Keywords

1 Introduction

In 2013, The World Economic Forum reported that “the greatest risk of hubris to human health comes in the form of antibiotic-resistant bacteria” [1].

Every year, approximately 100,000 Americans, 80,000 Chinese and 25,000 Europeans die from hospital-acquired antibiotic-resistant infections. Hospitals are often characterized as infection transmission systems where a considerable proportion of patients are infected with antibiotic-resistant bacteria. Low compliance with infection prevention evidence-based practices is one reason that five to ten percent of patients admitted to hospitals acquire at least one healthcare-associated infection [2, 3]. The general goal of infection prevention is to “minimize the risk of individuals acquiring infections during the course of care by preventing the transmission of infectious agents” [4]. An important piece of information from the infection prevention perspective is awareness of patient’s carriage of antibiotic resistant pathogens.

A situation awareness-oriented system design may improve healthcare workers’ knowledge about daily aberrations in infection burden and the risks associated with infection transmission and underuse of preventive interventions. Studies have shown that situation awareness correlates with higher performance [5–7], better clinical sense-making and improved patient outcomes [8–11]. However, there is a large gap in the field of healthcare context representation as no recommendations are available about the functional needs of the context. There is also a gap between fundamental research on context representation and actual context awareness prototype [12].

The objective of this research is to create a graphical representation of the infection prevention contextual data by applying the Rules of Image Construction, grounded into Matrix Theory and the physiology of a retinal perception [13]. The goal of this construction is to enhance clinical sense-making in a distributed group about the changing risks of infection transmission and spatially-linked infection prevention activity at a hospital unit level. The study graphical model should answer the question “At a given location, what are the risks of infection transmission there?” in an instant of vision. This research postulated that contextualization of the infection prevention data may provide critical cues for health care workers’ decision-making and detecting situations when prevalence and variety of antibiotic-resistant carriers reach dangerous thresholds. The study obtained the Institutional Research Board approval.

2 Effects of Data Representation on Human Performance

Ineffective data representation in electronic health records creates problems resulting in cognitive complexity [14–18]. Cognitive complexity is defined as activities related to identifying, perceiving, remembering, judging, reasoning, deciding, and planning [19]. In spite of years of research on human-computer interface, there is still a need to manage the information effectively in order to enable healthcare practitioners to gain a high level of understanding quickly. Representing information as a graphic is a form of information processing where a vast amount of data can be reduced to understandable and memorable information. Understanding of the graphically presented information can result in visual memorization, but there are the conditions of memorization; as the number of images and the amount of information increase, memorization becomes difficult. Cognitive psychology has described the difficulty with holding more than seven items in short-term memory [20]. Thus, effectively presented data will enable humans to interpret vast amounts of data, while ineffective data representation needs to be resolved.

A landmark study investigated memory for photographs and found that the performance on the recognition of 2,560 pictures, which were displayed for 10 s, exceeded 90 percent [21]. In healthcare, earlier studies explored that metaphor graphic offers a new form of medical knowledge representation [20, 22]. Metaphor graphic is defined as assemblies of icons for graphical representation of symptoms, signs, pathological situations, some components of diagnoses. A randomized trial on the effects of text, table, pie chart, and icon, on the efficiency of subjects’ assimilation of information identified that icons were superior to the other formats in speed (p-value < 0.001) and accuracy (p-value = 0.02) [23]. The researchers concluded that icons are a valuable representation of medical information. Other studies found the icon-based graphics were more effective than numerical formats in increasing risk-avoidant behaviors in patients [24–27].

3 Matrix Theory and Visual Perception

Experimental psychology explains that human visual perception interacts with the ability to understand and memorize the forms within an image [13]. The matrix theory of graphics is the application of this property of visual perception [28]. An image has three independent dimensions, X, Y, and Z. The eye perceives two orthogonal dimensions X and Y, while a variation in light energy produces a third dimension in Z. The third component necessitates the use of visual variables, such as color, texture, shape, orientation, and symbol, to enable visual ordered perception. Transcribing a set of information into an efficient three-component image depends on the application of the factors that enable human associative, selective, ordered, or quantitative visual perception. While the plane possesses all of these properties, the retinal variables hold only some of them.

Bertin emphasizes that the retinal variables are physiologically different from the planar dimensions [13]. The knowledge of the eye physiology explains what makes visual perception instant or non-instant. The reader perceives the planar dimensions through the intermediary of eye movement, so-called “muscular response”. In contrast, the retinal variables, inscribed “above” the plane, are independent of it. The eye perceives the retinal variation without eye movement. The retinal perception is called “retinal response.” Therefore, the retinal variables, such as size, value, color, texture, orientation, and shape become indispensable in the creation of the efficient image.

4 The Rules of Image Construction

4.1 Graphical Information Processing

To choose the right graphic formula for a set of information, a designer should identify the purpose of graphics, consider the number of concepts (components), their nature, the presence of a geographic component, and determine the most efficient image construction. Efficiency of the image is defined as “the most efficient construction that enables a person to answer any question in a single instant of perception” [13]. When one construction requires a shorter observation time than another construction to answer a given question, this construction is more efficient. A graphic can furnish the means of retaining information with the help of visual memory. The primary graphical problem encountered is identifying the best degree of data simplification that will still provide the substance for decision-making. This problem links to issues of visual selectivity and conceptual complexity of a graphic: each additional component increases the conceptual complexity and requires a new visual variable leading to an increase in visual variability and reduction in memorability of the graphic.

In order to reduce a set of comprehensive information to a meaningful minimum, the information elimination and processing are required. Reducing the number of correspondences and keeping essential ones can simplify the information conceptual complexities to a degree when an individual is able to retain information with the help of visual memory. The process of elimination of some information makes the image less comprehensive but easier for inscribing information in the viewer’s memory [13]. The mechanism of visual ordering and classing may decrease the overall information and increase the speed of discovering the groupings. Superimposing or overlaying several images in a figuration is the additional method of information reduction and simplification. Image construction involves the analysis of the transcribed information for identifying (1) an invariant defined as a component that is common to all the data, (2) the number of components, (3) the number of elements in each component, (4) the level of organization of each component (nominal, ordered, and quantitative), and (5) planar and visual variables that match these components.

4.2 Transcribing Information into Graphics, Length of the Component, and Graphic Processing of Information

The complexity of the image depends on the number of components and elements in each component. Short components, including up to four divisions, reduce visual variability and conceptual complexity of an image. “Long” components, when a number of divisions exceeds fifteen, not only increase the conceptual complexity of the image but also create a challenge for graphical representation. The visual perception is important for comparing the characteristics, discovering similarities and differences, and identifying areas of interest or exceptions. Therefore, the visual variables representing each component must permit visual selectivity, associativity, and ordering. The hierarchy of the visual variables that permit selectivity starts with the use of size and value, as the top choice, followed with color, texture, and orientation. Shape has no selectivity; however, it provides a base for symbolism. Selectivity is also enabled when characteristics are superimposed. The effectiveness of the chosen visual variables is measured by its capacity for enabling the reader to disregard everything else. In order to choose the most efficient retinal variables, it is important to determine the level of organization of each component and the length of each component. When the length of a visual variable matches the length of each component, diagrams and maps become visually retainable.

5 Methods

This paper is reporting research-in-progress. We present the first phase of our study: the conceptual model for infection prevention situation awareness model (IPSA). In phase two: the pertinent questions we intend to research are: To what extent can the 50-bed medical-surgical unit population IPSA information be graphically reduced? What is a meaningful minimum of information to be retained? What methods of information simplification can be used? The design process included: (1) development of an infection prevention conceptual model grounded in the epidemiology of nosocomial pathogens, (2) integration of the empirical data from the EHR for an analysis of the information complexity and the corresponding levels of visual variables, and (3) construction of the graphical interface.

5.1 A Conceptual Model for Situation Awareness Oriented System Design

In order to design a situation awareness-oriented (SA) system, it is important to develop a clear understanding as to what supporting SA means in a particular domain. Therefore, conceptualization was the first step in the image construction. Contextualization of the infection prevention data may provide critical cues that would capture health care workers’ attention during serious situations when prevalence and variety of antibiotic-resistant carriers reach dangerous thresholds. For this, an interface designer can utilize the knowledge developed in the epidemiologic risk models for predicting the movement of infection through populations. During the first phase study, the knowledge from infectious diseases was synthesized and translated into a conceptual model consisted of the following: Biological domain-antibiotic resistant agent (bacteria), patient infection state, hazard zone for infection transmission, infection burden, member at risk of exposer to infection, and allergy; Non-biological domain-social (contact repetitiveness), structural (location, proximity); Behavioral domain- receipt of intervention; and Temporal domain - time to delivery of intervention (Fig. 1).

Fig. 1.
figure 1

The infection prevention situation awareness conceptual model

Full size image

5.2 Study Setting and Data Sources

For the design phase, the study setting chosen is a 50-bed medical-surgical unit. The unit patient population consists of solid organ transplant patients (45 %), cancer patients (20 %), internal medicine patients (20 %), and general surgery patients (15 %). Each room is occupied with one patient. The first step to create a visualization was to abstract data elements from the distributed data sources. We have built the electronic report, an XLS file in the electronic health record system (EHR). The report abstracts patient’s name, medical record number, date of admission, date of discharge, date and time of a patient hygiene note, authorship of the note, the note content, antibacterial medication administered to a patient, and the infection surveillance data recorded during a 24-hour period in the EHR. All patient personal data is de-identified. The EHR data was analyzed on availability of data that corresponds to the conceptual model informational needs. It is important to clarify that the model represents a “container” that can be filled with the varying “content” depending on available knowledge, technology, quality of data, and users’ preferences. For example, “frequency of contacts” can be determined by using high radio-frequency wearable devices [29] or textual data indicating that a hospitalized patient requires maximum assistance. When the data was normalized, a decision tree (Fig. 2) was constructed to understand the daily aberrations in patient infection states, such as 1) the carriers of antibiotic resistant bacteria (ABR) who receive antibiotics IC + , 2) the ABR carriers who do not receive antibiotics I-C + , 3) the non-carriers who receive antibiotics I + C-, and 4) non-carriers who do not receive antibiotics IC-, for a period from June 7 through June 13, 2014.

Fig. 2.
figure 2

The study data elements, concepts, and empirical probabilities

Full size image

6 Image Construction Process: The Component Analysis

6.1 Defining the Number of Components and Invariant

Following the rules of image construction, we determined the components and invariant for the graphic. A process of data transcription requires a separation of “content”, or the information to be transcribed, from the “container”, which represents only the properties of the graphic system. In the first phase, the original content (context) included 16 concepts for the analysis of the data of interest. The presence of the geographic components, such as “location” and “proximity”, and the spatially-linked concepts, such as “risk of exposure” and “hazard zone”, informed the investigators that the most useful graphical construction would be a map that represents the unit’s physical layout. In this study, the unit geographic order becomes the invariant. The geographical element (e.g., a distinct geographic space representing variation of the locations) is the ward of the unit. The second component is the “common circuit” - a common area where the doors of two wards open out. This component includes two elements: “contaminated circuit” when, for example, one of the wards is occupied by an ABR carrier. The second element is “not contaminated circuit”. After a series of reviews, it was decided to keep the eleven essential components (Table 1). The elimination of some information would lead to constructing a graphic permitting visual memorization. This image can be customized to different interventions including various infection prevention and control activities.

Table 1 The summary of the graphic components and graphical information processing methods.
Full size table

6.2 Identifying Visual Variables Permitting Best Selectivity

The next step in our conceptual design process was the creation of the visual artifacts for the 11 components of the IPSA informational set and the continuous revision of the meaning of this information. By drawing different sketches and experimenting with the different visual variables, the investigators tried to identify the best variables that would reduce visual variability, permit visual selectivity and associativity, and reduce the conceptual complexity while preserving the original meaning. This analysis has conceptualized a set of the visual artifacts described in Table 1. It is recognized that ”Patient Infection State” permitted a creation of an independent visual component “Colonized” (a carrier of ABR) consisted of two elements “C+” a carrier of ABR vs. “C-“non-carrier. To permit visual selectivity, the most clinically significant but least prevalent phenotype, “C+”, was visualized with the use of red color, and the non-carriers “C-“, with the gray color. Each ward occupied by “C+” vs. “C-“patients will be color-coded correspondingly. The concept “Infected” has originally included two elements: a receipt of antibiotics “I+” vs. no receipt “I-“. The use of a texture for the element “I-“, which would retain the background color of the component “Colonized”, deems beneficial for several reasons.

First, such visualization reduced visual variability of the image, makes the most prevalent and a benign phenotype “I-” less salient, permitting a better selectivity for the phenotype “I+”. In addition, the use of texture permits a perceptual associativity when a reader can easily associate the sub-groups “I+” or “I-” among the phenotypes “C+” and “C-”. This approach reduced the original 4-division component “Patient Infection State” into two short components, including a two-division component for “Colonized” and a one-division component “Infected”. The components that are spatially linked to the “Colonized” patient’s location, such as “contaminated circuit” and “hazard zone”, inherited the red color of “Colonized” patients in order to enhance the visual selectivity. The component “Hazard zone: significance of risk of infection transmission” includes the three elements, such as significant, moderate, and low. The use of the retinal variable depicting “size” enables selectivity. “Size” intuitively corresponds to the amount of bacteria produced by patients and, respectively, the significance of infection transmission risk. As a result, this visualization eliminates a need for the users to calculate this measure and increases the speed of their comprehension.

The component “High Contact Patient” denotes a group of patients who require maximum assistance and experience very frequent contacts with healthcare workers. Frequent contacts increase the risk for infection transmission among members of a population. This component has one element represented with a symbol, a yellow dot. This visual artifact is planned to be superimposed over the patient’s location. In this study, “Receipt of an infection prevention intervention” is a component representing a receipt of chlorhexidine bathing. It includes only one element “No receipt of chlorhexidine bathing” and is represented with a symbol, a red circle. The red circle “informs” the reader about a lack of specific infection prevention intervention, allowing healthcare workers to recognize the underuse of the preventive intervention. The red circle and the yellow dot can be superimposed without overlapping each other. This information processing mechanism permits a visual selectivity for identifying a sub-group of patients who experience frequent contacts and lack infection prevention. When these two components are superimposed on the patient’s phenotype “C+” they create a critical cue to healthcare workers that these particular patients may increase the risk of infection transmission. The combination of the patient’s phenotype, contact frequency, and receipt of the infection prevention intention may provide a strong signal for actions, e.g., reinforcing the compliance with the existing policies or developing new tactics for specific situations. Finally, the concept “at risk of exposure” does not necessarily need visualization; a filter can be used to interactively present this group to the users. Figure 3 presents the visual cues explaining the most important IPSA concepts developed for the graphical design discussed above.

Fig. 3.
figure 3

The infection prevention situation awareness visual artifacts for user training

Full size image

7 Proposed Graphical IPSA Interface

As a result of the IPSA design process, a demonstrational unit for a given day is shown in Fig. 4. The infection prevention contextual information was transcribed into a cartographic message and implanted in the following structure: the invariant – a geographic order, which takes two orthogonal components (XY), and the nine components (Z) represented with the retinal variables, such as color, texture, size, shape, and symbol.

Fig. 4.
figure 4

A demonstrational unit

Full size image

8 Discussion

The development of an interface that would enable healthcare workers to comprehend the risks of infection transmission, properly allocate limited resources, and develop tactics maximizing the benefits of infection prevention and control in specific epidemiologic situations is desirable. Such an interface would be acutely important in emergencies when the intensity of work and monitoring needs rapidly increase. Mapped data provides instantaneous answers, making the groups and potential explanations appear with exceptions. The analysis of the empirical data helped the investigators to understand the scope of the meaningful minimum of informational to be remained, which deems sufficient for decision-making at the level of unit daily management. The next step in this research project will evaluate if the healthcare workers’ situation awareness increases with the use of the IPSA graphical interface in comparison with the current practice. The cartographic message aims at enabling the reader to locate high-risk for infection transmission patients, regarded as high priority for infection control services; to identify patients who are at high risk for exposure to pathogens, regarded as high priority for infection prevention services; to recognize the areas where the risk of infection transmission is significant, regarded as hazardous environments, and, ultimately, to assess the infection prevention needs in the context of these risks for work planning, patient arrangements, resources allocation, or targeted monitoring of compliance. The contribution of this research is the development of an innovative IPSA-oriented interface, a new form of medical knowledge representation where spatially-linked clinical data can be used for spatial decision-making in hospitals.