© 2018 by Triaged Data Display, LLC

|Web Design by Mark Husbands|

The goal of the present work is to improve the cognitive ergonomics for health care providers.  The  widespread  adoption of EHRs presents the provider with an ever escalating cognitive load from the ever increasing volume  of  archived  patient  data. An improved EHR may help lighten this cognitive load. The starting point of this work was the observation that optimal data visualiza- tion compared to tabular displays improves access to data and improves patient care delivery [13,14].

Edward Tufte in his classic study of the history of graphic dis- plays in print format stated that ‘‘graphical excellence is that which gives the viewer the greatest number of ideas in the shortest time with the least ink in the smallest space” [15]. With Tufte’s perspec- tive as background there have been many visual display systems implemented, often in the monitoring of health care and often  for use as heuristic tools. Scatter plots have been used in a visual display to monitor changes in multiple patients for a small number of characteristics over time [16]. With a hierarchical tree or tree map design a display has been used to monitor the quality of health care delivery [17,18]. Another data display system with tree map-based icons uses clustering to detect common features among a heterogeneous population to define more precisely different groups of patients [19]. Spider plots have been used with individ- ual patient data in the ICU against the historical patient data of many patients and many parameters to predict outcome [20]. A bubble chart design has been used to follow the quality of care adherence for colorectal cancer patients [21]. One limitation of these designs for use by practitioners in health care delivery is that only a limited number of parameters can be tracked (e.g. x-axis, y-axis, bubble size, and colour of bubble). This limitation exists also for data displays using histograms and contour maps [22].


A recent comprehensive review of EHR data visualization iden- tified several challenges to be overcome for optimal data visualiza- tion and health care delivery [23]. These challenges include: clutter-resulting from the size and complexity of the data, the difficulty of presenting a great deal of data on a single screen and compliance, the time it takes practitioners to learn to navigate a display system in an unfamiliar display format [23]. A successful display format for health care delivery must meet the twin chal- lenges of embracing data complexity while resisting display complexity.

The design simplicity of the dashboard may allow for these twin challenges to be met. A clinical dashboard ‘‘enables easy access to multiple sources of data being captured locally, in a visual, concise and usable format” [24]. To this point most medical dashboards have been used much more in patient care monitoring rather than in the improvement of the real time delivery of patient care. A recent review of the use of dashboards to improve medical care found 543 citations [25]. Of these citations, only  11  full  reports had data on the use of dashboards to improve medical care. Among the 11 reports only 1 report was assessed to be of high quality [26]. The high quality report used colour-coding to track level of compli- ance in real time with a patient management protocol.

For the present study the novel display design had to meet the requirement for the provider to handle correctly and in real time a huge volume of patient results. The display needed to provide the provider a ‘‘vectored alert” to the fraction of those results that are critical and require immediate provider acknowledgement.

New display design approaches may help providers manage the novel challenges in data acknowledgement posed by the comput- erized EHR. To that end Wickens et al. proposed thirteen principles of display design for best human–computer interaction [27]. These principles are: (1) legible displays, (2) avoid single variable (rather a dynamic range/analog display), (3) top down processing

(famil- iar/constant format), (4) redundancy gain (present data more than once), (5) similarity causes confusion-use clearly distinct elements, (6) pictorial realism (make display look like variable that it repre- sents), (7) principle of the moving part (moving design features should move as the measured element moves), (8) minimizing information access cost (convenient to use), (9) proximity compat- ibility principle (related items near each other but avoid clutter), (10) principle of multiple resources (different sensory input, as audio and visual), (11) replace memory with visual information: knowledge in the world (calibrate the need for background infor- mation on the basis of the user’s subject familiarity), (12) principle of predictive aiding (including information about possible implica- tions of data), and (13) principle of consistency (invariant design). Using the principles of Wickens et al., with the goal of improv- ing patient care delivery as well as optimizing health care resource utilization, we created a novel data visualization dashboard design that ‘‘minimizes access cost” (principle 8) and minimizes the num- ber of clicks. The design adheres to the principle of ‘‘multiple resources” (principle 10) in displaying reports from different EHR systems in their original format but representing them in the invariant novel display. Adhering to the principle of ‘‘predictive aiding” (principle 12) critical and non-critical reports are separate in the display. The novel display design is invariant and thus consistent with principle of consistency (principle 13). As mentioned before, the current data visualization designs share a common feature. With increasing data these designs become increasingly complex. This aspect of these designs violates Wickens et al. principles (9) avoiding clutter and (13) having an invariant design. We chose, instead, a display of absolutely invari- ant design. An invariant design allows an unlimited number of reports to be queued or stacked and the ‘‘above” view can still be intelligible. To minimize the hierarchical character of the data structure and the resultant time required for data discovery, we attempted to maximize the ‘‘active” surface of the display as a fraction of the total text on the screen. All reports in the data set  are enumerated in ‘‘buttons” by data category. This structure enables each click of any ‘‘button” to the display the first and then progressively all subsequent and underlying reports in the queue (or stack).


The current display partitions data into reports with non-critical reports represented on the circumference of the circle and reports with critical data partitioned and distributed around the outside of the circle. The display can be the basis for work triage. In the present case clinical reports with critical results that lie outside the circle are be dealt with immediately. The reports with non-critical results can be either dealt by the provider with at a later time or triaged to another member of the health care team. This display with its binary partition of reports is a model that can be applied in all other areas in health care where reports with critical results occur among a vast excess of reports with non-critical results. Areas such as compliance and accounting are examples.


At present, a major problem in EHRs is the lack of compatibility across different EHR systems. The novel design with its invariant display represents all reports both those with critical and those with non-critical results on a single screen. This structure enables the single screen display of all the reports from one or many EHR systems regardless of the primary EHR software. The requirement for a robust two way interface to the different EHR systems is the only modification needed for its deployment. The novel display design represents all reports but displays the actual report in the format of the original EHR system.

The current study is formative, and is limited to a test of labo- ratory data only. The value of the current display will be much clearer after it has undergone a real time  test  in  a  functioning EHR (see Section 5).