and Google Scholar and were included based on the title and abstract. The validation set was used as a basis for the search string. The search string was adjusted until at least 80% of the articles of the validation set were present in the search results. The remaining articles were then added to this result. The precise search string is presented in AppendixA.
Article inclusion
To perform article inclusion, the articles resulting from the two searches were imported into Rayyan QCRI [48]. At first, duplicates were resolved and articles that were published more than 15 years ago were excluded, to prevent including outdated information of old systems. This was followed by title and abstract inclusion and exclusion. If it was clear from the title and abstract that the article did not contain the sought-after information, it was excluded. On the residual articles, a full-text review was performed. An article was included if human factor instigated medication errors of the eMAR or BCMA system were mentioned.
Backward snowballing
To make sure no important articles were missed due to a too strict defined search strategy, back-ward snowballing was applied. In backback-ward snowballing, the reference lists of included papers were used to identify new papers to include [49]. First, the reference lists were examined, selecting arti-cles that might meet the inclusion criteria based on their title. If the selected article was available, not more than 15 years old, and not already examined, a full-text review was performed. If the article met the inclusion criteria, it was added to the set of included articles.
Analysis and classification
From the included articles, all eMAR and BCMA related medication errors were extracted. If a unique error was mentioned in more than one article, this was counted. A classification model was designed to provide a clear overview of these errors. Single errors were grouped if they belonged to the same error class. The definitions of these classes were retrieved during the literature review. Thereafter, these classes were again classified into broader categories. This classification was compared to the taxonomy of medication errors related to Information Technology (IT) usability by L.W.P. Peute, commissioned by the Dutch Ministry of Health, Welfare and Sport [50]. Classes and errors that both classifications agreed on could be validated, and classes and errors of the taxonomy that were not found in literature but met the inclusion criteria were added to the classification.
3.3
Results
3.3.1 Open Interviews
From the information gained by the open interviews with medication system experts, nine cat-egories of issues could be extracted. These nine catcat-egories could be induced to five catcat-egories: too safe choices in the system could lead to workarounds; decisions on medication safety in the system could impact the timely administration of medication; mistakes of nurses were not regu-larly corrected and learned from; it was hard to get insight into the needs of nurses due to an insufficient feedback loop; and different needs of nurses needed to be translated to the possibilities of the system. These five categories were then summarized into two issues; a lack of feedback loops and an imbalance between the safety and usability of the system. These two issues led to the overall problem; more insight into the current working process of nurses is needed to base policies, agreements, and systems on. The full description of categories is presented in AppendixB. 3.3.2 Literature Review
As presented in Figure5, 55 articles were found in PubMed and 52 in EMBASE. Of the resulting 107 articles, 44 were duplicates and therefore excluded. 63 articles remained for title and abstract review, of which 21 were excluded. Reasons for exclusion can be found in the figure. The validation set consisted of 10 relevant articles [51–60]. 8 out of 10 articles of the validation set were present in the search results, the remaining 2 articles were added to the results [51,52]. Subsequently, 44 articles were reviewed based on a full-text review, after which 21 were included [51–71]. By back-ward snowballing, 2 articles were found to meet the inclusion criteria and were therefore included in the study [72,73]. After one round of backward snowballing, no new information was found, only already known factors were confirmed. Therefore, the reference lists of the articles included after the snowball method were not further analysed.
Figure 5: Workflow of the review
Of the 23 included articles, 5 articles solely studied eMAR systems, and 18 articles studied eMAR systems in combination with barcode scanning. Table 1 provides an overview of the included articles and contains the general characteristics, study type, system type and the type of errors identified.
Table 1: Overview of the included articles
First author Year Country Study design System Identified errors
Franklin, B.D. [62] 2007 England Observational study eMAR/BCMA 6 types of errors after closed-loop system implementation includ-ing eMAR and BCMA (compared to 10 before implementation) Koppel, R. [53] 2008 USA Mixed methods study eMAR/BCMA 15 types of BCMA-related workarounds and 31 separate
proba-ble causes of the identified workarounds
Sakowski, J. [64] 2008 USA Review study eMAR/BCMA 4 types of medication errors that were not eliminated by BCMA Vogelsmeier, A.A. [71] 2008 USA Mixed methods study eMAR 3 types of workarounds associated with the implementation of
an eMAR
Helmons, P.J. [72] 2009 USA Observational study eMAR/BCMA 6 types of errors introduced by BCMA implementation
Morriss, F.H. [63] 2009 USA Observational study eMAR/BCMA 6 types of MAEs after implementation of BCMA (compared to 8 types pre-BCMA)
Young, J. [52] 2010 USA Systematic review eMAR/BCMA 11 MAE categories identified after BCMA implementation Guo, J. [57] 2011 USA Heuristic evaluation eMAR 10 usability related errors in eMARs
Harrington, L. [51] 2011 USA Systematic review eMAR/BCMA 7 types of BCMA related safety issues
Miller, D.F. [55] 2011 USA Observational study eMAR/BCMA 3 workaround categories containing 7 specific workarounds in BCMA use
Rodriguez-Gonzalez, C.G. [73]
2011 Spain Observational study eMAR 4 types of administration errors in combination with automated dispense systems
Tbale 1: Overview of the included articles (continued)
First author Year Country Study design System Identified errors
Szczepura, A. [69] 2011 England Observational study eMAR/BCMA 4 types of medication errors in BCMA which was confusing and open to error
Hardmeier, A. [67] 2014 USA Observational study eMAR/BCMA 5 types of medication errors that were not eliminated by BCMA implementation
Seibert, H.H. [68] 2014 USA Observational study eMAR/BCMA 7 types of medication errors after BCMA implementation (com-pared to 7 types pre-BCMA, but different types)
Tariq, A. [66] 2014 Australia Observational study eMAR 5 types of usability related errors of the eMARs Bowers, A.M. [61] 2015 USA Observational study eMAR/BCMA 4 types of medication errors in the use of BCMA Sakushima, K. [65] 2015 Japan Incident report analysis eMAR/BCMA 7 types of errors reported after BCMA implementation Staggers, N. [60] 2015 USA Heuristic evaluation eMAR/BCMA 4 types of usability related medication errors
Truitt, E. [70] 2016 USA Incident report analysis eMAR/BCMA 10 types of errors in administration using eMARs (compared to 8 types pre-eMAR)
Oliveros, N.V. [56] 2017 Spain Observational study eMAR 6 types of medication errors in eMARs
Lin, J.C. [58] 2018 Taiwan Survey study eMAR/BCMA 5 types of medication errors related to work process in the use of BCMA
Macias, M. [54] 2018 Spain Observational study eMAR/BCMA 6 types of medication errors in the use of BCMA
Risor, B.W. [59] 2018 Denmark Observational study eMAR/BCMA 10 types of medication errors in automated medication admin-istration including eMAR/BCMA
From the included articles, errors that could occur while using the medication administration system were extracted. AppendixCcontains a list of all errors that were found, including in how many articles the error was mentioned. These errors were classified into 21 subcategories, leading to 5 main categories. This classification is presented in AppendixD. An overview of the subclasses and main categories is presented in Table 2. The first class contains general medication errors that could occur during administration related to system use. The second class represent errors that could occur during the recording of administered medication into the system. The third class includes design issues of the system that influence its usability and safety. The fourth class consists of issues that could occur when medication administration is performed using barcode scanning. The fifth class contains external factors that showed to have a linkage with eMAR/BCMA system implementation and influence its usability and safety.
Table 2: Classification of human factor instigated medication errors
1. Medication errors 3.4 Noncompliance with standard procedures 1.1 Faulty order process 3.5 Issues with the system
1.2 Wrong medication order 3.6 Complicated screen 1.3 Faulty medication preparation 3.7 Inconvenient functions 1.4 Faulty medication administration 3.8 Navigation errors 1.5 Neglection of protocol
4. BCMA issues 2. Recording errors 4.1 Lack of scanning
2.1 Faulty recording 4.2 Wrong execution of scanning 2.2 Faulty workflow 4.3 Faulty label scanning
4.4 Neglection of the system 3. System usability issues
3.1 Non-intuitive design 5. External factors 3.2 Lack of system support 5.1 Distraction
3.3 Lack of functions 5.2 Lack of standard procedures
3.4
Discussion
In this first chapter, insight was provided into the current knowledge on human factor instigated medication errors of eMAR systems. These insights were obtained by conducting interviews with medication system experts and a literature study. In the interviews, two experts were interviewed to gain insight into experienced problems with the studied eMAR system. In the literature review, articles reporting human factor errors in eMAR and BCMA system use were used to design a classification model.
From the open interviews, it was found that the main experienced issue with the eMAR system was the system not being fully adjusted to the work process of its users. This was mainly caused by the lack of a feedback loop, leading to a lack of knowledge on the needs of nurses. Rutledge et al. performed a survey on barriers to medication error reporting among hospital nurses [15]. 18.6% of the nurses experienced a lack of feedback to themselves or their department on the reported medication errors. A lack of knowledge of the usefulness of reporting medication errors was experienced by 22.8% of the nurses and 12.2% believed that reporting medication errors had little contribution to improving the quality of care. This shows the need to inform nurses about the importance of reporting medication errors and keep them updated on the follow-up of the report. In the literature study, 23 articles were found reporting on human factor instigated medication errors in medication preparation and administration. Remarkably, literature was found from 2008 until 2019 reporting on the same type of mistakes. This shows that for at least ten years no sufficient improvement has been reached in this field. However, it may also show how complex this problem and healthcare process improvement in general is. Henri Manasse confirms this in his book ’Medication Safety: A Guide for Health Care Facilities’ by stating that making significant improvements in patient safety is complex, difficult, time-consuming, and, in some cases, costly [74]. Failure to implement improvement aims may result in negative outcomes and increased patient care costs. He also argues in his book that a system-based and human factors approach should be used for improving medication safety.
This first chapter had a few limitations. First, only two medication system experts were available for interviews on the experienced problems with eMAR systems. However, while the goal was to gain insight into the topic, and not to fully discover all experienced problems, this was found to be sufficient. A second limitation is that only one researcher performed article inclusion, which could lead to insufficient inclusion. However, the classification was compared to a similar taxonomy to find errors and classes that might have been falsely included or excluded in the literature study and the classification was discussed and adjusted with a second researcher.
3.5
Conclusion
In this chapter, answers were found on the questions which problems with the use and design of eMAR systems were known by medication system experts and reported in the literature. The results imply that the system should either be adjusted to the work process of its users or the user should be educated on how to make optimal use of the system. A classification model of human factor instigated medication errors was designed, based on earlier studies and the taxonomy by Peute. In the next chapter, this study will investigate if and how the findings of this chapter apply in practice in three Dutch hospitals, using the classification model as a solid base.
Identification of socio-technical barriers in safe eMAR use in
practice
Abstract
Introduction: The key to developing highly usable systems is to apply user-centered design. Performing usability evaluations early in the design process is a fundamental step in user-centered system design. Despite the critical role of eMARs in patient healthcare delivery and the crucial role in nurses’ work, few usability evaluations are available for eMARs. Insight is needed into the medication errors made due to an insufficient system, workarounds invented to cope with the system and user satisfaction of the system. By proactively identifying how and why users work around the system, it can be redesigned to improve the system, user satisfaction and patient safety. Methods: By conducting an observational study in three non-teaching Dutch hospitals, insight was provided into the use of an eMAR system during the medication preparation and administration process. By applying the SEIPS model on these observations, experienced barriers and their contributing socio-technical system components were identified.
Results: The neglection of protocols was a high-risk class for all hospitals. When the system was not adjusted to the workflow of its user, workarounds were introduced. A second big issue in all three hospitals was the communication between the doctor and the nurse within the system. Sometimes, the doctor did not order medication in time, ordered ambiguous, or did not change an order when this was needed.
Discussion: Instead of telling staff to just follow the protocols, repeated examinations and corrections of actual uses of the system are needed to optimize their role in preventing medication errors. Different studies showed with practical applications that the SEIPS model is a useful framework in medication safety for identifying the causes of errors, contributing factors related to errors, and interventions to reduce errors. However, no applications could be found of the SEIPS model applied to the process model, as proposed by Wooldridge et al., which was applied in this study. This analysis showed which elements should be optimised in order to improve medication safety.
4.1
Introduction
The electronic medication administration system is a key component of the medication preparation and administration process [5]. The use of eMARs has been recommended for preventing medication errors, especially if the system is linked to the CPOE system, in which the doctor prescribes medication orders [12]. Although eMARs can support the medication process, reduce medication errors, improve patient safety and advance workflow efficiency, poor eMAR design can place patients, nurses, and the organisation at risk of physical and legal harm [57,75] The most basic function of an eMAR is to display ordered medications for patients [76]. However, this description underestimates the complexity, both functionally and technically, for nurses. For example, the eMAR can display a large variety of medications with different administration routes, from oral pills to cytostatic drugs, which are integrated into a set of linked orders. Furthermore, some medications are prescribed as needed, while others may have a strict time schedule and can be accompanied by extensive instructions. These complexities introduce risks and should be accounted for in the design of the system.
The key to developing highly usable systems is to apply user-centered design, adjusting the system to the nurses’ medication activities [5,57]. Performing usability evaluations early in the design process is a fundamental step in user-centered system design [77,78]. The objective of these usability evaluations should be to assure that the eMAR fits nurses’ cognitive and behavioural requirements and positively impacts patient safety [79]. The complexity of eMARs has been underestimated by most designers in the past and have received little attention by developers and computer scientist [76,80]. Despite the critical role of an eMAR in patient healthcare delivery and the crucial role in nurses’ work, few usability evaluations are available for eMARs [60,76].
When a system is not designed with or around its users, the mismatch between user and system can introduce workarounds [57]. About workarounds, Berlinger et al. said: “Some nurses store extra pillows and blankets in a little-used closet to ensure they are available when needed. Others circumvent their institution’s medication administration system by prescanning patient bar codes instead of following proper procedure for scanning at the bedside. Ask a group of nurses to describe how they get the job done, and they’ll talk about workarounds: shortcuts and fixes for situations in which work rules or procedures don’t match work realities“ [26]. This shows that nurses, in all aspects of care, will invent workarounds if the situation, environment or system does not match their work processes.
The combination of a lack in eMAR usability evaluations, proof that eMAR system design is complicated and the experience that nurses currently need to deviate from the system lead to the need for more information on how the eMAR system is used by nurses in hospitals. Insight is needed into the medication errors made due to an insufficient system and workarounds invented to cope with the system. By proactively identifying how and why users work around the system, it can be redesigned to improve the system, user satisfaction and patient safety [81]. Therefore, this chapter focuses on discovering barriers in safe use of eMAR systems and the socio-technical barriers contributing to these barriers. An observational study will be performed, observing the use of an eMAR system in three Dutch hospitals at the moment of medication preparation and administration. These observations will be classified in the medication error model as designed in Chapter 2. Furthermore, observed problems encountered with the eMAR system will be modelled according to the SEIPS model, detecting barriers and their contributing system elements.
4.2
Methods
4.2.1 Observational Study
Methods on detecting medication errors date from 1962 where Barker et al. stated observational studies to be the best method, while it was independent of the subject’s willingness and ability to report medication errors [82]. Based on the observation method of Barker, it was found that only 0.07% of the medication errors observed in hospitals were reported on incident reports. An explanation for this was that nurses were afraid of the negative effects on their job or even legal issues [83]. Also, reporting mistakes was not included in the workflow of the nurse, resulting in an extra workload. Even though 75% of the nurses indicated finding it important to report mistakes, they sometimes did not know they made a mistake, while the mistake did not lead to negative outcomes for the patient. This last problem made anonymous reports, questionnaires and interviews a less suitable solution for finding medication errors than observational studies. In a more recent study by Flynn et al., twelve methods for detecting medication errors in hospitals were compared by six medication error experts [84]. The accuracy of the three methods that were expected to perform best were observational studies with 81.6%, chart review with 5.3% and incident report review with 0.2%.
Based on these findings, an observational study was chosen to find eMAR related medication errors in hospitals. The observations were conducted by one and the same observer in three non-teaching Dutch hospitals that used the same eMAR system for their medication preparation and administration. Table3shows which departments were observed per hospital. The different hospitals were visited for the same amount of time (one week), observing about the same amount of medication activities (n ≈ 294 per hospital). A medication activity was defined as the preparation or administration of a single medication for a single patient. The study was carried out in a natural setting, during actual medication rounds. The pharmacists and nurses knew they were observed to gain insight into the way they used the system, but not exactly what was checked, in order to minimize social desirable behaviour. Participants were assured that it was not a personal test and their actions and behaviour could not lead to sanctions. An observation was written down in a notebook if the participant experienced problems with the use of the system, performed an action that was not in line with the protocols or made a mistake. If an explanation on the actions and operations of the participant was needed to understand an observation at that moment, this was asked during the medication rounds. Questions that did not need direct explanation were asked afterwards.
Table 3: Departments where observational studies were performed
Hospital 1 Hospital 2 Hospital 3
Intensive Care Cardiology Cardiology, Gynaecology and Urology
Throat/Nose/Ear Surgery, Breast Clinic, Neurosurgery and Plastic surgery
Neurology Surgery,
Stomach/Bowel/Liver Diseases and Vascular Care Oncology and Internal
Medicine
Pharmacy Nephrology, Oncology and
Internal medicine
Pharmacy Pharmacy
Analysis of observations
After all three hospitals were visited, the observations from the notebook were copied to the software program Optimal Workshop, keeping the observations separate per hospital [85]. If a specific observation was captured multiple times in the same hospital, it was only copied once into the program. However, the numbers were saved for further analysis. In this program, the observations were assigned codes, representing the 21 human factor error classes found in the literature study. Each code represented one class and all observations were assigned one code that fitted best. A second coder performed this coding process in the same way. At first, the observations were classified individually followed by a discussion until consensus was reached. The results of both coders were compared to check the inter-coder reliability and inter-coder agreement. Intercoder reliability represented how equally skilled coders independently selected the same code for the same observations [86]. This measure gave an indication for the reproducibility of the coding if it would be executed by equally experienced coders. Intercoder agreement represented the consensus on coding after a discussion about the differences had taken place [86]. The intercoder agreement corrected for discrepancies that occurred from inexperienced or not equally skilled coders. After the observations were classified, the risk per class was determined for every single hospital. To calculate the risk, the variables frequency and severity were used. The frequency was determined by the number of observations that were assigned to that specific class, increasing if a single observation was seen multiple times in that hospital. While the severity of a problem is less objective than the frequency, an existing and validated severity rate by J. Nielsen was used [87]. Nielsen proposed a four-step scale in assigning a severity rate to a usability problem:
0 = I don’t agree that this is a usability problem at all
1 = Cosmetic problem only: need not be fixed unless extra time is available on the project 2 = Minor usability problem: fixing this should be given low priority
3 = Major usability problem: important to fix, so should be given high priority 4 = Usability catastrophe: imperative to fix this before the product can be released
While observations were only captured if they formed a problem, the scale of 0 was removed. The researcher who performed the observations, assigned the severity to the classes, discussed this with the second coder and made adjustments if needed. Based on frequency and severity, a 3x4 risk matrix was designed to determine the risk per class, following the recommendations on the use and design of risk matrices as proposed by N.J. Duijm [88].
4.2.2 SEIPS-based Analysis
As explained in the theoretical framework, the SEIPS model provides a framework for understanding the structures, processes, outcomes, and their interrelationships in healthcare [29]. In this study, the SEIPS framework was filled in based on the observations in the participating hospital. In a study by Wooldridge et al., in which one of the designers of the SEIPS model contributed, it was proposed to combine the SEIPS model with process mapping to represent the complex sociotechnical aspects of healthcare [32]. This would help to understand and analyse healthcare processes systematically and identify specific areas for improvement. Therefore, a process map was designed of the medication process, from the moment medication was ordered to the moment the medication administration was recorded. This process map was based on the procedure in the three visited hospitals and designed using the software Lucidchart [89]. Thereafter, all process steps were analysed as proposed by Wooldridge et al., mapping the barriers faced in the specific steps. This was followed by assigning the SEIPS work system components that contributed to these barriers. First, the main researcher mapped the barriers and assigned the components based on the observational study results. A SEIPS component was assigned to a barrier if an element of that component caused the barrier. For example, if a medication is recorded as prepared before actually preparing it, it would be assigned the element ’Technology & Tools’. The element, in this case, is the eMAR system allowing recording before the actual preparation occurred. Multiple SEIPS components could be assigned to a single barrier. Thereafter, this process-based SEIPS analysis was thoroughly discussed with a second researcher, making sure the root cause of each barrier was found. If information was lacking to get to this root cause, a system expert was consulted.
4.3
Results
4.3.1 Observational Study
The risk-analysis of the observational study is presented in Table4, followed by an explanation of the content and observations. Red, orange and green respectively represent a high, medium and low risk. When coding these observations, the intercoder reliability was 63%, and the intercoder agreement was 98%.
Table 4: Risk-analysis of the observational data
Medication errors
In all observed hospitals, issues with the order process of the doctor were experienced. The doctor used a CPOE system for medication orders, which was linked to the eMAR of the nurse. One of the experienced issues was that the doctor changed an already prescribed medication as a new order instead of adjusting the original order. The system raised an alert when this occurred, but somehow this was ignored by the doctor. If the doctor did not (timely) remove the original order, a too large dose of that medication was prescribed in the system. The nurse then needed to notice this and only administer the most recent order, ignoring the old one. Another issue was experienced with anticoagulants, which needed to be redosed every day based on measurements of the patient. The order was not displayed to the nurse until the dose was determined and ordered by the doctor. When the doctor did not order anticoagulants in time, the nurse could ask the doctor to do this. However, this was extra pressure for the nurse and not recognizing the need to do so could have led to serious harm to the patient. A last issue in the medication order process was experienced when the workflow of the doctor and the nurse were not in line with each other. For example, a patient could, in theory, change from Intravenous (IV) paracetamol to oral paracetamol, but this was not yet ordered by the doctor. The nurse decided to administer oral medication and to record intravenous medication. This did not directly harm the patient but was not desirable due to inadequate recording. Furthermore, it resulted in an incorrect fluid registration in the system, while the medication was administered as a pill, but recorded as a fluid.
’Neglection of protocol’ was a high-risk category for all hospitals. Examples of these errors were administering medication before approval of an authorized prescriber; staying logged in on the Computer On Wheels (COW) without supervision, making it possible for others to use the system unauthorized; unlocking all medication drawers and leaving them unsupervised; administering medication ad-hoc without consultation and faulty performance of the double check.
Recording errors
’Recording errors’ were only an issue in hospital 1, caused by differences in utilized technologies. In hospital 1, the COW had to be unlocked with username and password authentication. The COW locked after a fixed amount of time and logging back in took approximately ten seconds. In hospital 2 and 3, the COW could be unlocked with a staff pass. Logging in and out with a staff pass took around one second. While logging in into the system was time-consuming in the first hospital, this sometimes resulted in recording all medication of a patient at once before preparing or administering it.
System usability errors
In hospital 1, the usability of the system underperformed in general, and not on one specific issue. Examples of these issues were missing the possibility to cancel or suspend medication after a fixed amount of time, medication pumps that could only be recorded as single administrations instead of a continuous pump on some departments, a lack of dosing support based on measurements, and medication that was automatically set to home medication while the patient was still hospitalized. In hospital 2, lack of functions and system support was the biggest issue in the usability of the system. Not being able to record double-checks in the system was a main reason, while the nurses wanted some insurance that the double-check was performed in case something went wrong. Furthermore, insulin was still managed on paper and then copied to the system by both the doctor and nurse, which is sensitive to errors. A module to perform this in the system is available but was not yet found to be suitable and trusted.
In hospital 3, lack of system support and issues with the system were mainly caused by the use of the MedEye. MedEye is a complete closed loop medication safety suite for hospitals and healthcare institutions [90]. The nurse slid the patient’s medication in the MedEye and the scanner recognized the medicine using computer vision. If the medication matched the prescribed medication, it was automatically recorded in the eMAR as administered. This system raised an error when the medication was incorrect, for example, when the wrong dose was prepared. While this should improve medication safety, it did not yet work optimally. On some departments, the internet connection was suboptimal and made the MedEye slow or not working at all. This resulted in a bad score on ’issues with the system’. Furthermore, the MedEye sometimes raised an incorrect error saying the pill was another medication than it actually was. The inconsistency of the MedEye
resulted in a high frequency of observations in ’Lack of system support’, because the nurses could not fully rely on the MedEye.
Barcode Medication Administration issues
The main finding in barcode medication administration issues was that none of the hospitals used barcode scanners. In hospital 1, scanners were not available, in hospital 2, scanners were only used for blood transfusion products and in hospital 3, scanners were not used and/or did not work, varying per department.
External factors
Distraction was a big issue during medication preparation and administration, while the pharmacists and nurses were asked questions by patients and colleagues. This was a problem in all three hospitals and departments. Although this seemed to be hard to prevent with a system, one of the nurses commented that the distraction by patients increased when the COW was taken into the room instead of left in the hallway. She explained that the computer was a sign for the patient that the nurse was available for that patient at that moment.
4.3.2 SEIPS-based Analysis
In Table5, the SEIPS model applied to the medication process is presented, showing its components, processes, outcomes and a short overview of experienced barriers.
Table 5: Components of the SEIPS model in the medication process
In Figure6, the process map of the medication preparation and administration process is presented. The model starts at the moment of medication prescription and ends at the end of a medication round. The swimlanes of the model each contain a person/object; a doctor (assistant), pharmacist (assistant), nurse, administration system and patient. While differences between hospitals exist, this model shows the most general process. For example, in some hospitals nurses perform medication preparation instead of pharmacists or automated systems are used. Furthermore, systems as the MedEye could be used within the process but were not included, because it is not most common. The red coloured process steps represent activities where barriers were found during the observational study. In AppendixE, all activities are described, including their possible barriers, and the SEIPS components contributing to these barriers (Technology & Tools (S), Organisation (O), Environment (E), and Tasks (T)).
Figure 6: Medication preparation and administration process
In AppendixE, barriers were coupled to SEIPS components that contributed to these barriers. A full analysis of how these work system components affect the experienced barriers is included in Appendix F. When merging these results, it was seen that in the system component the biggest issues were encountered when the system gave its user too much freedom, was experiencing downtime, was not supporting its user enough, or was too complicated. In the organisation component, a lack was seen in making and maintaining rules and regulations, making nurses aware of specific risks, and providing training to nurses. The impact of the environment was seen when the pharmacists and nurses were distracted during medication preparation and administration, or when the system was not physically available. Tasks either caused a too high workload or were poorly planned. When these socio-technical elements were coupled to the most important human factor errors observed in eMAR use, ’neglection of protocol’ indicated to workarounds of nurses in the use of eMARs caused by a lack of support within the system and the complexity in usability of the system. Furthermore, the amount of human factor errors observed in the order process of the doctor were linked to a lack of communication with nurses within the system, where a supporting role of the system and educational interventions from the organisation were missing.
The SEIPS analysis showed how observed human factor errors were affected by socio-technical aspects. The Swiss Cheese model was in first instance applied to analyse how these human factors influenced each other, leading to a trajectory of errors. Only one instance was found which could be analysed in relation to the Swiss Cheese model. This happened when a patient was prescribed a medication by the doctor, however, during visitation by the doctor it became clear that the patient had an allergy for the prescribed medication. No COW was available at that moment, so the doctor discussed this with the nearest nurse. Before the doctor changed the order in the system, another nurse was administering medication to this patient according to the order in the eMAR. The patient told the nurse about the allergy, which prevented a possibly harmful event. While this was the only instance which could be analysed, the Swiss Cheese model could not be used to analyse the relation between human factors within this study.
4.4
Discussion
In this second chapter, insight was provided into the use of eMAR systems in hospitals, what barriers were experienced, and which system components caused these barriers. These insights were obtained by conducting an observational study in three Dutch hospitals and a SEIPS-based analysis of these observations.
In the observational study, several departments were visited where pharmacists and nurses were observed while they used the eMAR during medication preparation and administration. It was seen that the neglection of protocols and the order process of the doctor were main problems for all hospitals. From a socio-technical perspective it was seen that, when the system is not adjusted to the workflow of its user, workarounds will be introduced [26]. This finding was also reported by Koppel et al. who found 31 workarounds in the use of BCMA systems [53]. Koppel states it is not enough to tell staff to “do it right”. Instead, repeated examinations and corrections of actual uses of the system are needed to optimise their role in preventing medication errors. The opinions on the acceptance of workarounds are divided. M. Boonen, for example, found it surprising that when a nurse deviates from the system it is seen as a (negative) workaround, while if the organisation sees reasons to deviate from the system it is seen as an adjustment to not disturb the workflow [91]. He argues that a workaround by a nurse should not always be seen as something negative, while it is important that nurses always use their knowledge, experience and clinical reasoning and not solely rely on the system. However, W. van der Veen showed that workarounds led to 7% more medication errors than if the system would have been followed according to protocol [25]. A second big issue in all three hospitals was the communication between the doctor and the nurse within the system. Sometimes, the doctor did not order medication in time, ordered ambiguous, or did not change an order when this was needed. Other studies have also shown that CPOE systems may undermine the efficiency and safety of the medication process, by disrupting the collaboration and communication between doctors and nurses [92,93]. However, Hoonakker et al. concluded that few studies have examined long term effects of CPOE implementation on communication aspects [94]. Different studies showed with practical applications that the SEIPS model is a useful framework in medication safety for identifying the causes of errors, contributing factors related to errors, and
