(J Med Internet Res 2020;22(6):e15547) doi: 10.2196/15547
KEYWORDS
mHealth; crowdsensing; tinnitus; machine learning; mobile operating system differences; ecological momentary assessment;
mobile phone
Introduction
Background
Mobile health (mHealth) uses smart mobile devices to address
various questions in the context of neuroscience, psychology,
and medicine. New paradigms, such as ecological momentary
assessment (EMA), mobile crowdsourcing, and mobile
crowdsensing, as well as mHealth apps, in general, have enabled
data collection procedures that surpass many existing methods
in gathering valuable medical data by several orders of
magnitude [1]. Among others, by using smart mobile devices,
data can be gathered in everyday life, on a cost-effective basis,
and by adding contextual information sources, such as Twitter
or Facebook. As many medical phenomena pose daily variations
[2], mHealth technology is predestined to be utilized in this
context. Along these trends, many insights have been presented
by researchers that show that smart mobile devices can help to
establish new data sources in many scenarios [3].
In these data collection scenarios, which are built on the usage
of mobile devices and their sensors, one dimension has been
less considered so far. It refers to the question of whether the
operating system (OS) of the mobile technology being used (eg,
iOS or Android) constitutes a valuable information source or
confounder for medical data analyses. Or, as another example,
is it possible to derive insights if a patient changes the OS during
a study when using mHealth apps? As Android and iOS
dominate the mobile OS market [4]—with a market share of
99.32% in May 2020 (72.52% Android and 26.80% iOS)—any
insights gained based on differences from users regarding these
OS types could provide a representative picture for the OS
market. Following this, data that were gathered with the
TrackYourTinnitus (TYT) mHealth crowdsensing platform for
tinnitus patients over 5 years of age are analyzed in this paper.
TYT is an mHealth crowdsensing platform that offers iOS and
Android apps that can empower patients to learn more about
their tinnitus symptoms over time. Tinnitus is the phantom
perception of a sound and it is experienced by 5.1% to 42.7%
of the population worldwide at least once during their lifetime
[5]. The symptoms often reduce the patient’s quality of life. As
tinnitus constitutes a chronic condition for which currently no
cure or general treatment exists, patients suffering from it crave
for new treatment procedures or at least new medical insights.
With the idea of EMA, also known as ambulatory assessment
or experience sampling, and mobile crowdsensing techniques
in mind, TYT was developed by an interdisciplinary team of
medical experts, psychologists, and computer scientists.
The development of TYT was motivated by the clinical
experience that among many tinnitus patients, tinnitus loudness
and tinnitus annoyance vary over time and that patients’
experiences differ in the pattern of these fluctuations. Therefore,
the variations are considered to provide new valuable insights
in the pathophysiological mechanisms of this chronic condition
[6]. To learn more about these fluctuations, TYT applies EMA
and mobile crowdsensing to capture them. In EMA, the variable
in question (eg, a symptom) is assessed repeatedly in daily life
[7]. In mobile crowdsensing, only mobile devices are used for
the data collection procedure, while their sensors are used to
capture, for example, the GPS position or the external sound
level [8]. In contrast, in mobile crowdsourcing, tasks are
proposed by a crowdsourcer to a group of individuals, who
voluntarily undertake tasks. The undertaking of the task always
entails mutual benefit. The user will receive the satisfaction of
a given type of need, while the crowdsourcer will obtain and
utilize to their advantage what the user has brought to the
venture [9]. In contrast to mobile crowdsourcing, mobile
crowdsensing relies solely on mobile technology and integrates
sensors to collect data. Two recent works that discuss mobile
crowdsensing in the context of health care can be found in Kraft
et al [1] and Pryss [10]. In TYT, the users fill in a registration
questionnaire (ie, static data) and can provide repeated
assessments in daily life (ie, dynamic data) afterward [11].
Objectives
Compared to the existing studies on TYT, this work investigates
repeatedly provided EMA datasets from TYT users (ie, dynamic
data) and their relation to the mobile OS used. While this study
analyzes this dynamic data, a previous study focused on
differences between Android and iOS users in the static data
given at registration [12]. Contrary to the Android versus iOS
comparison of the SmokeFree28 (SF28) smoking cessation app
[13], in our study we found no differences in gender, but we
did find differences in age for TYT users. However, in Pryss et
al [12], we found differences that might be of interest for
medical purposes. More specifically, we revealed that Android
users reported a significantly longer tinnitus duration than did
iOS users, cross-sectionally. Future longitudinal research is
necessary to address the question of whether users with longer
tinnitus duration prefer Android to iOS or whether users of
Android tend to develop longer tinnitus durations than iOS
users. In another recent work [14], we investigated differences
in Android and iOS users of the TrackYourHearing (TYH)
mHealth crowdsensing platform. This platform aims to measure
fluctuations in hearing of users with hearing loss. In the TYH
study, we found no differences in gender or age, but significant
differences were revealed in three questions of the dynamic data
that were repeatedly provided. This shows that the dynamic
data in combination with the OS are worth being investigated
more deeply.
As another current trend, the application of machine learning
techniques in different fields is promising. In the medical field,
there is a remarkable discrepancy between huge expectations
in the potential of machine learning on one side and the current
application of this technique on the other [15-19]. Importantly,
there is an increasing consensus about its potential in the context
of mobile technology [20-23]. However, the application of
J Med Internet Res 2020 | vol. 22 | iss. 6 | e15547 | p. 2http://www.jmir.org/2020/6/e15547/
(page number not for citation purposes)
Pryss et alJOURNAL OF MEDICAL INTERNET RESEARCH
XSL
•
FO
RenderX