Assessing performance of an Electronic Health Record (EHR) using Cognitive Task Analysis
Introduction
Many health information system projects fail due to lack of systematic consideration of human-centered computing issues such as usability, workflow, organizational change, and process reengineering [1]. This paper evaluates the usability of the AHLTA (Armed forces Health Longitudinal Technology Application) EHR (Electronic Health Record) system that is used by the US military which connects 412 medical clinics, 414 dental clinics and 65 military hospitals. According to Dr. Casscells, assistant secretary of defense for health affairs, AHLTA is difficult to learn, and once you have learned it, it is cumbersome and difficult to navigate [2]. Clinician users of AHLTA also report seeing fewer patients and having longer workdays, largely because of the extra time needed to use the system [3]. Two major factors that lead to sluggish performance of this EHR system are complexity of the GUI (Graphical User Interface) and system response time [4]. This paper analyzes the complexity of the GUI of AHLTA independent of its system response time.
The usability and usefulness of EHR can be evaluated through UFuRT (User, Function, Representation, Task) analysis [5], [6], [7], which is a process for the design and evaluation of work-centered products through User, Function, Representation, and Task analysis. User analysis identifies the categories of users; Functional analysis identifies the ontology of a given work domain; Representational analysis identifies an appropriate information display format for a given task, and Task Analysis identifies the procedures and actions to be carried out for a given task goals by using specific representations. Task analysis of AHLTA was the main focus of this study. Specifically, we used a Cognitive Task Analysis called GOMS (Goals, Operators, Methods, and Selection rules) to analyze a set of prototypical tasks of AHLTA users. In addition, we applied an associated GOMS method – KLM (Keystroke Level Model) to evaluate the execution time required for the given task [8].
- 1.
Cognitive Task Analysis (CTA) is a core methodology used in cognitive science to study human performance in both laboratory and real-world settings [9]. It can uncover the underlying knowledge, skills and structures of task performance by characterizing the decision making and reasoning skills and information processing needs of subjects as they perform tasks [10]. Compared to other methods for system design and assessment, CTA is superior for its predominance in analyzing both physical and mental procedures in a task. There are various methods to conduct CTA. In this paper, we use GOMS – a goal–subgoal structured model to assess the cognitive complexity and task performance in AHLTA.
- 2.
Distributed cognition: Distributed cognition is the theoretical development in the distributed system approach [11], [12], [13]. Recent studies by Patel et al. extended the distributed cognition approach to the medical domain [14]. Distributed cognition emphasizes the inherently social and collaborative nature of cognition and also characterizes the mediating effects of technology or other artifacts on cognition [10], [13]. The paper focuses on analyzing how human cognition factors are distributed across the task performance in AHLTA system, using the Extended Hierarchical Task Analysis method developed by Chung et al. [15]. In this study, distributed cognition is constrained and characterized as a process of coordinating distributed internal (mental) and external (physical) representations and factors in the interaction between physicians and the EHR system. In the task performance, physicians perform mental actions (i.e., retrieving information from memory, making a choice) and also physical actions (i.e., scanning a patient list, writing a note). External representations can minimize the difficulty of a task by supporting recognition-based memory or perceptual judgments rather than recall [16].
Section snippets
Methods
This section describes the overall approach used to evaluate the user interface of AHLTA. First, GOMS analysis was performed to identify all the sub-tasks of a given task and to classify them into mental or physical operators. Inter-rater reliability was then calculated to determine agreement between two evaluators who independently conducted GOMS analysis for each task. Finally, one associated GOMS analysis technique – KLM – was used to predict the execution time required to perform each given
Results
As shown in Table 2, the total number of steps for a given task ranged from a minimum of 41 for “Review Coding of Medical Encounter” to a maximum of 466 for “entering vital signs”. The mean number of steps for a task was 106.
Steps for each of the 14 tasks were further classified as either mental or physical operators depending upon their cognitive distributions based on the GOMS classification. The results show that of the total operators, 37% of the steps were mental.
The second half of Table 2
Discussion
This paper identified the following factors related to the performance of AHLTA: (1) large number of steps to complete a task, (2) long execution time and (3) high percentage of mental operators. The inter-rater reliability analysis showed that the method used in this study can be reliably used for evaluating the usability of the health information systems. The results demonstrate that the user has to perform around 106 steps on an average to complete one prototypical task. Secondly, the
Conclusion
This study investigated the current user interface of AHLTA based upon the two important user concerns obtained from discussion forums, i.e., large number of steps to perform a task and mental workload. GOMS model was applied to classify the steps as mental (internal) operator or physical (external) operators. This study suggests that the amount of work to be done to perform basic tasks is deterrent to use the system and the key to improve the user interface is to reduce the overall number of
Contributions
Himali Saitwal performed GOMS analysis on all tasks as a co-analyst, calculated inter-rater reliability, derived time-calculation equations for KLM analysis, performed KLM analysis on all tasks, generated final results and discussed pros and cons, and wrote the original draft of the paper. Xuan Feng performed GOMS analysis on all tasks as a co-analyst, wrote CTA and distributed cognition related text, and suggested changes to improve the current user interface. Muhammad Walji guided on KLM
Acknowledgements
This study was supported in part by US Army Telemedicine Advanced Technology Research Center (W81XWH-08-2-0025 and W81XWH-07-2-0108). We would also like to thank Ron Gimbel for providing valuable comments; Brian Bowers for providing the list of tasks for task analysis; and the program officer, Nancy Connolly, for her support.
References (19)
Human-centered computing in health information systems. Part 1: analysis and design
Journal of Biomedical Informatics
(2005)- et al.
Operationalization of the UFuRT methodology for usability analysis in the clinical research data management domain
Journal of Biomedical Informatics
(2009) - et al.
Representations in distributed cognitive tasks
Cognitive Science
(1994) - et al.
Translational cognition for decision support in critical care environments: a review
Journal of Biomedical Informatics
(2008) - et al.
Representations in distributed cognitive tasks
Cognitive Science
(1994) AHLTA under attack. EHR system stays despite blog complaints: defense
Modern Healthcare
(2008)- T. Philpott, Doctors see bigger role in electronic health record reform, 2009, available at...
- DoD Electronic Health Record System is a headache for military doctors,...
- et al.
UFuRT: a work-centered framework and process for design and evaluation of information systems
Cited by (67)
Cognitive task analysis and workload classification
2021, MethodsXA cognitive decomposition to empirically study human performance in control room environments
2020, International Journal of Human Computer StudiesCitation Excerpt :Clark et al. (2008) identified 5 steps common to most methods: 1) Collect preliminary knowledge; 2) Identify knowledge representation; 3) Apply focused knowledge elicitation methods; 4) Analyze and verify acquired data; and 5) Format results for intended application. CTAs have been shown to be useful in various domains, including medicine (Fackler et al., 2009), instructional design (Tjiam et al., 2012; Tofel-Grehl and Feldon, 2013), health-care systems (Saitwal et al., 2010), and autonomous vehicle engineering (Papautsky et al., 2015). The classification of tasks can help in generalizing CTA results across cases (Rantanen et al., 2006).
Design for improved workflow
2020, Design for Health: Applications of Human FactorsNavigation in the electronic health record: A review of the safety and usability literature
2017, Journal of Biomedical InformaticsUser Experience Research and Usability of Health Information Technology
2024, User Experience Research and Usability of Health Information TechnologyEstimating Cognitive Load in a Mobile Personal Health Record Application: A Cognitive Task Analysis Approach
2023, Healthcare Informatics Research