Clinically Relevant Data Visualization

From Clinfowiki
Jump to: navigation, search


  • Data visualization is complex with little formal training focused on use in a clinical environment available
  • To add value, data presentation must be actionable and at least maintain overall workload, ideally decreasing workload
  • Data visualization can support decision-making and knowledge discovery for diverse stakeholders including providers, administrators, informaticians, patients, and the public; each party has unique needs for how data are visualized


The practice of data visualization seeks to communicate meaningful information through graphical depiction. Successful visualizations apply a hierarchical interpretation of data to bridge gaps in understanding.[1] Creating easily understood, meaningful graphics within biomedical disciplines, there is little evidence of formal training programs or methodologies to improve data visualization skills.[2] Clinical use of data visualization does already exist. Pediatric growth charts and labor curves are two examples of current use. At a glance, these tools convey important information with a clinical impact.

Data visualization literacy is defined as "the ability to confidently use a given data visualization to translate questions specified in the data domain into visual queries in the visual domain, as well as interpreting visual patterns in the visual domain as properties in the data domain, the ability and skill to read and interpret visually represented data in and to extract information from data visualizations, and the ability to make meaning from and interpret patterns, trends, and correlations in visual representations of data."[3] Systematic evaluation of visualization literacy enables the development of competencies and curricula designed to improve data communication.

As a next step, Cabanski, Gilbert, and Mosesova describe five principles in constructing visualizations.[4]

  • Focus on the message
  • Displays must be fit-for-purpose
  • Simplify
  • Every graph should stand on its own
  • Avoid deception

Though not clinical, some university libraries do hold "Data Visualization Clinics," which are opportunities to develop and critique visualizations while learning the principles of construction.[5]

Historical Context

Data visualization as we know it was pioneered by the English engineer and economist William Playfair, who in 1786 published a treatise arguing that information is more easily understood when represented graphically. He was the first to present data in line graphs, bar graphs, and pie charts.[6]

Graphical representation of medical data specifically was pioneered by Florence Nightingale, the English nurse and statistician whose 1858 accounts of abysmal conditions in an army hospital during the Crimean War led to policy changes concerning hygiene and sanitation that led to a dramatic reduction in wartime mortality. Her accounts were crucially supported by polar area (or “coxcomb”) diagrams that represented monthly soldier mortality with shaded areas of representative sizes corresponding to frequency of cause of death. From her diagrams, it was clear that preventable infections killed more soldiers than battlefield wounds.[7]

Data visualization improves interpretability. As data become more complex, visualization becomes indispensable to extract meaning. Computerization in American healthcare has accelerated since the 2009 Health and Information Technology for Economic and Clinical Health (HITECH) Act promoting meaningful use of electronic health records (EHR).[8] This increased capacity to collect and store data has led to an increase in volume and complexity that has outstripped our natural capacity to analyze it, thus the need for effective data visualization.[9]

Philosophical Impact

Increasing computerization of all facets of life brings increased data. Smart watches provide data on location, steps, heart rate and rhythm, sleep patterns. Digital imagery provides petabytes of detail of our planet from subterranean to satellite mapping. Beyond the mechanisms of creating and interpreting data visualizations, understanding the broad implications carried by big data is an important discussion point. Rodenbeck discusses his experiences through three communications principles which help define the larger theoretical framework around data usage.[10]

  • Principle 1: Public conversations about science are never just about the truth. It’s wise to plan for this, and not shrink from it
  • Principle 2: Data visualization can invite more questions than it answers
  • Principle 3: Data visualization communication is never context-free. There’s no neutral or correct way to do this work

As we develop more methods of measuring, recording, and storing data, contextualizing efforts to communicate will continue to develop.

Bridging Text into Graphics

Most journal articles present information solely through text, due in part to the difficulty in creating accurate visual representations.[11] In several randomized controlled trials, Buljan, et. al. demonstrated no difference in knowledge acquisition between text and image modalities, noting the increased readability and user-preference of graphics.[12] Creating visualizations which update clinical knowledge is a subset of data visualization with clear clinical relevance. Given the large volume of published studies, identifying relevant articles can be challenging. The American Academy of Family Physicians (AAFP) created a framework to identify and present studies impacting clinical care. Patient Oriented Evidence that Matters (POEMs) meet this requirement. The AAFP defines POEMs as "summaries of research [...] relevant to physicians and their patients and meeting three criteria:

  • Address a question that primary care physicians face in day-to-day practice;
  • Measure outcomes important to physicians and patients, including symptoms, morbidity, quality of life, and mortality;
  • Have the potential to change the way physicians practice.[13]

Recognizing the limited time for reading, evaluating, and applying even this summarized data, visualization of the information can help clinicians maintain evidence-based practice.

Integration into Workflow

Target audiences for data visualization are wide-ranging and include individuals in private institutions, academicians, and the general public. Applying the colloquialism, "a picture is worth a thousand words," visual representations can communicate complex ideas quickly. This is particularly the case when data are presented in harmony with human factors such as existing associations, visual perception, and the limits of memory. For example, the color red is already associated with “stop” or “emergency” and should be used only to represent these concepts instead of, for example, alerts of user errors. Related information should be grouped together on the basis of relevance, (e.g. all data relevant to heart failure), as opposed to topic (e.g. labs, history, specialty tests). To take into account limits of memory, reference ranges can be displayed graphically, with the patient’s value overlaid. Icons should be meaningful (e.g. mortar and pestle for “pharmacy”) and be accompanied by redundant text labels.[14] Clinically engaged professionals usually have too many requirements for the time allotted. The addition of another source to access and evaluate is only worthwhile when it decreases the amount of total work. Simply adding another system without allowing for the change management process stymies successful implementation.[15]

Use cases include:

  • Improving workflow

Node-link analysis aids in defining process relationships. The linked image displays a node-link diagram describing the process of developing the plan of care for admission of patients into home care.[16]

  • Patient communication

Disease specific information can be impactfully communicated. AIDSVu is an interactive map providing visualizations at the national, state, and local levels, such as this one. The potential impact of data visualization on public health is tremendous.[17] At an individual level, graphical depiction of weight trends is useful in monitoring fluid balance in heart failure. Growth charts are a visualization commonly used in providing well-child care and evaluating childhood failure-to-thrive.

  • Practice management

Referral patterns, Patient demographic distribution, and HEDIS trends can aid in understanding care requirements and aid in resource allocation.

Visual Analytics

Visual analytics involves automated analyses leading to data visualization to support analytical reasoning. It goes beyond simple data visualization to include interactive interfaces that lend themselves better to exploration and hypothesis generation. In an era of big data, the goal of visual analytics is to “turn...information overload into an opportunity by enabling decision-makers to examine this massive information stream to take effective actions in real-time situations”.[18] More information per se is not helpful for decision-making but rather adds to cognitive load. Visual analytics offers a solution for this, particularly when it comes to applying analytical reasoning to problems that involve complicated timeline variables and aggregated data from large samples.

For clinical decision support

Clinical reasoning involves the type of analytical thinking that is well-supported by visual analytics. It requires analysis of complex timeline data (for example, timing of symptoms, overlapping diagnoses, correlations between medications and clinical signs) and group trends (for example, common presentations among a specific demographic, endemic illnesses, similar patients a clinician has encountered prior). Therefore, interactive data visualization may allow patterns to emerge from data that are otherwise too complex to analyze unsupported.

Rostamzadeh and colleagues[19] propose four primary dimensions to characterize and evaluate EHR visual analytics that are helpful to understand their use:

  • Tasks supported by the application: understand progression of disease, identify and explore cohorts of interest, create prediction models, detect adverse events
  • Analytics approaches: classification, clustering, pattern discovery, dimensionality reduction, regression, inference
  • Visualization approaches: relationship-based, hierarchy-based, flow-based, time-based
  • Interactions between user and data: arrange, compare, drill (down/up), filter, search, select, transform/translate, animate/freeze, collapse/expand, insert/remove, link/unlink

An early example of software applying visual analytics to clinical decision support is LifeLines, developed by Plaisant and colleagues in the early 1990s.[20] It was designed for longitudinal representation of patient histories, with time on the horizontal access and events listed on the vertical axis, represented by colored lines extending over time. LifeLines was initially built to visualize single patient histories but was further developed as a visual analytics tool with aggregation of thousands of patient histories to explore population trends.[21]

Other visual analytics tools for clinical decision support accomplish such tasks as:

  • Risk prediction based on medical code or patient visit patterns (sequence and timing) in the patient’s history (Retain VIS)[22]
  • Exploration of the interaction between disease progression and patient characteristics, considering disease states co-occur with symptoms and other variables (DPVis)[23]
  • Pharmacovigilance involving comparison between a reference sequence (e.g. laboratory tests, drug administration) and the patient’s sequence[24]
  • Identification and description of subpopulations of patients with similar characteristics to minimize number of viable treatment options (VisualDecisionLinc)[25]
  • Extraction of patterns from numerous records to determine how patients with chronic disease might develop co-morbidities later in life[26]

For public health

Florence Nightingale’s use of statistics and data visualization to improve sanitary conditions in hospitals is an illustrative example of how well-presented data informs policy and the public. Visual analytics, i.e. automated analysis of big datasets underlying interactive visualization interfaces, has been used to the same end, largely in the form of public health dashboards. Dashboards interactively represent multi-dimensional data from sources such as governmental institutions and health organizations, and cover topics including infectious disease (e.g. Ebola), non-infectious disease (e.g. diabetes), and natural disasters or hazards that can affect human health (e.g. floods, air pollution).[27]

The goals of such data visualization are to improve surveillance of public health risks, improve management of these risks, provide public access to information, and, in some cases, enable public participation in surveillance.[28][29] COVID-19 dashboards were plentiful during the global Sars-CoV-2 pandemic, during which challenges that face all visual analytics implementations, for example quality of data and its sources, presenting data in a manner that is meaningful for the lay person, and updating the interface as data becomes obsolete, were particularly salient.[30]

Some examples of dashboards with visual analytics relevant to public health include:

  • America’s HIV Epidemic Analysis Dashboard (AHEAD) (
  • Tuberculosis Data, Impact Assessment, and Communication Hub dashboard (
  • The Health of People in Canada dashboard (
  • The National Institutes of Health’s the All of Us data browser (


Computerization of healthcare continues to grow, and the quantity and dimensionality of healthcare data grows along with it. Visualization is a critical tool for deriving meaning from such data. In the face of big (and complex) data, challenges associated with visualization include data (quality, size, diversity), users (needs, skills), design (taking human factors into account when considering how data and users interact), and technology (tools, infrastructure).[31] As with all change processes, forward movement requires a significant amount of work to demonstrate realized potential and build a growing coalition of users and consumers.


The Data Visualization Catalogue

Data visualization:Wikipedia


  1. Frické M. The knowledge pyramid: a critique of the DIKW hierarchy. Journal of Information Science. 2008;35(2):131-42. doi:10.1177/0165551508094050.
  2. Elgendi M. Scientists need data visualization training. Nat Biotechnol. 2017;35(10):990-1. doi:10.1038/nbt.3986.
  3. Borner K, Bueckle A, Ginda M. Data visualization literacy: Definitions, conceptual frameworks, exercises, and assessments. Proc Natl Acad Sci U S A. 2019;116(6):1857-64. doi:10.1073/pnas.1807180116.
  4. Cabanski C, Gilbert H, Mosesova S. Can Graphics Tell Lies? A Tutorial on How To Visualize Your Data. Clin Transl Sci. 2018;11(4):371-7. doi:10.1111/cts.12554.
  5. LaPolla FWZ, Rubin D. The "Data Visualization Clinic": a library-led critique workshop for data visualization. J Med Libr Assoc. 2018;106(4):477-82. doi:10.5195/jmla.2018.333.
  6. Spence I. William Playfair and the psychology of graphs. American Statistical Association JSM Proceedings; 2006: 2426–2436.
  7. O’Connor S, Waite M, Duce D, O’Donnell A, Ronquillo C. Data visualization in health care: The Florence effect. J Advanced Nursing, 2020: 76(7): 1488-1490. doi: 10.1111/jan.14334
  8. Burde H. The HITECH Act: An overview. AMA J Ethics Virtual Mentor 2011:13(3):172-175. doi: 10.1001/virtualmentor.2011.13.3.hlaw1-1103
  9. West VL, Borland D, & Hammond WE. Innovative information visualization of electronic health record data: A systematic review. J Am Med Inform Assoc 2015;22:330-339. doi: 10.1136/amiajnl-2014-002955
  10. Rodenbeck E. Communications Principles for Inviting Inquiry and Exploration Through Science and Data Visualization. Integr Comp Biol. 2018;58(6):1247-54. doi:10.1093/icb/icy105.
  11. Buljan I, Malicki M, Wager E, Puljak L, Hren D, Kellie F, et al. Response to letter to the editor by Mc Sween-Cadieux et al. (2017). J Clin Epidemiol. 2018;100:133-4. doi:10.1016/j.jclinepi.2018.02.021.
  12. Buljan I, Malicki M, Wager E, Puljak L, Hren D, Kellie F, et al. No difference in knowledge obtained from infographic or plain language summary of a Cochrane systematic review: three randomized controlled trials. J Clin Epidemiol. 2018;97:86-94. doi:10.1016/j.jclinepi.2017.12.003.>
  13. POEMs Department Collection. American Family Physician; 2019 Web page. Available from: Accessed 18 Oct 2019.
  14. Pruitt ZM, Howe JL, Bocknek LS et al. Informing visual display design of electronic health records: A human factors cross-industry perspective. Patient Safety, 2023;5(2):32-39. doi: 10.33940/001c.77769
  15. Ulaganathan S, Fatah T, Schmidt T, Nohr C. Utilization of a Novel Patient Monitoring Dashboard in Emergency Departments. Stud Health Technol Inform. 2019;262:260-3. doi:10.3233/SHTI190068.
  16. Sockolow PS, Yang Y, Bass EJ, Bowles KH, Holmberg A, Sheryl P. Data Visualization of Home Care Admission Nurses' Decision-Making. AMIA Annu Symp Proc. 2017;2017:1597-606.
  17. Valdiserri RO, Sullivan PS. Data Visualization Promotes Sound Public Health Practice: The AIDSvu Example. AIDS Educ Prev. 2018;30(1):26-34. doi:10.1521/aeap.2018.30.1.26.
  18. Keim DA, Mansmann F, & Thomas J. Visual analytics: How much visualization and how much analytics? ACM SIGKDD Explorations Newsletter 2010:11(2): 5-8. doi: 10.1145/1809400.1809403
  19. Rostamzadeh N, Abdullah SS, Sedig K. Visual analytics for electronic health records: A review. Informatics 2021: 8(12). doi: 10.3390/informatics8010012
  20. Plaisant C, Milash B, Rose A, et al. LifeLines: visualizing personal histories. SIGCHI Conference on Human Factors in Computing Systems Proceedings; 1996:221–227.
  21. Wongsuphasawat K, Guerra Go´mez JA, Plaisant C, et al. LifeFlow: visualizing an overview of event sequences. SIGCHI Conference on Human Factors in Computing Systems Proceedings 2011:1747–1756.
  22. Kwon, B.C.; Choi, M.-J.; Kim, J.T.; Choi, E.; Kim, Y.B.; Kwon, S.; Sun, J.; Choo, J. Retainvis: Visual Analytics with Interpretable and Interactive Recurrent Neural Networks on Electronic Medical Records. IEEE Trans. Vis. Comput. Graph. 2018, 25, 299–309.
  23. Kwon, B.C.; Anand, V.; Severson, K.A.; Ghosh, S.; Sun, Z.; Frohnert, B.I.; Lundgren, M.; Ng, K. DPVis: Visual Analytics with Hidden Markov Models for Disease Progression Pathways. IEEE Trans. Vis. Comput. Graph. 2020.
  24. Ledieu, T.; Bouzille, G.; Plaisant, C.; Thiessard, F.; Polard, E.; Cuggia, M. Mining Clinical Big Data for Drug Safety: Detecting Inadequate Treatment with a DNA Sequence Alignment Algorithm. AMIA Annu. Symp. Proc. 2018, 2018, 1368–1376.
  25. Mane, K.K.; Bizon, C.; Schmitt, C.; Owen, P.; Burchett, B.; Pietrobon, R.; Gersing, K. VisualDecisionLinc: A Visual Analytics Approach for Comparative Effectiveness-Based Clinical Decision Support in Psychiatry. J. Biomed. Inform. 2012, 45, 101–106.
  26. Huang, C.-W.; Lu, R.; Iqbal, U.; Lin, S.-H.; Nguyen, P.A.A.; Yang, H.-C.;Wang, C.-F.; Li, J.; Ma, K.-L.; Li, Y.-C.J.; et al. A Richly Interactive Exploratory Data Analysis and Visualization Tool Using Electronic Medical Records. BMC Med. Inform. Decis. Mak. 2015, 15, 92.
  27. Schulze A, Brand F, Geppert J, & Bol GF. Digital dashboards visualizing public health data: A systematic review. Front Public Health 2023; 11: 999958. doi: 10.3389/fpubh.2023.999958
  28. Rees EE, Gendron B, Lelièvre F, Coté N, Bélanger D. Advancements in web-database applications for rabies surveillance. Int J Health Geogr. (2011) 10:48. 10.1186/1476-072X-10-48
  29. Tegtmeyer R, Potts L, Hart-Davidson W. Tracing and Responding to Foodborne Illness. In: Proceedings of the 30th ACM International Conference on Design of Communication. Seattle, Washington, USA: Association for Computing Machinery; (2012). p. 369–70. 10.1145/2379057.2379131
  30. Clarkson MD. Web-based COVID-19 dashboards and trackers in the United States: Survey study. JMIR Hum Factors 2023; 10: e43819. doi: 10.2196/43819
  31. Keim DA, Kohlhammer J, Ellis G, et al., eds. Mastering the information age-solving problems with visual analytics. Florian Mansmann, 2010.

Submitted by Barrett Campbell

Summary and Conclusion edits, paragraph on human factors in Integration into Workflow section, Historical Context section, and Visual Analytics section submitted by Jessica Keating