| The meaningful visualization of data was of primary importance to this project. Our development group needed to create a display that assisted in drawing conclusions about resource utilization. Our visualization was grounded in two areas of expertise: Edward Tufte’s (1990) core principles of Information Visualization and Colin Ware’s (2008) work in Visual Design heuristics derived from studies in cognitive science. Central to our graphical design philosophy was basing any design decisions on the limitations inherent in Human perception. Cognitive research shows that a person’s perception is limited to storing a mere three elements in “visual working memory” (Ware, 2008, p. 10). Understanding this as a first principle, we systematically designed our graphic to address this limitation and to encourage simple interpretation of the displayed information by our user group.
Our graphic presents three layers of information: the time of the scan, the type of scan with which it can be associated, and the resource that completed the scan.
By layering these three areas of information (Figure 1), we were able to assist in deriving meaning from the specific data sets, thus “visually stratifying various aspects of the data” (Tufte, 1990, p. 53). This “visual stratification” allows for an observer to easily ascertain context of the graphical segment in question as it relates to the totality of the data presented.
Small Multiples, defined by Tufte (1990) as “visually enforcing comparisons of changes” (p. 67), were utilized in the representations of each instance of a scan. The various spatial qualities of each of these glyphs allow for quick visual comparison to the other elements displayed, and creates yet another avenue for one to decipher meaning (Figure 1). Color was used to classify types of scans, and adds yet another layer of information to the display. The color palette of the background was chosen carefully, so as not to distract the reader of the graphic and lose the visual significance of the chronological data stored in the columnar elements of the display. “Color Clutter” was avoided (Tufte, 1990, p. 82), where strong colors were reserved for the scans, but only if their presence did not distract from the chronological information that provides context in the background of the display. Finally, the ability to document specific events during the twenty-four hour period was included. This feature allows users to communicate information about a specific scan that may not be easily decipherable in the graphic. For example, an extremely long procedure, one that encompasses the entire twenty-four hour period, could be annotated with this feature to better communicate the context around the scan in question.
The graphic is wrapped in a web based dashboard built using the Ruby on Rails programming language and framework. A high level view of volumes over time is given to the user with the ability to filter by modality type and time period. The data points on each graph can be clicked on to drill through to the information which they represent. As stated above, the initial level of granularity is a specific twenty-four hour period of time. This diagram provides a chronological snapshot of all the scans that took place, delineated by resource. The user can click on each scan to see the information about the procedure(s) that make up the scan, including exact time stamps and the performing technologist (Figure 2). Further documentation can be provided by filling out a web form and submitting the information to the specific period of time that is being viewed (Figure 3). To gain an accurate and complete picture of scanner utilization across multiple modalities, data were collected from the radiology information system (RIS) and the picture archiving and communication system (PACS). The extraction, transformation, and loading (ETL) of these systems occurs in two phases. In phase 1, an ETL process occurs nightly to gather new studies and update data from previously gathered studies. Using heuristics, multiple modalities are combined into a single resource and categorized into a modality type. In phase 2, an ETL process occurs every hour to gather new studies to provide tactical information for the utilization dashboard.
|
| The Scanner Utilization Web Tool has become a critical resource for departmental leadership. Total access to the tool by user is significantly higher than most other departmental applications (Figure 4). Retrospective analysis and contextual inquiry into departmental meetings has shown an increasing culture of reliance in data to drive strategic and tactical discussion about resource usage. The documentation tool has not been widely adopted by the user base. There are many possible explanations for this. Cases that need to be accounted for are probably not as prevalent as those that do not. A dedicated analysis of this phenomenon has yet to be conducted, but it is possible to extrapolate this meaning from a law of averages. It is possible that users fear a punitive measure for documenting an outlying case; again, this would have to be quantified by contextual inquiry and analysis into the data about documentation actions. Finally, it is possible that the design of the graphic does not support the right kind of documentation for this display. The documentation tool does not allow for the ability to mark-up a graphic, or point text to specific graphical elements. Therefore, there could be a disconnect between what the graphic is representing, and deciphering what an annotator meant by a comment about a specific case. Clearly, the tool’s usage statistics define wide acceptance of the core informative features of the Scanner Utilization web application.
|