Conference PaperPDF Available

Patient Journey Visualizer: A Tool for Visualizing Patient Journeys

Abstract and Figures

Patient journeys indicate journeys of patients from sickness to recovery. Recent policy initiatives have provided opportunities to researchers to visualize and data- mine patient journeys. In this paper, we aim to understand patient journeys via certain powerful visualization charts, which help mine patterns in Big-Data concerning patients at the individual and aggregate levels. We developed a Patient Journey Visualizer (PJV) tool that helped visualize patient journeys via Parallel Coordinates, Sankey, and Sunburst charts. Parallel Coordinates assist in visualizing multivariate data concerning patient journeys in PJV at the individual level. Sankey charts help to visualize the flow of patients between various phases of patient journeys in PJV. Sunburst charts contribute to representing hierarchical relationships between diagnoses, procedures, and prescription medications in PJV. We performed a comparison of different visualization charts in PJV across increasing number of data points. Results reveal that Parallel Coordinates chart performs better than Sankey and Sunburst charts when data set size increased. We highlight the implication of our results for visualizing patient journeys in health care.
Content may be subject to copyright.
Patient Journey Visualizer: A Tool for Visualizing Patient Journeys
Kulendra Kumar Kaushal1, a, Shruti Kaushik1, b, Abhinav Choudhury1, c, Krish Viswanathan2, d, Balaji Chellappa2, e, Sayee
Natarajan2, f, Larry Pickett2, g, and Varun Dutt1, h
1Indian Institute of Technology Mandi, India 175005
2Rx Data Science, USA27709
akulendra_kumar_kaushal@students.iitmandi.ac.in, bshruti_kaushik@students.iitmandi.ac.in,
cabhinav_choudhury@students.iitmandi.ac.in, dkrish@rxdatascience.com,
ebalaji@rxdatascience.com, fsayee@rxdatascience.com, glarry@rxdatascience.com, hvarun@iitmandi.ac.in
AbstractPatient journeys indicate journeys of patients
from sickness to recovery. Recent policy initiatives have
provided opportunities to researchers to visualize and data-
mine patient journeys. In this paper, we aim to understand
patient journeys via certain powerful visualization charts,
which help mine patterns in Big-Data concerning patients
at the individual and aggregate levels. We developed a
Patient Journey Visualizer (PJV) tool that helped visualize
patient journeys via Parallel Coordinates, Sankey, and
Sunburst charts. Parallel Coordinates assist in visualizing
multivariate data concerning patient journeys in PJV at the
individual level. Sankey charts help to visualize the flow of
patients between various phases of patient journeys in PJV.
Sunburst charts contribute to representing hierarchical
relationships between diagnoses, procedures, and
prescription medications in PJV. We performed a
comparison of different visualization charts in PJV across
increasing number of data points. Results reveal that
Parallel Coordinates chart performs better than Sankey and
Sunburst charts when data set size increased. We highlight
the implication of our results for visualizing patient
journeys in health care.
Index TermsKDB+, Q, Parallel Coordinates, Patient journeys,
Sankey, Sunburst, visualizations.
I. INTRODUCTION
A healthy population is an asset to society, and providing
personalized and timely health care is an important
objective for policymakers in today’s world [1]. For
achieving these goals, modern healthcare has started
analyzing patient journeys, which document how patients
move in a health care system from their sickness to
recovery [2]. Patient journeys are likely to involve all the
sequential steps starting from visiting doctors, filling
prescriptions at pharmacies, and getting lab tests done, to
getting treatment and recovering from sicknesses.
Patient journeys cut across different medications,
procedures, and diagnoses with large volumes of data that
gets generated at great velocities. Given this big patient
journey data, we urgently need powerful visualization
techniques which can help us visualize multivariate data
and hierarchical relationships in data at both the individual
as well as the aggregate levels. Visualization techniques
like Parallel coordinates charts, Sankey charts, and
Sunburst charts could help in visualizing multivariate
patient journey data and the hierarchical relationships
present in this data. This paper aims at visualizing patient
journeys from the perspective of various stakeholders like
physicians, drug manufacturers, medical insurer through
Parallel coordinates, Sankey, and Sunburst charts [3].
Visualization proposed in this paper can significantly
contribute towards physicians understanding of a disease
and performance of various drugs and procedures in
treating diseases. For example, the visualization tool,
Patient Journey Visualizer (PJV), reported in this paper
helps in visualizing patients consuming drugs before and
after certain diagnoses or procedures. In summary,
visualization tools are likely to help drug manufacturers to
analyze the performance of their drugs against other
medicines in the market as well as the performance of
medications across certain procedures and diagnoses [4].
Beyond drug manufacturers, visualizations would also be
helpful to various other stakeholders like physicians and
health insurers.
In what follows, we first provide background literature
regarding the use of visualization techniques for patient
journeys. Next, we detail how newer visualizations like
Parallel Coordinates, Sankey, and Sunburst charts could be
utilized for display of quantitative patterns in patient
journeys in the PJV tool. Here, we divide our analysis into
three sections: The first section deals with patient journeys;
the second section deals with the procedure we used to
extract data from the database; and, finally the last section
shows the visualization of multivariate data sets. We close
this paper by discussing the significance of our results for
improving patient care.
II. BACKGROUND
There is an urgent need to visualize patient journeys from
different stakeholder perspectives because of the insights
they provide related to patient care [5]. Process mapping [6]
is a clinical audit which is used to build patient journeys
based on the patient’s feedback about various healthcare
services like x-ray diagnoses [7]. It helps in understanding
various issues related to the patient experience by listing out
all the subsequent events [8]. Though the process-mapping
technique maps the entire patient’s problem with different
sequential steps involved [6]; however, to visualize patient
journeys of various stakeholders’ perspectives one needs to
use innovative visualization tools [5]. These tools, when
embedded in software, help in the communication of ideas
between various stakeholders involved. Furthermore, these
tools are likely to explicitly help figure out the variables of
interest for different interested parties and increase the
involvement of each of them for making the patient
experiences smoother [5].
Furthermore, due to the continuous real-time and
multivariate nature of patient journey data, one-time 2-D X-
Y visualizations are not sufficient. Rather, we need to have
an automated software solution which can help us in
visualizing the multivariate patient journey data in real time.
Conventionally, researchers have used visualization
methods like 2-D X-Y plots, scatter-grams, histograms,
which are limited in the sense that they can be used to
visualize only two-dimensions at a time [8]. In contrast,
proposed visualizations like Parallel coordinates, Sankey,
and Sunburst charts can handle multivariate data and could
also show the hierarchical relationships present in the data.
For example, [9] proposed a visualization tool called
“CareFlow” using the Sankey charts where the system is
designed to help doctors and patients communicate about
possible care plans and their outcomes. CareFlow allows
physicians and patients to understand which treatments have
historically been most effective.
Another patient journey analyzer tool called, “Patient-
viz,” has been proposed [10]. In this tool, electronic medical
records (EMRs) and administrative data containing distinct
events, like diagnoses and laboratory tests, are used to tell
the story of a patient by using a visual representation of data
using charts over time.
Furthermore, [11] have introduced “EventFlow,” a
visualization tool that transforms an entire dataset of
temporal event medical records into an aggregated display,
allowing researchers to analyze population-level patterns
and trends. As per [11], when data becomes large, a series
of user-driven data simplifications allow researchers to pare
event records down to their core elements.
Inspired by the successes with Patient-viz, CareFlow,
and EventFlow visualization tools [09]-[11], [10], in this
paper, we present a software tool called, “Patient Journey
Visualizer (PJV).” The PJV tool simplifies the temporal
health data through different filtering mechanisms like
brushing and data selection. The tool has different
visualizations implemented that help graph multivariate data
and has many exciting features unique to each chart. In the
PJV tool (referred to as the “software tool” hereafter), users
can select a dimension for visualization and use various
demographic filters. Certain visualizations like Sunburst
charts in the tool can help change the perspective of the user
in real time. In the section that follows, we present
methodological details of building visualizations in our
software tool.
III. METHOD
In this paper, we visualize patient journey big-data for
different stakeholders like pharmaceutical firms and
physicians. The data contains attributes like Patient ID, age,
gender, race, International Classification of Diseases (ICD-
9) codes, Healthcare Common Procedure Codes (HCPC),
National Drug Code (NDC) codes of prescribed drugs, and
duration of the medication. In the software tool, a Big-Data
architecture based on Q-query language was used to query
an in-memory column-based KDB+ database (from Kx
Systems) to find various patterns in the underlying data.
Figure 1 shows the architecture sketch of the software tool
for visualization of patient data. We chose KDB+ database
package because, in contrast to other distributed systems, it
is an in-memory column-based database that is fast and
could process a billion records on a single machine with
large amounts of RAM [12]. After the database is queried,
the results of the query are given to JavaScript regarding
JSON object and JavaScript plots the visualizations (see
Figure 1). For visualizing patient journeys, Parallel
Coordinates, Sankey, and Sunburst charts have been used
on top of JavaScript. Different visualizations gives different
insights and patterns about patient journeys, which helps
different stakeholders in visualizing their variables of
interest.
A. Patient Journeys
For building patient journeys in the software tool, we
focused our analyses on patient diagnoses, procedures
performed on patients, and medications refilled by patients.
This data was obtained from a proprietary claims dataset
consisting of more than 100,000+ patients. This dataset
contained following fields: patient’s ID, International
Classification of Diseases (ICD-9) codes, date of diagnoses,
HCPC procedure codes, time of procedures, and
medications refilled by patients along with their NDC drug
codes. For Parallel-Coordinate charts, attributes like Patient
ID, ICD9 codes, HCPC codes, Date of Diagnosis, Date of
Procedure, NDC codes, age, and gender are considered as
separate vertical axes. For Sankey charts, ICD9 Diagnosis
codes, HCPC Procedural codes have been considered as
nodes and number of patients from an ICD9 code to an
HCPC code have been found as weight on the link between
nodes. For Sunburst charts, ICD9 codes, HCPC codes, NDC
codes and number of patients belonging to each of these
codes have been considered.
Figure: 1. Architecture Diagram of the PJV Tool
1) Patient journeys for Parallel Coordinates Charts:
Parallel coordinates charts could be used to visualize
multivariate data [13, 14]. This chart was first proposed by
Inselberg for analysis of hyper-dimensional geometry [15].
In the Parallel Coordinates Chart, to represent an N-
dimensional point, we have N-equally spaced vertical
parallel lines. The N-dimensional data point is represented
as a polyline with vertices on vertical parallel lines. The
vertex’s position on ith axis determines the ith coordinate
of that point. For the Parallel Coordinates chart, patient
journeys were summarized as the relationship between the
patient id, diagnoses codes, date of diagnoses, medications
refilled and procedures performed. Thus, the Parallel
Coordinates chart visualized patient journeys on an
individual patient level. Stakeholders like physicians could
visualize meaningful patterns like medicine performance
between the different set of patients diagnosed with the
same disease or medication performance across several
demographic variables belonging to patients.
2) Patient journeys for Sankey Charts: Sankey charts
[16] are a particular type of flow diagram where the
thickness of each link between connecting entities is
proportional to data being represented. Sankey diagram was
first used by Riall Sankey to study the flow of energy in
steam engines and thus was named after him [17]. In our
case, the link between two entities shows the flow of
patients between combinations of medications refilled,
diagnoses made, and procedures performed when these
three are taken two at a time. For the Sankey chart, patient
journeys were analyzed over a specified duration. The
duration was defined as the date of analysis ± an
observation period. Journeys were visualized as the flow of
some patients over the duration of diagnoses made to
medications refilled; procedures conducted to medications
refilled; and, from diagnoses made to procedures
conducted. The flows were an aggregated versions of
patient journeys, i.e., here we did not investigate individual
patients. Thus, the Sankey chart shows patterns which are
useful for stakeholders like pharmaceutical firms, which
may be interested in aggregate patterns over patient
journeys. Moreover, using the Sankey charts, one could
also compare medications’ performance across several
groups of patients with different demographic variables.
3) Patient journeys for Sunburst Charts: Sunburst charts
are used to visualize hierarchical data in the form of
multilevel pie-chart or concentric circles, where the
thickness of each segment represents the proportion of
different values found at each level of hierarchy [18]. These
charts have been used in cases involving hardware
components (e.g., for showcasing hard-disk contents) [19].
We have used these charts to summarize patient journeys
for a different group of patients involved in different events
like diagnoses made, procedures performed, or medications
refilled. The chart is useful to show the variation among
different groups for an event like medications prescribed. It
also helps in grouping various diseases that occur in a
sequence. Overall, this chart could help stakeholders like
insurers and physicians in predicting ailments that could
affect several patients.
B. Data format and queries to extract data from KDB+
Data format required for all three charts, Parallel
Coordinates, Sankey, and Sunbursts, are different. For
Parallel coordinates charts, we need a single-level JSON
structure containing all the relevant fields. For Sankey
charts, we need a two-level JSON structure. For Sunburst
charts, we need a multilevel JSON structure. We created
data for different charts in the required format by writing
chart-specific q-scripts in KDB+. In this section, we
describe the JSON format used for creating different charts.
1) Data Format for Parallel Coordinates Charts: Figure 2
shows the data format for the Parallel Coordinates chart.
The format contains fields that refer to a single patient.
Among the fields, Patient_ID relates to the id, which is
unique to each patient. Diagnosis_code relates to the
International Classification of Diseases (ICD-9) codes for
diagnoses made for a patient. Diagnosis_Date refers to the
date on which diagnosis was performed. Procedure_Code
refers to the Healthcare Common Procedure Codes (HCPC)
for procedures carried out on a patient, Procedure_date
refers to the date of procedure and Drug_NDC relates to the
National Drug Code (NDC) of the drug prescribed to the
patient. A plugin was written using JavaScript library d3js
[13, 20] which take the JSON object in Figure 2 format and
plots the Parallel Coordinates chart. The interface was
made more interactive by adding several features like
brushing experience in the d3js library.
2) Data Format for Sankey Charts: Figure 3 shows the
data format for Sankey charts. Here, nodes represent the
name of the entities between which Sankey chart is plotted,
and links show the relationship between them. In our case,
nodes can be diagnoses, procedures, or medications. Figure
4(A) shows the q-snippet used to extract drugs and
diagnosis from a table X and then combine them as node
objects. Whereas, Figure 4(B) shows the q-snippet used to
compute link objects concerning source nodes, target nodes
and the number of patients involved between origin and
destination nodes. X used in Figure 4 is the table which
contains both drug code represented by ndc_code and
International Classification of Diseases (ICD-9) codes
represented by diagnosis_code. Figure 4(C) shows the q-
snippet used to combine nodes and links object obtained
from Figure 4(A) and 4(B) into one object named graph.
Nodes1 and links1 used in Figure 4(C) are q-tables that
contain the names of nodes and the relationship between
them. We can also put filters on various parameters while
plotting Sankey charts. This feature can be implemented
{
"Patient_ID":"1",
"Diagnosis_Code":"V5812",
"Diagnosis_Date":"2014.08.01",
"Procedure_code":"J4568",
"Procedure_Date":"2014.08.22",
"Drug_NDC":"00054455025"
}
Figure: 2 Data format for parallel coordinate axes
using a q-query that takes the selected parameter say the
name of the node as an argument in the q-query, runs the
query and returns the results filtered on the parameter
passed to the query.
3) Data format for Sunburst Charts: Figure 5 shows the
JSON format for Sunburst charts regarding the events like
diagnoses made, procedures performed and medication
refilled in their hierarchical order of occurrence.
Diagnosis1 represents the diagnosis for a group of patients.
In Figure 5, Medication1 refers to the drug prescribed to a
sub-group of patients, who have been diagnosed with
diagnosis1. Furthermore, out of this sub-group, 3,938
patients were stored
in a variable named
“size” and were
operated under
procedure1. These
charts have zoom-in
and zoom-out
feature which makes
the charts more
interactive. By
selecting any
segment, the
sunburst chart is
redrawn, and it starts showing hierarchy from that selected
segment [21].
IV. RESULTS
C. Patient Journey with Parallel Coordinates Chart
Patient journeys visualized with the Parallel coordinates
charts highlight the medical history of an individual patient
by visualizing the relationships between diagnoses made
and procedures performed (see Figure 6). In Figure 6, hvid
represents individual patient id, diagnosis_code represents
the ICD9 code of the diagnosis made, short_description
refers to the procedure performed, strength_mg represents
the strength of medication prescribed, days_supply
represents the duration of the procedure, and
proprietary_name represents the name of the medication
prescribed. As shown in Figure 7, the highlighted polyline
shows an individual patient journey (patient id 54924452),
where the table below the chart shows the data row
corresponding to the polyline in the graph.
Furthermore, the tool also provides features like brushing
and correlations in the Parallel coordinates chart [13] (see
Figure 8 and Figure 9). The brushing helps to apply filters
on the data in real time within a certain range. For example,
the brush on data on the short_description axis restricts the
polylines (patients) to only those passing the brush (see
Figure 8). Brushing can be applied on as many axes as the
user wants and the user can visualize a reduced subset of
data points on the axis on which brushing is applied.
{
"nodes":[
{"name":"node0"},
{"name":"node1"}
],
"links":[
{"source":"node0", "target":"node1"
,
"value":"2"}
]
}
Figure: 3. Data Format for Sankey Diagram
l1: (select distinct name: ndc_code from x);
l2: (select distinct name: diagnosis_code from
x);
nodes1: l1,l2;!
Figure: 4(A). Q Snippet used to derive dr ugs and diagnosis as nodes object
"#$%&'(!select&&)*+,-(!$.,/,).-0!12+3-1(!
.#23$)&#&/,).-0!45#.!from&6!7!
"#$%&8(!9:!select&;52"*-(!count&45#.!by&&)*+,-012+3-1!
from&"#$%&'7!
Figure: 4(B). Q Snippet used to derive links object consisting of
drug as source node, diagnosis as target node and count of
patients between them as w eight.
!
`nodes`links!(x ; y)
}[nodes1;links2];
:graph
Figure: 4(C). Q Snippet us ed to derive graph object by combining
{
"name": "Patient Journey",
"children": [
{
"name": "diagonosis1",
"children": [
{
"name": "Medication1",
"children":
[
{"name": "Procedure1", "size": 3938}
]
}
]}]}
Figure: 5. Data Format for Sunbu rst Visualization
Moreover, the user can simply decrease the number of
dimensions in real time through un-marking the dimension
check box. The combination of reduced dimension and
brushing can allow the user to see the visualization for the
relationship it wants. The correlation features as shown in
Figure 9 enables users to see different correlations and their
p-values (in brackets) between the various axes in the chart.
Significant positive or negative correlations are marked in
green and red color respectively (significance is tested at
0.05). The tool also provides the ability to users aggregate
an axis in number and percentages (see Figure 10). For
example, for the diagnoses Rheumatoid Arthritis, the tool
allows the user to see the proportions of different
procedures performed.
D. Patient Journey with Sankey Chart
Sankey charts are a specific type of flow diagram, in which
the width of arrows is shown proportional to the flow
quantity. The chart is used for visualizing the movement of
the patients connecting procedures performed, diagnoses
made, and medications refilled. In the software tool, while
working with the Sankey diagram, the user first selects a
date of analysis say d1 and the duration of analyses (the
observation period). Next, the flow of patients from
diagnosis made to medications refilled within range (d1 -
user selected observation period, d1 + user selected
observation period) is visualized. For example, as shown in
Figure 11, the tool allows users to visualize the patient
flows within a specified observation period of an analysis
date. In this case, one could easily see a vast majority of
patients being mapped between the diagnosis Rheumatoid
Arthritis and the medication Methotrexate. Similarly, the
chart could be used for visualizing the flow of patients
between
procedures
performed
and
medications
refilled
around an
analysis date.
In summary,
Sankey charts enable users to observe dependent variables
aggregated over many patients and visualize large-scale
patterns in data.
E. Patient Journey with Sunburst Chart
Patient journeys in Sunburst Chart can be summarized
as a sequence of events like diagnosis made, procedures
performed and medications refilled depicted by concentric
rings with increasing diameter. For all the sequence of
events happening at one level in the hierarchy, the chart
creates one concentric ring. For each group of patients
belonging to that level of the hierarchy, there is a segment
in that concentric ring represented by different colors (see
Figure 12). Also, by selecting any segment on the sunburst
circles, the chart is redrawn, and it starts showing hierarchy
in concentric circles from the selected segment as its center.
Overall, like the Sankey chart, the Sunburst chart allows
one to visualize the aggregate patterns and sequences in
data. However, one could select different segments
(representing diagnoses, procedures, or medications) for
visualizing via this chart in real time.
F. Performance Analysis of Various Visualization Charts
We performed an analysis of the capacity of different web-
based applications to handle data from various sizes. For
this purpose, we took data sets of 200 points, 2,000 points,
and 20,000 points and computed the influence of these data
sizes on the DOM loading time, window loading time, and
graph rendering time. The DOM loading time is the time
taken by the browser in parsing the HTML tags and
building the DOM tree. The window loading time refers to
the time adopted by the browser in loading different scripts,
parsing the HTML tags, and creating the DOM tree.
Finally, the graph rendering time is the windows loading
Figure: 7. Individual Journey of a patient with id 54924452
Figure: 8. Brushing and Visualization of selective dimension in parallel coordinate axes
time plus the time
taken for
combining CSSOM
and DOM trees into
a render tree and
computing the
layout of each
visible element on
the screen. Thus,
the graph rendering
time tells the entire
time taken for the
visualization to
appear after going
through all the
mentioned
processes.
Here as we
wanted to measure the performance of visualization charts
only and not the networking bandwidth related aspects.
Thus, we used a local server to keep the DOM loading time
and the window loading time about constant. Also, we
made sure that there were no external files for our browser
to load and we focussed on the graph rendering time with
increasing data set size. Figure 13 shows
different time durations with different
data set sizes and chart types. As can be
seen in the figure, the Parallel
Coordinates chart possessed the most
optimal rendering as it took the least
graph rendering time. The graph
rendering time of the Sankey charts was
worst; whereas, the Sunburst charts
possessed an intermediate graph
rendering time.
V. DISCUSSION AND CONCLUSION
Recent policy initiatives have made
health care data digitized, and this
digitization has provided researchers the
ability to create impacting visualizations
on medical data. As seen from our
results, different visualizations allowed
us to visualize multivariate patient
journeys both at the individual patient
level as well as at the aggregate level.
Overall, these visualizations in Patient
Journey Visualizer (PJV) tool allow
different stakeholders to analyze patient
journeys at various levels to identify
bottlenecks in the health care
system, boost the quality of
health care, and eliminate
unnecessary activities.
First, we believe that
patient journeys are
important to several stakeholders, and it is
important for stakeholders to understand their roles in these
journeys to improve them. Furthermore, one visualization
may not be sufficient for visualizing the patient journey for
different stakeholders. In fact, various stakeholders would
likely have different variables of interest and thus they
would need different visualizations to understand their
roles better. To satisfy different stakeholder interests, in
this paper, we highlighted the potential of using three
different visualizations in PJV: Parallel Coordinates charts,
Sankey charts, and Sunburst charts. The Parallel
Coordinates charts show the patient journeys on an
individual patient level and thus could help physicians in
understanding the journeys of patients across several
demographic variables, medications, procedures, and
diagnoses. The PJV tool also provides correlations with p-
values between different axes in the chart. However, this
correlation matrix may become significant for several
attributes. Thus, providing a simple ranked list of paired
dimensions that are significantly correlated may suffice in
future.
The PJV tool allows one to use filtering with the
Parallel Coordinates charts as well as aggregation of data
based upon this filtering. For putting filters on Patient IDs a
brushing feature was included; however, for Patient IDs, a
Figure: 9. Correlation matrix show correlation among
dimensions
Figure: 10. Findin g most popular procedure
Used for Rheumatoid Arthritis
Figure: 11. Showing the Journey and flow of patients from Rheumatoid Arthritis to Drug Methotrexate.
search box is likely to be more appropriate. We plan to
include this search box functionality as part of our future
work with PJV tool.
As discussed above, Sankey charts allow stakeholders
like pharmaceutical companies to visualize the mapping of
medications across a group of patients who have undergone
specific diagnoses and specific procedures. Thus, these
visualizations may help stakeholders like medicine
manufacturers for comparing their drug performance with
those of their peers. Also, they would help in improving the
reach of medications across several patients with varied
diagnoses and procedures.
Furthermore, our results also showed that Sunburst
charts could help stakeholders like health care insurers in
visualizing patient journeys in terms of hierarchy of events
like diagnoses made, procedures performed, and
medications refilled. Also, these charts allow stakeholders
to change the perspective and visualize the aggregate
patterns from any segment in the chart.
Lastly, the performance of various visualization charts
on different data set sizes revealed Parallel Coordinates
charts to be highly scalable on increasing data size: The
increase in its rendering time is almost linear as we increase
the data size. While in case of Sankey charts, the rendering
time increases almost 6 times on increasing data size from
2,000 points to 20,000 points. There is also a sharp increase
in rendering time for Sankey charts; but this increase is not
as much compared to Sunburst charts. Overall, our results
highlight the potential of using software tools with a mix of
different visualizations for analyzing patient journeys.
VI. FUTURE WORK
Patient journeys can help provide a roadmap to identify
bottlenecks and create better health care plans for patients.
In future, we plan to enhance the capabilities of our
software tool with more innovative visualizations that help
predict the impact of newer medications on patients as well
as analyze patient journeys of patients who are diagnosed
with similar diseases. For example, to explain the potential
of Sankey diagrams, we focused on showing only one step
in a patient's journey. As part of the future research, we
would like to extend this demonstration to include the
entire patient’s journey in the tool. We would also enhance
the visualizations of patient journeys in our software tool to
predict certain rare diseases and the diagnoses and
procedures associated with them.
ACKNOWLEDGEMENT
The project was supported from a grant (awards:
#IITM/CONS/RxDSI/VD/07) from Rx Data Science Inc,
USA, to Varun Dutt.
REFERENCES
[1] Yagi, E. (2016.March 25). Healthcare is an asset not a
liability (Online), Available:
https://councilcommunity.com/2016/03/25/healthcare-is-an-asset-
not-a-liability/
[2] Curry, J., McGregor, C., & Tracy, S. (2006, August). A
communication tool to improve the patient journey modeling
process. In Engineering in Medicine and Biology Society, 2006.
EMBS'06. 28th Annual International Conference of the IEEE (pp.
4726-4730). IEEE.
[3] Ben-Tovim, D. I., Dougherty, M. L., O'Connell, T. J., &
McGrath, K. M. (2008). Patient journeys: the process of clinical
redesign. Medical Journal of Australia, 188(6), S14.
[4] Macdonald, N., Shapiro, A., Bender, C., Paolantonio, M., &
Coombs, J. (2012). Experiences and perspectives on the GIST
patient journey. Patient Prefer Adherence, 6, 253-62.
Figure: 12. Patient Journey Visualized with Sunburst Chart.
Diagnosis I represe nts the ith diagnosis code. Similarly procedure
I refer to ith proce dure code. While Medication I refers to the ith
medication corresponding t o certain diagnoses and proc edures .
9
'999
8999
<999
=999
>999
?999
@999
A999
B999
'9999
Time&(in&ms)
Different&Visualization&Chart
C+2D4!E-$.-+#$3!
F#G-
H#$.)I!;)2.#$3!
F#G-
JKL!
;)2.#$3F#G-
Figure: 13. Performance analysis of various
visualization charts on varying data sizes
[5] Schwartz, E. (2017, May 1). Mapping the Patient Journey: A
Case Study Available: http://www.tandemseven.com/journey-
mapping/mapping-patient-journey-case-study/
[6] Trebble, T. M., Hansi, N., Hydes, T., Smith, M. A., & Baker,
M. (2010). Process mapping the patient journey through health
care: an introduction. BMJ, 341(7769), 394-397.
[7] NHS Modernization Agency. Process mapping, analysis and
redesign. London: Department of Health. 1-40 (2005).
[8] Yuan, X., Guo, P., Xiao, H., Zhou, H., & Qu, H. (2009).
Scattering points in parallel coordinates. IEEE Transactions on
Visualization and Computer Graphics, 15(6), 1001-1008.
[9] Perer, Adam, and David Gotz. "Visualizations to Support
Patient-Clinician Communication of Care Plans." In Proceedings
of ACM CHI Workshop on Patient-Clinician Communication
(2013). 2013.
[10] Krause, Josua, Narges Razavian, Enrico Bertini, and David
Sontag. "Visual Exploration of Temporal Data in Electronic
Medical Records." In AMIA. 2015.
[11] Monroe, Megan, Rongjian Lan, Hanseung Lee, Catherine
Plaisant, and Ben Shneiderman. "Temporal event sequence
simplification." IEEE transactions on visualization and computer
graphics 19, no. 12 (2013): 2227-2236.
[12] Big Data Analytics, Time-Series Database, Kx Systems,
http://www.kx.com.
[13] Chang, K., Parallel coordinates: A visual toolkit for
multidimensional detectives, (2014), GitHub repository,
https://github.com/syntagmatic/parallel-coordinates
[14] Kosara, R. Nutrient Contents-Parallel Coordinates, (2010)
GitHub repository,
https://syntagmatic.github.io/exposedata/parallel/
[15] Inselberg, A. "The plane with parallel coordinates." The
visual computer 1, no. 2 (1985): 69-91.
[16] Bostock, M. "Sankey Diagrams, 2012." URL: http://bostocks.
org/mike/Sankey.
[17] Sankey, H. Riall. "THE THERMAL EFFICIENCY OF
STEAM-ENGINES. (INCLUDING APPENDIXES)." In Minutes
of the Proceedings of the Institution of Civil Engineers, vol. 125,
no. 1896, pp. 182-212. Thomas Telford-ICE Virtual Library,
1896.
[18] Rodden, K. (2017, March 17) Sequences Sunburst (Online),
Available: https://bl.ocks.org/kerryrodden/7090426
[19] Stasko, J., & Zhang, E. (2000). Focus+ context display and
navigation techniques for enhancing radial, space-filling hierarchy
visualizations. In Information Visualization, 2000. InfoVis 2000.
IEEE Symposium on (pp. 57-65). IEEE.
[20] Bostock, M. (2017, March 08) Parallel Coordinates, (Online)
Available: https://bl.ocks.org/mbostock/1341021
[21] Bostock, M. (2017, April 26) Zoomable Sunburst (Online),
Available: https://bl.ocks.org/mbostock/4348373
[22] Perer, Adam, and David Gotz. "Visualizations to Support
Patient-Clinician Communication of Care Plans." In Proceedings
of ACM CHI Workshop on Patient-Clinician Communication
(2013). 2013.
... In this particular field, experts have a crucial task that heavily depends on how to process and understand the data insights. In this context, Kaushal et al. [71] and DeCroon et al. [31] use the 2D-NP-GLC to analyze the data corresponding to clinical trials. In addition, Nguyen et al. [104] present a system of data analysis where the PCPs are used to remark the pattern of patient profiles. ...
Article
Full-text available
Multidimensional data visualization is one of the primary foundations supporting data analysis used for understanding the hidden relationships between items and dimensions of complex data. The line-based visualization techniques are a fundamental class of multidimensional visualization techniques and cover an important set of methods that are relevant to the visual exploratory analysis. Recently, General Line Coordinates (GLCs) were introduced. These are losslessly line-based visualization techniques for multidimensional data. Particular cases of GLCs are the non-paired GLCs, which generalize the radial and parallel coordinates and have proved to be highly suitable for visualizing multidimensional data. In this context, we conduct a systematic paper review of the 2D non-paired GLC (2D-NP-GLC) visualization techniques present in the literature. We organize the 2D-NP-GLC contributions in a unified reference framework in which both the representations and the associated interactions are considered. Focusing jointly on these two criteria, we provide a useful common space for the design and development of 2D-NP-GLC techniques. Besides, this framework integrates the 2D-NP-GLC contributions and helps to identify under-explored areas that may be candidates for further research.
Article
Background: Simple visualizations in health research data, such as scatter plots, heat maps, and bar charts, typically present relationships between 2 variables. Interactive visualization methods allow for multiple related facets such as numerous risk factors to be studied simultaneously, leading to data insights through exploring trends and patterns from complex big health care data. The technique presents a powerful tool that can be used in combination with statistical analysis for knowledge discovery, hypothesis generation and testing, and decision support. Objective: The primary objective of this scoping review is to describe and summarize the evidence of interactive visualization applications, methods, and tools being used in population health and health services research (HSR) and their subdomains in the last 15 years, from January 1, 2005, to March 30, 2019. Our secondary objective is to describe the use cases, metrics, frameworks used, settings, target audience, goals, and co-design of applications. Methods: We adapted standard scoping review guidelines with a peer-reviewed search strategy: 2 independent researchers at each stage of screening and abstraction, with a third independent researcher to arbitrate conflicts and validate findings. A comprehensive abstraction platform was built to capture the data from diverse bodies of literature, primarily from the computer science and health care sectors. After screening 11,310 articles, we present findings from 56 applications from interrelated areas of population health and HSR, as well as their subdomains such as epidemiologic surveillance, health resource planning, access, and use and costs among diverse clinical and demographic populations. Results: In this companion review to our earlier systematic synthesis of the literature on visual analytics applications, we present findings in 6 major themes of interactive visualization applications developed for 8 major problem categories. We found a wide application of interactive visualization methods, the major ones being epidemiologic surveillance for infectious disease, resource planning, health service monitoring and quality, and studying medication use patterns. The data sources included mostly secondary administrative and electronic medical record data. In addition, at least two-thirds of the applications involved participatory co-design approaches while introducing a distinct category, embedded research, within co-design initiatives. These applications were in response to an identified need for data-driven insights into knowledge generation and decision support. We further discuss the opportunities stemming from the use of interactive visualization methods in studying global health; inequities, including social determinants of health; and other related areas. We also allude to the challenges in the uptake of these methods. Conclusions: Visualization in health has strong historical roots, with an upward trend in the use of these methods in population health and HSR. Such applications are being fast used by academic and health care agencies for knowledge discovery, hypotheses generation, and decision support. International registered report identifier (irrid): RR2-10.2196/14019.
Article
Background Sacubitril/valsartan, a first‐in‐class angiotensin receptor neprilysin inhibitor, received US Food and Drug Administration approval in 2015 for heart failure with reduced ejection fraction (HFrEF). Our objective was to describe the sacubitril/valsartan initiation rate, associated characteristics, and 6‐month follow‐up dosing among veterans with HFrEF who are renin‐angiotensin‐aldosterone system inhibitor (RAASi) naïve. Methods and Results Retrospective cohort study of veterans with HFrEF who are RAASi naïve defined as left ventricular ejection fraction (LVEF) ≤40%; ≥1 in/outpatient heart failure visit, first RAASi (sacubitril/valsartan, angiotensin‐converting enzyme inhibitor [ACEI]), or angiotensin‐II receptor blocker [ARB]) fill from July 2015 to June 2019. Characteristics associated with sacubitril/valsartan initiation were identified using Poisson regression models. From July 2015 to June 2019, we identified 3458 sacubitril/valsartan and 29 367 ACEI or ARB initiators among veterans with HFrEF who are RAASi naïve. Sacubitril/valsartan initiation increased from 0% to 26.5%. Sacubitril/valsartan (versus ACEI or ARB) initiators were less likely to have histories of stroke, myocardial infarction, or hypertension and more likely to be older and have diabetes mellitus and lower LVEF. At 6‐month follow‐up, the prevalence of ≥50% target daily dose for sacubitril/valsartan, ACEI, and ARB initiators was 23.5%, 43.2%, and 47.1%, respectively. Conclusions Sacubitril/valsartan initiation for HFrEF in the Veterans Administration increased in the 4 years immediately following Food and Drug Administration approval. Sacubitril/valsartan (versus ACEI or ARB) initiators had fewer baseline cardiovascular comorbidities and the lowest proportion on ≥50% target daily dose at 6‐month follow‐up. Identifying the reasons for lower follow‐up dosing of sacubitril/valsartan could support guideline recommendations and quality improvement strategies for patients with HFrEF.
Article
Full-text available
The tyrosine kinase inhibitor (TKI) imatinib has improved outcomes for patients with unresectable or metastatic gastrointestinal stromal tumors (GIST), and for patients receiving adjuvant therapy following GIST resection. This qualitative study explored the experiences and emotions of patients through GIST diagnosis, treatment initiation, disease control, and in some patients, loss of response and therapy switch. Ethnographic investigations were conducted, including semi- structured qualitative interviews of patients with resected or metastatic/unresectable GIST and their caregivers, from Canada (n = 15); the United States (n = 10); and Brazil, France, Germany, Russia, and Spain (n = 5 each). Some interviewees also kept 7-day photo journals. Responses were qualitatively analyzed to identify gaps and unmet needs where communication about disease, treatments, and adherence could be effective. Patients shared common experiences during each stage of disease management (crisis, hope, adaptation, new normal, and uncertainty). Patients felt a sense of crisis during diagnosis, followed by hope upon TKI therapy initiation. Over time, they came to adapt to their new lives (new normal) with cancer. With each follow-up, patients confronted the uncertainty of becoming TKI resistant and the possible need to switch therapy. During uncertainty many patients sought new information regarding GIST. Cases of disease progression and drug switching caused patients to revert to crisis and restart their emotional journey. Patients with primary or unresectable/metastatic GIST shared similar journeys, especially regarding uncertainty, although differences in the scope and timing of phases were observed. Strategies patients used to remain adherent included obtaining family support, setting reminder mechanisms, taking medicine at routine times, and storing medicine in prominent places. Physicians and support staff can manage patient expectations and encourage adherence to therapy, which may facilitate optimal patient outcomes. Patient education about current GIST developments and adherence across all phases of the patient journey are of benefit.
Article
Full-text available
Quality improvement is high on the agenda of Health Care Organisations (HCO) worldwide. Patient journey modeling is a relatively recent innovation in healthcare quality improvement that models the patient's movement through the HCO by viewing it from a patient centric perspective. Critical to the success of the redesigning care process is the involvement of all stakeholders and their commitment to actively participate in the process. Tools which promote this type of communication are a critical enabler that can significantly affect the overall process redesign outcomes. Such a tool must also be able to incorporate additional factors such as relevant policies and procedures, staff roles, system usage and measurements such as process time and cost. This paper presents a graphically based communication tool that can be used as part of the patient journey modeling process to promote stakeholder involvement, commitment and ownership as well highlighting the relationship of other relevant variables that contribute to the patient's journey. Examples of how the tool has been used and the framework employed are demonstrated via a midwife-led primary care case study. A key contribution of this research is the provision of a graphical communication framework that is simple to use, is easily understood by a diverse range of stakeholders and enables ready recognition of patient journey issues. Results include strong stakeholder buy-in and significant enhancement to the overall design of the future patient journey. Initial results indicate that the use of such a communication tool can improve the patient journey modeling process and the overall quality improvement outcomes.
Article
Full-text available
*Clinical process redesign is a successful improvement method that has been used to increase access to health services in 60 public hospitals across New South Wales, and at Flinders Medical Centre (FMC) in South Australia. *The method focuses on the patient journey as the primary improvement locus, and uses process mapping to identify the value-adding steps in that journey; it involves redesign teams identifying and eliminating non-value-adding steps to improve flow and reduce delays in access to emergency and elective care. *The method engages clinicians, managers, patients and carers, and delivers real gains in health care delivery. *This article outlines the clinical process redesign programs being used by NSW Health and at FMC.
Conference Paper
Full-text available
Radial, space-filling visualizations can be useful for depicting information hierarchies, but they suffer from one major problem. As the hierarchy grows in size, many items become small, peripheral slices that are difficult to distinguish. We have developed three visualization/interaction techniques that provide flexible browsing of the display. The techniques allow viewers to examine the small items in detail while providing context within the entire information hierarchy. Additionally, smooth transitions between views help users maintain orientation within the complete information space
Article
Electronic Health Records (EHRs) have emerged as a cost-effective data source for conducting medical research. The difficulty in using EHRs for research purposes, however, is that both patient selection and record analysis must be conducted across very large, and typically very noisy datasets. Our previous work introduced EventFlow, a visualization tool that transforms an entire dataset of temporal event records into an aggregated display, allowing researchers to analyze population-level patterns and trends. As datasets become larger and more varied, however, it becomes increasingly difficult to provide a succinct, summarizing display. This paper presents a series of user-driven data simplifications that allow researchers to pare event records down to their core elements. Furthermore, we present a novel metric for measuring visual complexity, and a language for codifying disjoint strategies into an overarching simplification framework. These simplifications were used by real-world researchers to gain new and valuable insights from initially overwhelming datasets.
Article
By means ofParallel Coordinates planar “graphs” of multivariate relations are obtained. Certain properties of the relationship correspond tothe geometrical properties of its graph. On the plane a point ←→ line duality with several interesting properties is induced. A new duality betweenbounded and unbounded convex sets and hstars (a generalization of hyperbolas) and between Convex Unions and Intersections is found. This motivates some efficient Convexity algorithms and other results inComputational Geometry. There is also a suprising “cusp” ←→ “inflection point” duality. The narrative ends with a preview of the corresponding results inR N .
Healthcare is an asset-not a liability (Online) Available: https://councilcommunity.com
  • E Yagi
Yagi, E. (2016.March 25). Healthcare is an asset-not a liability (Online), Available: https://councilcommunity.com/2016/03/25/healthcare-is-an-assetnot-a-liability/
May 1) Mapping the Patient Journey: A Case Study Available: http://www.tandemseven.com/journeymapping/mapping-patient-journey-case-study
  • E Schwartz
Schwartz, E. (2017, May 1). Mapping the Patient Journey: A Case Study Available: http://www.tandemseven.com/journeymapping/mapping-patient-journey-case-study/