Instrumentation & Measurement Magazine 24-6 - 71

Fig. 2. The CART initiative platform architecture. The front end provides an overview of sensors, the hub computer, and applications operating in participant
homes. The back end provides an overview of the data storage and processing systems for data received from the hub computer. API, application programming
interface; CART, Collaborative Aging Research Using Technology; DMZ, demilitarized zone; EHR, electronic health record; GPS, Global Positioning System; VPN,
virtual private network. (Reproduced with permission from S. Karger AG, Basel. The final, published version of this figure is available at
http://karger.com/?doi=10.1159/000512208.)
development of technology assisted remote monitoring
platforms needs to consider issues related to validation of sensor-based
data, the scalability of platforms, standards for data
collection and security, technical issues with platform development,
and the selection of sensors to collect data on specific
physical and health-related measures. Variations in these aspects
of platform development have been a barrier for larger
scale implementation of these systems and reproducibility of
data collection between studies.
The need for accurate methods to assess and monitor individuals
outside of the clinic and hospital settings becomes
even more pressing in light of the implementation of physical
distancing measures due to the COVID-19 pandemic. This
led to a rapid shift in clinical assessments being performed
through telemedicine and virtual platforms and necessitated
modifications of protocols for data collection in research studies.
The CART initiative provides an open, sharable platform
to develop digital technologies and tools that can complement
standard methods of patient assessment and advance methodologies
related to in-home monitoring. Issues related to remote
monitoring and how these have been addressed in the development
of the CART platform are discussed in the following
section.
Validation of Sensor-based Outcomes
For home-based sensor technologies to be integrated into clinical
trials and monitoring of chronic diseases, there needs to
September 2021
be evidence of accuracy and reliability in detecting an outcome
measure. Currently, sensor-derived measures are often
compared against conventional in-clinic tests that traditionally
represent the " gold standard " for that variable. However,
this becomes more difficult to evaluate when there is evidence
that the objective measurements derived from digital technologies
outperform standard clinical assessment [33]. Previously
validated research grade devices, e.g., actigraphs, can be worn
simultaneously over a period of a few days in a subset of cohort
participants to evaluate the accuracy of sensors used in
the home-based platform for validation studies. For long-term
monitoring of general activity and step counts, commercially
available activity monitoring wristwatches are more practical.
The Withings Activité (used widely in CART) has good correlation
with step counts from a pedometer and the ActiGraph
but slightly underestimates total step counts [34]. An additional
method is to use witnessed or video recorded ground
truth training examples to validate sensor-based outcomes
being tested. For example, the accuracy of an algorithm derived
from in-home PIR sensors and door contact sensor data
to determine when subjects were out-of-home was evaluated
against ground truth data from cameras installed over doors
that captured subjects entering and exiting the house, was
found to be 94% sensitive and 97% specific [35]. Novel sensor-based
outcome measures will require validation through
comparison to other validated research devices or ground
truth data.
IEEE Instrumentation & Measurement Magazine
71

http://karger.com/?doi=10.1159/000512208

Instrumentation & Measurement Magazine 24-6

Table of Contents for the Digital Edition of Instrumentation & Measurement Magazine 24-6