“We failed to anticipate Pearl Harbor not for want of the relevant materials, but because of a plethora of irrelevant ones.” Attributed to Roberta Wohlstetter
Clinicians can obtain data with incredible ease—unlike the days when, as a medical student, my job was to go down to the laboratory in the afternoon to search the handwritten log, listed in order of specimen receipt rather than patient name, and hand copy the test results for each of the patients my team was treating. Over time, our access to data has improved and it is now available—literally—at our fingertips. However, the volume of data to be retrieved, interpreted, and synthesized has grown monumentally and can create data overload that shifts our challenges. Clinicians are confronted by unrelenting streams of data about individual patients, including written or verbal observations and reports, ordinal data, images, point-in-time numeric laboratory values, and continuous waveform data,1 as well as data about the logistical management of groups of patients (i.e., whiteboards and operating room schedules). During rounds on critically ill patients, clinicians may be confronted with more than 200 variables.1,2 In addition to thresholds, clinicians need to understand trends and complex combinations of factors.1 Electronic health records have increased access to data, but have also increased its volume, such as by adding metadata to many displays. Our ability to acquire data often exceeds our ability to fully understand and integrate it; this information overload can lead to preventable medical errors.1,2
One strategy to manage the abundance of data is to use established best practices for effective data display, because the way information is represented significantly affects problem solving.2 Short-term memory quickly becomes overloaded and degrades when searching multiple data display screens; improved search time and efficiency may free cognitive resources that can be used for other interpretive or planning tasks.3 Understanding and applying principles of data visualization can improve our ability to access and interpret data to provide safe patient care.
Whether visual data is presented in electronic or paper format, its abundance can contribute to cluttered data display, and clutter contributes to attentional and performance costs.4 As clutter increases in visual displays, search time tends to increase while accuracy tends to decrease.4 Clutter contributes to distraction, uncertainty, and confusion.4 Clutter degrades object recognition and detection, information interpretation, and detection of unexpected events.4 Most of the displays Kamaleswaran and McGregor analyzed in their review of visual representations of physiologic data contained more than 20 variables per screen.5 However, information that is cluttered to one user may be meaningful to another. Clutter is not solely related to the number of items in a display; clutter is also impacted by factors including item density, arrangement, color, display organization, structure, order, conceptual grouping, background noise, and task relevance.4
Kamaleswaran and McGregor5 describe four types of visual display for physiologic data; familiarity with their concepts may help efforts to improve the effectiveness of data display. First, tabular or text displays include tables and may mimic traditional flowsheets5 that provide data about a series of variables (i.e., temperature, laboratory test results) over time, using text and numbers. Second, graphic displays may use waveforms to display data,5 such as real-time presentation of cardiac rhythms on vital signs monitors; some present complex data derivations.6 Third, object-oriented displays manipulate two-dimensional graphical characteristics such as color, size, and shape to produce dynamic representations of changing properties or characteristics that emerge from integrating information from combinations of sources.5 Distinctive patterns or visible changes—such as a flashing visual signal—may activate pre-attentive processing,5 so that the observer is alerted. Recognizing that a change has occurred precedes interpreting the specific information being presented. Finally, metaphoric displays include images related to the organ system associated with the data, such as a series of pipes and baffles representing blood flow into, through, and out of the heart5,7 or a dynamic image of lungs with varying volumes of inspiratory and expiratory fullness during different phases of respiration. Graphic and integrated displays have been shown to decrease response time, improve recall, and improve user satisfaction compared with traditional text displays.2
Optimal data display depends in part on the eye of the beholder. Information that is “signal” to one person may be “noise” to another. Data display issues include
Data display needs and preferences can vary based on the task, the user, and the situation.4
Information displays that enhance agreement among individuals about current physiologic states may better support continuity of care across personnel and across work shifts.3
Ideally, data displays integrate goals directly with information needs and also represent these relationships over time.3 Different users may have different goals, related to patient care, process improvement, or other purposes.
Eventually predictive modeling may be able to represent the patient’s anticipated future state, based on current physiologic and treatment parameters, and compare this to the patient’s goal or desired state.
Following an iterative, human-centered design method1,2 during the development of data visualization can help manage the potential for competing requirements and unintended consequences. Moacdieh and Sarter’s review article, addressing research on clutter primarily from fields outside of healthcare, provides synopses of a variety of measurement approaches and techniques that could be applied to help improve healthcare data display, including performance evaluations (i.e., search time and accuracy), subjective assessments, and eye tracking metrics.4
Nonvisual data presentation formats, such as audible alarms, can also be valuable. In the operating room, pulse oximeters use the rate and tone of audible emittances to provide dynamic information about a patient’s heart rate and oxygen saturation that can be understood without requiring clinicians to divert their visual attention from procedural tasks. However, similar to visual information clutter, excessive or inconsequential audible alarms can be problematic.
The challenge in managing the potential for data overload is to determine the ideal middle ground between excessive data and insufficient information and to arrange the relevant information in a manner that supports clinicians’ cognitive processes.1,2,4 Understanding the relevant established best practices will help us design and implement data display in a manner that contributes to efficient, effective application to patient safety.
Citerio G, Park S, Schmidt JM, Moberg R, Suarez JI, Le Roux PD, Second Neurocritical Care Research Conference Investigators. Data collection and interpretation. Neurocrit Care. 2015 Jun;22(3):360-8. Also available: http://dx.doi.org/10.1007/s12028-015-0139-4. PMID: 25846711 Schmidt JM, De Georgia M, Participants in the International Multidisciplinary Consensus Conference on Multimodality Monitoring. Multimodality monitoring: informatics, integration data display and analysis. Neurocrit Care. 2014 Dec;21 Suppl 2:S229-38. Also available: http://dx.doi.org/10.1007/s12028-014-0037-1. PMID: 25208675 Miller A, Scheinkestel C, Steele C. The effects of clinical information presentation on physicians’ and nurses’ decision-making in ICUs. Appl Ergon. 2009 Jul;40(4):753-61. Also available: http://dx.doi.org/10.1016/j.apergo.2008.07.004. PMID: 18834970
Moacdieh N, Sarter N. Display clutter: a review of definitions and measurement techniques. Hum Factors. 2015 Feb;57(1):61-100. PMID: 25790571
Lowe A, Jones RW, Harrison MJ. The graphical presentation of decision support information in an intelligent anaesthesia monitor. Artif Intell Med. 2001 May;22(2):173-91. PMID: 11348846
Agutter J, Drews F, Syroid N, Westneskow D, Albert R, Strayer D, Bermudez J, Weinger MB. Evaluation of graphic cardiovascular display in a high-fidelity simulator. Anesth Analg. 2003 Nov;97(5):1403-13. PMID: 14570658