Clinicians must analyze large amounts of patient data from various sources to assess health and develop treatment plans. This process is becoming more complex with new sources of data. The study explored creating data visualization tools to help clinicians make sense of growing data volumes and varieties. Participants provided feedback on prototype tools, identifying challenges and beneficial design approaches. The tools showed potential to improve the speed and accuracy of healthcare decisions by making implicit thought processes more explicit.

1. Each day, clinical healthcare providers must review and analyze a large
volume and variety of patient information to assess health, anticipate
potential issues, diagnose illnesses, and develop the most effective treatment
plans. Making sense of patient information and acting on it are complex tasks.
Clinicians must integrate data from multiple sources, judge the integrity of
data, take into account a variety of circumstances that might not be reflected
in the medical record, adhere to patient wishes, and more.
Clinicians use a wide array of processes and tools to help collect potentially
useful patient information—from taking a patient’s medical history to
employing sophisticated imaging systems. Yet there are relatively few tools
to help sort through the accumulation of patient data. Empirical population
studies can provide guidance for identifying the risk of illnesses, for example,
but clinicians must determine the relevance of that information for individuals.
The introduction of new types of healthcare data promises to help clinicians—
and patients—better monitor health, better identify future health problems, and
better evaluate the effectiveness of treatments. For example, the data created
by wearable devices can help individuals track their progress toward health
goals and gauge responses to medications. The data produced from genomic
testing can help clinicians spot predispositions to particular illnesses.
However, the addition of this new data to traditional health data can also
increase the complexity of clinical decision-making. Doctors and nurses will
have to review and analyze more information from a wider variety of sources.
Numerical data from new sources might be difficult to decipher and integrate
with existing information.
Data visualization tools can help address the challenges clinicians face as
they try to make sense of a rapidly growing volume and variety of data.
Researchers at University College London, in collaboration with advisors
at Intel, conducted a study in which they created concepts for visualization
tools and evaluated the effectiveness of those tools for determining the risk
of illnesses. The study suggests that by making implicit thought processes
explicit, visualization tools can ultimately help improve the speed and
accuracy of healthcare decision-making.
Nicholas Tenhue,
UX Architect,
Genospace
Chiara Garattini
(correspondence to chiara.garattini@intel.com),
Anthropology & UX Research,
Health and Life Sciences,
Intel Corporation
Ann Blandford,
Professor of Human–Computer Interaction,
Director of UCL Institute of Digital Health,
University College London (UCL)
white paper
Enhancing Medical Decision-
Making Through Visualization
Visualization tools could help healthcare providers
make sense of large volumes of complex health data
and improve the speed and accuracy of decisions

2. How Do Clinicians Typically
Make Sense of Information?
To understand information—in healthcare
or any other field—an individual must
seek out appropriate information and then
interpret it. In many cases, the person
trying to make sense of information
places it into pre-existing frames. As
new information is collected, the person
continues to assimilate it into pre-existing
frames in a cyclical, iterative process.
Over time, the frames are modified to
reflect all the information collected.
For the most part, making sense of
information is an internal process
for clinicians. They review patients’
medical histories, evaluate test results,
note patients’ complaints, consider
physical signs that they observe, and
attempt to identify other contextual
information—from social factors to
employment status—that might affect
health. They apply this information to
their pre-existing frames for health,
disease, diagnostic categories, treatment
plans, and more—frames that have been
created and developed through extensive
education and clinical experience.
This complex internal process is
becoming more difficult as new types
of data are added to the mix. Data
collected from wearable computing
devices and data produced through
genomic testing, for example, can
substantially increase the complexity
of clinical decision-making.
The format of that new data adds yet
another level of difficulty. Studies have
shown that statistical or numerical
information can be particularly difficult
for people to understand. And yet a
large portion of risk-related evidence is
based on statistical data.
Intel researchers wanted to know whether
presenting healthcare data in a visual
manner, through a data visualization
tool, could improve clinicians’ ability
to understand information. Could a
visualization tool make it faster or
easier for clinicians to interpret a larger
volume and wider variety of data?
Could data visualization lead to better
risk assessments, allowing clinicians
to more easily identify future risks or
even to discount some indicators of risk
when more genetic, environmental, and
behavioral factors are taken into account?
Designing the Study
Intel researchers constructed a
multiphase study in which they
designed concepts for visualization
tools and worked with clinicians to
evaluate the potential benefits of using
data visualization to evaluate health
risks in a preventive care setting.
Study Participants
The study involved 10 participants.
All were primary-care clinicians or
specialists who see patients and
evaluate disease risk as part of their
daily practice. Six provided feedback on
initial tool designs, and all 10 provided
feedback on an optimized tool.
Data-Gathering Methods
Researchers collected information
from the participants through semi-
structured interviews. Participants also
contributed to “think-aloud” sessions in
which they attempted to verbalize their
thoughts as they completed tasks.
Patient Personas
Rather than using real patient
information, researchers created three
patient “personas”—fictional patients
whose profiles included information
synthesized from medical research.
The personas were used to populate the
visualizations with data and to facilitate
the think-aloud sessions with study
participants, as they evaluated each
patient persona for risk of developing a
specific condition.
Researchers selected three conditions
for which the prototype visualization tool
would help evaluate risk: lung cancer,
Table of Contents
How Do Clinicians Typically Make
Sense of Information? . . . . . . . . . . . . 2
Designing the Study. . . . . . . . . . . . . . 2
Study Participants. . . . . . . . . . . . . . 2
Data-Gathering Methods. . . . . . . . 2
Patient Personas. . . . . . . . . . . . . . . 2
Developing Concepts for Data
Visualization Tools. . . . . . . . . . . . . . . 3
Exploring the Visualization Tool. . . 4
Identifying Visualization
Challenges and Valuable
Design Approaches. . . . . . . . . . . . . . . 6
Underscoring the Value
of Visualization . . . . . . . . . . . . . . . . . . 7
Transforming Healthcare
and Improving Outcomes. . . . . . . . . 7
2Enhancing Medical Decision-Making Through Visualization

3. skin cancer, and type 2 diabetes. These
conditions were selected in part because
they have strong biomarkers—biological
signs that might indicate the presence of
a disease or a predisposition to a disease.
At the same time, these conditions
generally develop with the contribution
of environmental and behavioral factors,
such as smoking, excessive exposure to
sunlight, or obesity.
Selecting these conditions, which might
have a variety of contributing factors,
could help highlight the value of using
data visualization. Accurately identifying
the risk of developing disease depends
on interpreting information from
multiple sources.
Researchers also constructed the
personas so that the risks of developing
disease were not clear-cut. For example,
the patient at risk of developing lung
cancer was not a smoker. Researchers
hoped to present complex scenarios
that mirror real-life cases. Only by
taking into account a large amount
and wide variety of data can clinicians
accurately measure risk for these cases.
Developing Concepts for
Data Visualization Tools
In a three-stage process, Intel
researchers worked with study
participants to design and fine-tune a
concept prototype data visualization
tool. The goal was to create a concept
that could help researchers explore the
value of data visualization rather than
to produce a fully functioning tool.
Stage 1: In the first stage of the
study, researchers worked with a data
visualization expert to produce two
design concepts for a visualization tool
(Figure 1). Once the first version of the
tool was produced, the expert then
provided feedback for elements that
should be redesigned.
Researchers then worked in parallel
to create two independent designs
based on the initial expert feedback.
This parallel approach helped test
and compare visualization types,
presenting the same data set in
different visual structures. Each
design used the data from the three
patient personas.
Stage 2: In the second stage of the
study, six of the study participants—
practicing clinicians—evaluated
the designs. Researchers first
interviewed the participants, asking
Figure 1. In one of the initial data visualization concepts, bubbles were used to represent
patient data. The color of each bubble indicates whether it is a changeable or non-
changeable factor and signifies the type of information the bubble represents.
about their current workflows and
decision-making processes so
they could better understand how
clinicians make sense of data. Then
researchers led participants through
a “think-aloud” session with the
two different visualizations. The
participants assessed a patient at risk
of developing a condition using the
visualization designs (Figures 2 and 3)
and verbalized their thinking.
Figure 2. For this data visualization concept, colors were assigned to different categories
of information to help reviewers quickly find what they needed. This concept provides the
name of the persona and the condition he is at risk of developing, and enables access to
radiographic scans.
3Enhancing Medical Decision-Making Through Visualization

4. Stage 3: Researchers used the input
and feedback from participants to
merge the best features of the two
visualization designs, creating a single
tool (described more fully in the next
section). The goal was to apply findings
about how clinicians think about and
evaluate risk, producing a tool that
reduced the gap between the clinicians’
mental model and the external
representations of data.
Researchers discarded one of the three
personas in this stage to accelerate the
design process. The designers created
one visualization design using the two
remaining patient personas.
Exploring the Visualization Tool
The visualization tool created in the
third stage of the study incorporates
input from the data visualization
expert, information about workflows
and thought processes drawn from
study participants, and feedback from
participants about the initial designs.
Rather than a final, finished product, it
offers a strong starting point for future
visualization tools.
Drawing on previous research
about effective approaches to data
visualization, the visualization tool
created for this study offers an overview
screen with patient data presented in a
macro view. From an overview screen,
clinicians could quickly identify key
information and risk factors, and then
zoom in to find more detailed data.
Patient information and risk factors:
The top of the overview screen provided
the name, age, sex, and a picture of the
patient (Figure 4). On the left, in colored
boxes, the screen presented data for
numerous factors that could play roles
in the development of disease, including
age, gender, postal code, socioeconomic
status, environment, ethnicity, job,
marital status, education, family history,
and genetic factors. The right side
presented colored boxes with both
non-changeable and changeable risk
Figure 3. In one of the second-generation data visualizations, the severity of risk is represented
by the size of a block in the “Increase Risk” column. Lines are drawn from the central “Patient
Information” column to show relationships between factors that reduced, increased, or had no
direct correlation to risk. Design by Marisa Parker.
Figure 4. The overview screen shows patient information and risk factors. For this patient
persona, the tools suggest a combination of non-changeable and changeable factors put the
patient at high risk of type 2 diabetes.
4Enhancing Medical Decision-Making Through Visualization

5. factors for a specific disease relative
to empirical population data. Non-
changeable factors included sex and
genetic markers while changeable
factors included exercise, weight, and
job attributes. Boxes with deeper red
hues indicated increased risk.
Optimizing placement of the data
visualizations on the screen was
essential: Researchers wanted to
ensure the most important data was
displayed prominently so clinicians
would not have to expend mental
energy searching for information.
Risk calculator: From the overview
page, participants could select “Apply
Calculator” to access a risk calculator,
which would appear superimposed
over the page (Figure 5). Using a risk
calculator can help clinicians quickly
assess a patient’s risk of developing
diseases (Figure 6).
Detailed information: By clicking a
patient information box, participants
could zoom in on data, accessing more
detail. For example, clicking “Postcode”
would show a map of the area where
the patient lived. Information about the
patient’s neighborhood might provide
vital information not otherwise
contained in the medical record. For
example, a clinician might discover
that a nonsmoking patient lives in
Figure 5. Participants could access a risk calculator to provide a
general overview of risk while also helping them recognize potentially
important factors that were not included in the calculation.
Figure 6. The risk calculator presents the odds of developing a
specific disease.
5Enhancing Medical Decision-Making Through Visualization

6. an area where he might be exposed
to radon, which can contribute to
the development of lung cancer.
Participants could also review genetic
information and identify any factors
that might affect a patient’s risk for a
particular disease (Figure 7).
This detailed view also provided
additional buttons at the bottom of
the screen:
• Notes: Clinicians could view
and contribute to the patient’s
healthcare record.
• Guidelines: Clinicians could access
healthcare guidelines available from
the National Institute for Health and
Care Excellence.
• Evidence: Clinicians could search
for available evidence and the latest
literature on a risk factor.
• Change risk classification:
Clinicians could flag a piece of
information with a risk factor.
• More: A button for more
information enabled participants
to quickly jump to other parts
of the clinical workflow, such as
investigations or diagnosis.
Identifying Visualization
Challenges and Valuable
Design Approaches
Interviewing and soliciting feedback
from the study participants was helpful
in understanding how clinicians think
and what workflows they use to assess
risks and make decisions. For example,
researchers confirmed that to assess
risks, clinicians routinely use data from
a number of different sources, including
case studies, risk calculators, and
guidelines, in conjunction with patient
data related to a particular condition.
This diverse data is usually presented in
a way that makes it difficult to discover
relationships among potentially
significant pieces of information.
In working with participants on the study,
researchers uncovered some of the
challenges for developing visualization
tools. For example, researchers found
discrepancies between how clinicians
describe their thought processes (how
they think they work) and what they
actually do when evaluating risk. Those
discrepancies came to light when
researchers compared the information
drawn from interviews with the
information derived from think-aloud
sessions. Participants also showed
apprehension in mixing evidence-
based calculators and contextual
information, since doing so highlighted
the discrepancies. One of the goals of
designing a visualization tool is to reflect
the internal thought process, making the
implicit more explicit in an effort to
enhance the efficiency and effectiveness
of decision-making. But the tool design
process will be challenging if the implicit
decision-making process is complex,
highly internalized, and difficult for
clinicians to fully verbalize.
If researchers discovered potential
obstacles to effective tool design, they
also identified some useful design
Figure 7. While zooming in, participants could also view genomic information and identify
potential genetic risk factors.
6Enhancing Medical Decision-Making Through Visualization

7. approaches. For example, presenting
all of the patient information in a single
interface helps highlight potentially
significant contextual factors that
might otherwise be overlooked.
Providing information about a patient’s
socioeconomic status and home
environment close to blood test
results and other clinical findings can
help ensure all relevant information
is considered as part of the decision-
making process.
Some study participants noted that
superimposing the risk calculator on
the overview page could help them
determine whether key factors were
absent from that risk calculation. For
example, clinicians might see that the
risk calculator used factors such as age,
ethnicity, and family history to determine
the risk of type 2 diabetes. But with a
quick glance back at the overview page,
clinicians could quickly recognize that the
patient’s high fitness level and relatively
low weight lowered the risk.
Enabling clinicians to zoom in for
detailed data was also found to be
useful. By zooming in, clinicians can
better understand why certain factors
might contribute to or reduce risk.
Underscoring the Value
of Visualization
The study suggests a number of ways
that visualization tools help clinicians.
For example, visualization tools enable
clinicians to accommodate a greater
volume of data in their decision-making
processes. As long as data is presented
in a way that is easily consumed (as it
was in the visualization tool created
for this study), clinicians will not
be overwhelmed by the volume of
information. The visualization tool used
in this study enabled clinicians to have
a quick, high-level view of data and risk
based on a broad range of factors. They
could then dig deeper into data that
might be relevant.
Visualization tools can also support
a greater variety of information,
ranging from quantifiable, “hard”
data that typically drives evidence-
based medicine to “soft,” qualitative,
contextual data. From interviewing
participants, researchers found that
clinicians generally claim to rely on
“validated clinical data,” such as data
drawn from use studies based on
specific populations and tools such
as risk calculators. However, clinicians
also accept that research evidence is
an abstracted generalization that does
not always represent the actual risk of
individual patients. Consequently, they
often use soft data in their decision-
making process and in their treatment of
patients. For example, as one participant
remarked, “Knowing that [a patient] is
widowed is fairly important. Her husband
might have just died last week, and [if so,]
you don’t really want to necessarily be
talking to her about her risk of melanoma
[now]. She won’t care.” Including soft
data as part of a visualization tool makes
it easier for clinicians to access that soft
data when evaluating risk and interacting
with patients.
The feedback received from
participants suggests that visualization
tools could help clinicians improve
the speed of decision-making. Having
all relevant information in one place,
available at a quick glance, could enable
clinicians to more quickly move through
their decision-making processes. One
participant noted: “[E]xperienced
clinicians will have these gestalts….
You see someone and you recognize a
certain sort of configuration of features.
And if you then focus on bringing out
important risk factors in this patient…
and here’s the evidence base behind it...
you are facilitating that process.”
Visualization tools could also help
accelerate processes by making
implicit, internal thought processes
more explicit. Clinicians would not need
to expend as much time and mental
effort assembling and assessing data.
The study also suggests that
visualization tools could enhance the
accuracy of decisions. Clinicians could
take more information into account and
uncover relationships among factors
that might ordinarily be hidden. Having
an efficient way to integrate a large
volume and wide variety of information
would be especially useful as new types
of data—such as data from wearables
or genomics—become more readily
available to clinicians. With tools to help
them understand and integrate this data
into their risk assessment processes,
clinicians could produce more accurate
assessments of risk and subsequently
enhance treatment decisions.
Transforming Healthcare and
Improving Outcomes
The rapidly increasing volume
and variety of data available today
presents tremendous opportunities for
transforming healthcare. If healthcare
providers can successfully integrate
this data into their practice, they can
generate new insights that help better
identify health risks for individual
patients, optimize treatment plans,
enhance the efficiency of healthcare,
and more.
Intel and other companies are producing
leading-edge hardware technologies
to give healthcare providers the
horsepower to rapidly analyze and
efficiently manage this data, whether
those providers are using large-scale
data centers or office workstations.
New visualization tools can provide
important complements to those
technologies. As this study suggests,
visualization tools can help clinicians
more fully capitalize on available data
to make smarter, faster decisions that
ultimately improve patient outcomes.
7Enhancing Medical Decision-Making Through Visualization

8. For More Information
To learn more about the ways in which
Intel is contributing to new breakthroughs
in health and life sciences, visit:
www.intel.com/content/www/us/en/
healthcare-it/healthcare-overview.html
Acknowledgments: This paper is a
condensed version of the Master of
Science degree thesis in ICT innovation
at University College London (UCL)
presented by Nicholas Tenhue. The
work was conducted while he was an
intern at Intel Health and Life Sciences.
Chiara Garattini, Anthropology UX
Research, Health and Life Sciences, Intel
Corporation, served as the industry
supervisor; Ann Blandford, Professor
of Human–Computer Interaction and
Director of UCL Institute of Digital
Health, was the faculty supervisor for
the thesis. Intel thanks all contributors
and participants who gave their time
to help with this project, including
Marisa Parker, a senior user experience
designer formerly at Intel Corporation
and currently at Microsoft, who was a
key designer for the project.
Enhancing Medical Decision-Making Through Visualization
INFORMATION IN THIS DOCUMENT IS PROVIDED IN CONNECTION WITH INTEL® PRODUCTS. NO LICENSE, EXPRESS OR IMPLIED, BY ESTOPPEL OR OTHERWISE, TO ANY INTELLECTUAL PROPERTY RIGHTS
IS GRANTED BY THIS DOCUMENT. EXCEPT AS PROVIDED IN INTEL’S TERMS AND CONDITIONS OF SALE FOR SUCH PRODUCTS, INTEL ASSUMES NO LIABILITY WHATSOEVER, AND INTEL DISCLAIMS ANY EX-
PRESS OR IMPLIED WARRANTY, RELATING TO SALE AND/OR USE OF INTEL PRODUCTS INCLUDING LIABILITY OR WARRANTIES RELATING TO FITNESS FOR A PARTICULAR PURPOSE, MERCHANTABILITY, OR
INFRINGEMENT OF ANY PATENT, COPYRIGHT OR OTHER INTELLECTUAL PROPERTY RIGHT. UNLESS OTHERWISE AGREED IN WRITING BY INTEL, THE INTEL PRODUCTS ARE NOT DESIGNED NOR INTENDED
FOR ANY APPLICATION IN WHICH THE FAILURE OF THE INTEL PRODUCT COULD CREATE A SITUATION WHERE PERSONAL INJURY OR DEATH MAY OCCUR.
Intel may make changes to specifications and product descriptions at any time, without notice. Designers must not rely on the absence or characteristics of any features or instructions marked “reserved” or
“undefined.” Intel reserves these for future definition and shall have no responsibility whatsoever for conflicts or incompatibilities arising from future changes to them. The information here is subject to change
without notice. Do not finalize a design with this information.
The products described in this document may contain design defects or errors known as errata which may cause the product to deviate from published specifications. Current characterized errata are available
on request. Contact your local Intel sales office or your distributor to obtain the latest specifications and before placing your product order. Copies of documents which have an order number and are referenced
in this document, or other Intel literature, may be obtained by calling 1-800-548-4725, or by visiting Intel’s Web site at www.intel.com.
© 2015, Intel Corporation. All rights reserved. Intel and the Intel logo are trademarks of Intel Corporation in the U.S. and other countries.
* Other names and brands may be claimed as the property of others.
Printed in USA 1015/CG/TDA/XX/PDF Please Recycle 332722-001US