Stephanie Guerlain

This chapter covers principles of software user interface design that are particularly relevant to avoid: unnecessary workspace navigation, loss of situational awareness, the getting-lost phenomenon, and display thrashing. Unfortunately, software systems are sometimes designed in such a way as to require scrolling or paging when a redesign would avoid such a need. Well-designed software systems are those that were designed with major goal in mind: to effectively and efficiently support intended user tasks. Designing to minimize workspace navigation is of particular importance as users will change their behavior in response to the ease or difficulty with which they can gather necessary information. Poorly designed systems, with multiple operating modes, complex data-navigation schemes, and poor feedback about system state can all lead to predictable errors with their use. People may adopt coping strategies for handling an overly complex system.

Although computers and other automated devices have been designed to play a variety of roles in the medical field, this discussion will focus on the design of decision support systems, specifically those designed to aid human problem-solving using a critiquing approach. The history of critiquing, its proposed advantages and disadvantages, relevant design criteria, and evaluation of such systems will be reviewed. The design of one particular critiquing system is described, along with the results of two empirical studies showing it to be a significantly better alternative to both partial automation of the task and to current practices (no decision support at all).

The consequences of error-prone human performance is to automate as much as possible in a system. Other classic problems with automation include loss of human skill as tasks become automated and brittleness. Manual solution strategies are by no means foolproof, but they are as generic as the problem of how to design for safety when humans and automated agents are involved. Clearly, due to the well-known differences between information systems’ and humans’ strengths, weaknesses, and means to: sense information, make judgments, and execute actions, both the types of information required and the means for gathering and communicating that information will necessarily be different for each type of agent. The goal of representation aiding is to represent relevant domain, task, and system constraints through visual properties of the display, and thus encourage people to perceive these relationships with little cognitive effort. Workspace management refers to the window manipulation, command input, and navigation activities required when working with computer-based systems.

Researchers at the University of Virginia are studying human performance in the operating room in an effort to understand the factors that lead to or inhibit patient safety in this environment. In general, observation ‘in the wild’ is extremely difficult as there is no transcription or video recording upon which to base an analysis or to validate the result. Or, even with such recordings, the work to code and analyse these recordings is extremely labour intensive and thus may not be practical for ‘real-world’ projects that do not have enormous budgets (and time) to accommodate such analyses. We have developed and tested several methodologies that enable collecting team data ‘on the fly’, some of which are ‘summative’ evaluations and others that are more detailed process tracing of events and communications yielding directly usable data that can be analysed to characterize team behaviours immediately following the team process that was observed. We specifically report here on the design of a portable and customizable hardware/software system called Remote Analysis of Team Environments (RATE) that we have used to run two studies designed to improve team communication and coordination in the operating room (Guerlain et al. 2004, 2005, 2008), one handoff-of-care study for paediatric residents (Sledd et al. 2006) and one usability study (in progress). This chapter is based upon work supported by the National Science Foundation and describes how the system works; it has been used in our operating room studies. The hardware enables digitally recording up to four video feeds and eight audio feeds. The event-recording software enables observers to create a customized set of events of interest that can be logged into a time-stamped database when watching a case live. The playback software synchronizes the event-recording database with the audiovisual (AV) data for efficient review of the case.

As robotic systems are integrated into mission operations, they provide key benefits over traditional manned systems such as increased endurance, versatility, and risk reduction for personnel. In order for robotic systems to become fully integrated into United States Army missions, they will need to gain a level of autonomy that is closer to that of human personnel. Under Mission Command, the current system of issuing orders, an Operational Order (OPORD) contains the information required to execute a mission. The structure allows for authors of OPORDs to leave many details of the operation open to the discretion of the reader, making task interpretation and execution difficult for an autonomous system. The goal of this project, and one step of many in the process of automating systems to read OPORDs, is to show how OPORDs can be annotated with relevant information (such as locations, coordinates, and organizations) using the natural language processing techniques of tokenization and named-entity recognition (NER). Using A-fold cross validation of a Conditional Random Field (CRF) sequence model on 9 OPORDs containing 38,551 tokens, we were able to extract entities with an overall precision of 0.702, recall of 0.478, and F-measure of 0.569.

The value of virtual reality simulation for emergency response training and improvement has great potential, but obstacles remain. The utility of simulators is dependent on creating a highly realistic environment. Rather than building a virtual world from scratch, using existing architectural models may be a viable option due to the large quantity of data embedded within three-dimensional architectural designs. However, it can be difficult to maintain and translate these huge and highly detailed models into an interactive, user friendly interface. The MITRE Intelligent Cities project aims to blend methods of data science with interactive 3D modeling tools, using software packages like Revit, 3DS Max, and the Unity gaming engine. The goal of the project is to provide a procedure for visualizing and interacting with the structural properties and metadata of facilities encoded in architectural designs. Using MITRE campus data as a test case, several options were explored to preserve the metadata across software platforms, and a means of easily and efficiently accomplishing this task was discovered. However, limitations occur due to the large size of the model, slowing down the gaming engine and requiring the use of a data subset. This procedure allows anybody with a standard Revit architectural model to visualize their environment in a 3D simulation. Interactivity can be added by creating scripts to highlight key features of the models. These implementations can be used by emergency personnel or facility operators to peer through walls, or to find locations of key facility systems, such as HVAC or data centers.

Anastomotic leak, a potentially deadly postoperative occurrence, particularly interests surgeons performing gastrointestinal procedures. We investigated incidence, cost, and impact on survival of anastomotic leak in gastrointestinal surgical procedures at an academic center. We conducted a chart review of American College of Surgeons NSQIP operative procedures with gastrointestinal anastomosis from January 1, 2003 through April 30, 2006. Each case with an American College of Surgeons NSQIP 30-day postoperative complication was systematically reviewed for evidence of anastomotic leak for 12 months after the operative date. We tracked patients for up to 10 years to determine survival. Morbidity, mortality, and cost for patients with gastrointestinal anastomotic leaks were compared with patients with anastomoses that remained intact. Unadjusted analyses revealed significant differences between patients who had anastomotic leaks develop and those who did not: morbidity (98.0% vs. 28.4%; p < 0.0001), length of stay (13 vs. 5 days; p ≤ 0.0001), 30-day mortality (8.4% vs. 2.5%; p < 0.0001), long-term mortality (36.4% vs. 20.0%; p ≤ 0.0001), and hospital costs (chi-square [2] = 359.8; p < 0.0001). Multivariable regression demonstrated that anastomotic leak was associated with congestive heart failure (odds ratio [OR] = 31.5; 95% CI, 2.6-381.4; p = 0.007), peripheral vascular disease (OR = 4.6; 95% CI, 1.0-20.5; p = 0.048), alcohol abuse (OR = 3.7; 95% CI, 1.6-8.3; p = 0.002), steroid use (OR = 2.3; 95% CI: 1.1-5.0; p = 0.027), abnormal sodium (OR = 0.4; 95% CI, 0.2-0.7; p = 0.002), weight loss (OR = 0.2; 95% CI, 0.06-0.7; p = 0.011), and location of anastomosis: rectum (OR = 14.0; 95% CI, 2.6-75.5; p = 0.002), esophagus (OR = 13.0; 95% CI, 3.6-46.2; p < 0.0001), pancreas (OR = 12.4; 95% CI, 3.3-46.2; p < 0.0001), small intestine (OR = 6.9; 95% CI, 1.8-26.4; p = 0.005), and colon (OR = 5.2; 95% CI, 1.5-17.7; p = 0.009). Significant morbidity, mortality, and cost accompany gastrointestinal anastomotic leaks. Patients who experience an anastomotic leak have lower rates of survival at 30 days and long term.

Recent in-hospital studies of artificial pancreas (AP) systems have shown promising results in improving glycemic control in patients with type 1 diabetes mellitus. The next logical step in AP development is to conduct transitional outpatient clinical trials with a mobile system that is controlled by the patient. In this article, we present the user interface (UI) of the Diabetes Assistant (DiAs), an experimental smartphone-based mobile AP system, and describe the reactions of a round of focus groups to the UI. This work is an initial inquiry involving a relatively small number of potential users, many of whom had never seen an AP system before, and the results should be understood in that light. We began by considering how the UI of an AP system could be designed to make use of the familiar touch-based graphical UI of a consumer smartphone. After developing a working prototype UI, we enlisted a human factors specialist to perform a heuristic expert analysis. Next we conducted a formative evaluation of the UI through a series of three focus groups with N = 13 potential end users as participants. The UI was modified based upon the results of these studies, and the resulting DiAs system was used in transitional outpatient AP studies of adults in the United States and Europe. The DiAs UI was modified based on focus group feedback from potential users. The DiAs was subsequently used in JDRF- and [email protected] transitional outpatient AP studies in the United States and Europe by 40 subjects for 2400 h with no adverse events. Adult patients with type 1 diabetes mellitus are able to control an AP system successfully using a patient-centric UI on a commercial smartphone in a transitional outpatient environment.

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.