Designing Free-Living Reports for
Faculdade de Ciências
Universidade de Lisboa
Faculdade de Ciências
Universidade de Lisboa
Faculdade de Medicina
Universidade de Lisboa
Faculdade de Medicina
Universidade de Lisboa
Parkinson’s disease is a progressive neurodegenerative disorder that is also characterized by its motor
fluctuations throughout the day. This makes clinical assessment to be hard to accomplish in an
appointment as the patient status at the time may be largely dierent from his condition two hours
before. Clinicians can only evaluate patients from time to time, making symptom fluctuations diicult
to discern. The emergence of wearable sensors enabled the continuous monitoring of patients out of
the clinic, in a free-living environment. Although these sensors exist and they are being explored in
a research seing, there have been limited eorts in understanding which information and how it
should be presented to non-technical people, clinicians (and patients). To fill this gap, we started by
Permission to make digital or hard copies of part or all of this work for personal or classroom use is granted without fee
provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the
full citation on the first page. Copyrights for third-party components of this work must be honored. For all other uses, contact
CHI’19 Extended Abstracts, May 4–9, 2019, Glasgow, Scotland Uk
©2019 Copyright held by the owner/author(s).
ACM ISBN 978-1-4503-5971-9/19/05.
performing a focus group with clinicians to capture the information they would like to see devised
from free-living sensors, and the dierent levels of detail they envision. Building on the insights
collected, we developed a data-driven platform, DataPark, that presents usable visualizations of data
collected from a wearable tri-axial accelerometer. It enables report parameterization and includes
a baery of state-of-the-art algorithms to quantify physical activity, sleep, and clinical evaluations.
A two-month preliminary deployment in a rehabilitation clinic showed that patients feel rewarded
and included by receiving a report, and that the change in paradigm is not burdensome and adds
information for clinicians to support their decisions.
Parkinson’s disease; Free-living; wearable technology; Data-driven;Accelerometer
Parkinson’s disease (PD) is a frequent progressive neurodegenerative disorder, aecting about 1% of
the world population. The ageing of the population will increase the number of people living with
that disease in the following years. PD manifests itself by the reduction of dopamine levels, due to
the death of the brain cells that produce it. These phenomenon occurs only if more than seventy or
eighty percent of cells die [
]. This disease is characterized by tremors, rigidity of the trunk and
limbs and low movements. With the progression of the disease, the postural instability can be very
disabling, creating diiculties in the tasks of standing, siing, and walking.
One of the characteristics of PD is that the disease progression is highly variable and the symptoms,
alongside the degree of disability, are likely to fluctuate over the duration of a day [
can only evaluate patients from time to time, making symptom fluctuations diicult to discern.
Challenges for clinical practice include understanding the progression of the disease, the response
to pharmacological and non-pharmacological interventions, and the fluctuations the patient goes
through alongside their explanations .
One way for clinicians to understand what happened with the patients in a free-living environment
is by asking them questions, but this can be less precise than needed because recall is unreliable [
Recurring to diaries to help patients resume their day and provide useful information to the clinicians
is an alternative. However, there is a compliance issue when using diaries as well as a subjectivity one.
The democratization of sensing wearable technologies opened several possibilities in the continuous
monitoring of people, and particularly in what relates to their health and wellbeing [
]. The amount
of data these devices can produce, and the rich insights that can be derived from it, ask for a shi to a
data-driven consultation paradigm, that needs to be carefully designed .
We developed DataPark, a web platform with the purpose of helping clinicians obtain more infor-
mation about patients. Objective data is obtained with the usage of a tri-axial accelerometer sensor
(AX3). Clinicians can generate a personalized report according to the needs of each patient or the
analysis they want to perform.
We chose an iterative co-design approach where clinicians and HCI researchers craed all elements
and aesthetics of the platform and its reports. In addition, we performed a preliminary study with a
stable version of the platform that showed the benefits and limitations of DataPark and oered new
perspectives for its future.
Our findings suggest that a data-driven approach for geing a summarized analysis of objective
data from dierent environments can help clinicians to beer understand patients’ fluctuations during
their daily life. It also showed patients more engaged with therapy when they knew they were going to
receive a report. Conversely, albeit valuing the benefits of sharing reports with patients, the clinicians
also showed some concerns with providing patients with clear evidence of their motor decline.
Figure 1: The setup in the focus group
Figure 2: An example of some of the digital
boards created with Trello (trello.com; last
visited on 07/01/2019)
Our first goal was to understand what were the needs of clinicians, which events or measures they
would like to see collected, and how they would envision it to be presented. First, we conducted
informal interviews with clinicians and observations of clinical assessments with patients. This enabled
us to understand the complexity of assessing a person with PD, even when in close proximity. Our
main study was then a focus group to capture a broader perspective of what were their needs.
Five participants took part in the session. It was composed of one neurologist, two physiotherapists,
and two nurses. The focus group was prepared around a set of dierent boards (Devices, Activities,
Data) that were iteratively filled with information (e.g., post-its) by the participants or the researchers
as they emerged. For each one of the dierent boards we asked participants to write down examples.
For Devices, we asked participants to write down devices (or objects) they would like to sense, i.e.
where they know that understanding usage paerns of that object could be relevant to monitor as
a proxy for disease progression; examples of such devices were mobile phone, TV remote, clothes,
glass, toothbrush. As for Activities, we were looking for daily Activities that they found relevant to be
reported in a free-living assessment context; examples were sleep, walk, swim, get dressed, ride a bike,
among others. In the Data board, they were asked to think about data they could see derived from
the aforementioned devices and activities, and write down / discuss the most relevant they could
think of; Sleep and Gait measures were the most discussed, which is aligned with the focus of the
literature in the area of sensing technologies for PD [
]. Gait measures include step length, step
time, number of steps, for example (Figure 1).
In the end, we challenged each of the participants to list or draw a possible report and enumerate
some important points that should be part of it. Sleep, gait, and physical activity were referenced by
all the clinicians.
To allow for participants to continue the discussions aer the session and consolidate the infor-
mation, we created a digital version of the paper boards as they were at the end of the meeting in
Trello (Figure 2). We asked participants to complete all the information they could and asked them
questions to elicit action (e.g., "you mention a couple of objects (e.g., a book). What type of useful
data you see being collected from those objects that can inform on the status of a person with PD?").
A new board called Scenarios was created with the purpose of capturing real-world scenarios that
could be relevant for PD, so we could retrieve more data from those examples. One example was the
usage of sensors to detect freezing of gait that could then issue a request for the patient to confirm
Figure 3: Omgui tool and the output: a
chart of vector magnitude distribution
and a CSV file about the same vector mag-
Figure 4: An example template of a Data-
Park report, containing energy, charts, a
physical activity chart and measures
We grouped all the information from the boards and the example reports and informedly designed
a first prototype with digital reports derived from wearable accelerometer data. Our purpose, in this
initial phase, was to gather as many data as possible so we could have a picture of the clinicians’
We created a data-driven platform, DataPark, to collect objective data from uncontrolled environments.
The output is a report that includes measures about physical activity and sleep.
AX3 and Data
We used an inertial sensor, Axivity AX3, for geing the objective movement data from the patients
day-to-day. This device already makes available with an analysis open-source platform. However,
the data is not easy to understand for non-technical people. The output provided consists in a chart
with the raw data and a set of CSV files with the analyzed data (see Figure 3). They make available
algorithms about vector magnitude, wear time and sleep analysis. The output needs to be prepared
and exhaustive analyzed for extracting relevant information from it. This is not easy to do by clinicians,
mostly because they do not have the time needed but also, in some cases, the skill set to get meaningful
information from the raw or the processed but text-based data.
Our approach focused on oering clinicians the data in an easy understandable way. For that purpose,
we gathered all the data we found relevant and focus our platform in the design process. For each
data measure we presented, we had to decide what was the best way for presenting it. We did this
together with our clinical partners as we learned that their experience and habits with visualizing
medical data had a strong influence in the acceptance and understanding of dierent visualizations.
Each report is composed by a set of charts, tables and measures that gives a processed analysis
over the raw data (Figure 4). In this proof-of-concept, our analysis focused on physical activity, energy
and sleep. The first one give us the dierent positions patients were during each day. Energy shows
how much kcal were spent by the patients. Sleep consists in understanding the fluctuations patients
may undergo during the night period. It measures changes in sleep positions and wake-ups.
All the dierent modules produce a variety of charts. We use dierent types of views for data, for
example by day, or week, or period of the day. If needed, clinicians can adjust each report according
to the patients or the type of analysis they want to see or discuss. Also, we give the change for
clinicians to build their own reports, by choosing what information should appear and how it should
be presented. In the platform, we save the previous evaluations of each patient. We allow comparisons
between dierent evaluations. Reports can be printed or saved as PDF for further analysis.
We deployed DataPark in a rehabilitation clinic and residential unit for neurodegenerative diseases,
for a period of two months. We wanted to understand the qualities and limitations of the output
data and understand how the platform influences the clinical practice. Twenty-two patients with PD
agreed to participate in our IRB-approved study; one neurologist and three therapists used DataPark
during the study. Patients were asked to wear an inertial sensor for a period of three or seven days,
depending on the rehabilitation program they were associated with. At the end, clinicians had access
to a report about the period in evaluation. We conducted semi-structured interviews to understand
how the platform influenced the normal workflow of dealing with the patient.
We received the feedback from the clinicians’ perspective and their perception on the patients’
perspective. Patients enjoyed to have an overview of how their week was. Examples of what they like
the most are energy expenditure (expressed in kcal) and sleep analysis, as they can provide meaning
to it. Some of the patients had previously participated in another evaluation and get excited for being
able to participate again. The participants have been anxious for receiving a summary report that
somehow can give them a perspective of how their week was.
This study was performed in a free-living context (not controlled task), however patients stayed at
the clinic, so they were outside their home environment. Patients knew they were being evaluated,
and their behavior could be dierent if they were at their homes.
Clinicians reported no changes in taking care of patients by using sensors. This shows that inertial
sensors do not influence the normal procedure. The use of the web application did not show to have
an influence in the time to perform the tasks. In the same way, there were no major diiculties on
using the web application neither the learning time was large.
The functional reports were of easy comprehension. All the data presented in the reports was
important for clinicians and they considered that having this type of monitoring gives the opportunity
for beer understanding patients in a free-living context. Clinicians pointed negative aspects of using
"The comparison of the data in dierent periods, if there was a regression in the results, it could lead to
"If geing the data and understand it is a very complex process, it would not work and could harm the
They also pointed positive aspects:
"Allow to have a more objective perception of the results of each stage of the evaluation."
"Possibility to have a more real perspective on the functional state of the patient in their environment."
CONCLUSIONS AND FUTURE WORK
There is a variety of objective sensor data that can be useful for monitoring and assessment of PD.
However, there is the need to design interfaces that allow for clinicians to interpret and benefit from
this data. If we are able to achieve this early, then clinicians will be beer equipped to collaborate with
researchers and designers in developing the next generation of data-driven consultation platforms
This work was was partially supported by FCT through the LASIGE Research Unit, ref. UID/CEC/00408/2019.
We would like to thank all participants of our studies.
Silvia Del Din, Alan Godfrey, Brook Galna, Sue Lord, and Lynn Rochester. 2016. Free-living gait characteristics in ageing
and Parkinson’s disease: impact of environment and ambulatory bout length. Journal of neuroengineering and rehabilitation
13, 1 (2016), 46.
Silvia Del Din, Alan Godfrey, Claudia Mazzà, Sue Lord, and Lynn Rochester. 2016. Free-living monitoring of Parkinson’s
disease: Lessons from the field. Movement Disorders 31, 9 (2016), 1293–1313.
Jacqueline M Graham and Harvey J Sagar. 1999. A data-driven approach to the study of heterogeneity in idiopathic
Parkinson’s disease: identification of three distinct subtypes. Movement disorders: oicial journal of the Movement Disorder
Society 14, 1 (1999), 10–20.
Michael R Huord, Arthur A Stone, Saul Shiman, Joseph E Schwartz, and Joan E Broderick. 2002. Paper vs. electronic
diaries. Applied Clinical Trials 11, 8 (2002), 38–43.
Yoojung Kim, Eunyoung Heo, Hyunjeong Lee, Sookyoung Ji, Jueun Choi, Jeong-Whun Kim, Joongseek Lee, and Sooyoung
Yoo. 2017. Prescribing 10,000 steps like aspirin: designing a novel interface for data-driven medical consultations. In
Proceedings of the 2017 CHI Conference on Human Factors in Computing Systems. ACM, 5787–5799.
Helena M Mentis, Anita Komlodi, Katrina Schrader, Michael Phipps, Ann Gruber-Baldini, Karen Yarbrough, and Lisa
Shulman. 2017. Craing a view of self-tracking data in the clinical visit. In Proceedings of the 2017 CHI Conference on
Human Factors in Computing Systems. ACM, 5800–5812.
Alan Michael Woods, Mariusz Nowostawski, Elizabeth A Franz, and Martin Purvis. 2014. Parkinson’s disease and essential
tremor classification on mobile device. Pervasive and Mobile Computing 13 (2014), 1–12.