Evaluation of a novel canine activity monitor for at-home physical activity analysis

Activity data provided by the Whistle device includes the total activity in minutes
per day and a visual activity intensity display (Fig. 2). This study revealed a strong correlation for total activity obtained by the two
accelerometers. TAW is a single number and can be recorded (but not exported) from
the Whistle application. TAW is also emailed weekly to the ‘owners’ of a specific
dog which may provide another means of obtaining this data in clinical research. We
also found a strong correlation for activity intensity data between the two devices.
However, the method used to derive the intensity data from the Whistle is time-consuming
and cumbersome, limiting the use of this feature.

The Whistle device pairs with any smartphone/tablet via a free application allowing
dog owners to create a profile for their dog, set daily PA goals, track their dogs
PA, log food consumption, track medication administration (as well as receiving medication
reminders), add photos, and enter notes throughout the day. The application also allows
for multiple owners to log in, track, and manage activity through their mobile device.
PA data collected from the Whistle is synced hourly with the owners’ smartphone or
tablet which the Whistle is linked to. This eliminates the need for device-removal
for data retrieval and allows real-time monitoring of PA. Real-time monitoring and
the collection of objective and subjective data in one location offers a wide variety
of opportunities for future research/clinical use. This may include tracking of post-operative
clinical progression, response to specific treatments (since all data is ‘time-stamped’)
and two-way communication with veterinarians including the sharing of pictures (such
as for evaluation of wound healing etc.). Furthermore, measurement of joined activity
between people and dogs is feasible using the Whistle device and Jawbone wristband.
This may be of interest for evaluation of the impact of the human-animal bond and
activity on both, human and dog health 15]–17].

We chose the Actical activity monitor for comparison since it is a widely researched
accelerometer that has been previously validated for the use in dogs 2]–4], 6], 10], 11] and cats 5]. A recent study that evaluated the Actical in dogs found that there was 100 % specificity
and 100 % sensitivity in distinguishing sedentary activity from walking activity and
a 92 % specificity and 92 % sensitivity in distinguishing trotting activity from walking
activity 4]. Another study found that the Actical is highly effective in differentiating sedentary
activity from various degrees of activity in healthy dogs moving on a treadmill 11]. If the Whistle were to be used for activity intensity data collection, further investigations
determining cut points between light, moderate, and vigorous activities may be indicated.
Limitations of the Actical include its cost (at the time of writing priced at $450),
the lack of real-time monitoring and the restriction to PA/energy expenditure measurement
only. To access Actical data, the device must be physically removed from the protective
housing/collar and placed onto the ActiReader, which must be connected to a PC. The
data is then imported into the proprietary Actical Activity and Energy Expenditure
analysis software and can be interpreted using this software or be exported into Microsoft
Excel for further analysis. There are several disadvantages of the Whistle device
including the lack of exportable data and battery life. As mentioned previously, to
record TAW the researcher has to manually record this number either from the smartphone
app or the weekly email sent to ‘owners’. The Whistle also uses proprietary software
and the algorithms behind the data calculation are unknown. The battery life of the
Whistle is approximately 7 days (compared to 240 days of the Actical), however, the
device comes with a USB-charger for at-home use and charges within approximately 2 h.
For long-term clinical studies the short battery life proposes a significant challenge
for multiple reasons: Firstly, if owners forget charging their respective device,
PA activity data recording is interrupted. Secondly, the dog’s activity during charging
is not recorded. Lastly, while real-time data acquisition is a potentially useful
feature, it eliminates the possibility of owner ‘blinding’ to their dog’s PA. Blinding
would be feasible if owners were not allowed to pair their dog’s device, however,
owner access to the smartphone app is necessary to check the device’s battery status.
Since the battery lasts only a week, a study design where owners would be provided
with a fully charged device would not allow for long-term outcome measurement. Furthermore,
real-time monitoring is only feasible if the device is paired to the owner’s Wi-Fi
network which requires some technical knowledge and a Bluetooth connection. This may
limit participation of owners without Wi-Fi availability or the technical knowledge
to perform the set-up.

The current study has several limitations. The Actical epoch length was set to 1 min
whereas the Whistle is pre-set for an epoch length of 3 min. Since raw data was not
available for the Whistle device, it is difficult to evaluate whether activities are
recorded at the exact same moment. Data exploration suggested that less active dogs
showed higher Whistle scores, however, this is difficult to confirm without availability
of raw data. A second limitation is that all dogs participating in this study were
large breed dogs. Therefore the results of this study may not apply to smaller dogs
or cats. Only one device was used for this study, hence we can not comment on inter-device
variability. Data evaluation for total activity data was performed over a 24-h time
period. Previous studies have suggested a 7-day sampling interval for long-term clinical
studies to account for differences in activity observed between week-days and weekend-days
12]. However, the purpose of this study was comparison of the two devices rather than
evaluation of the study participants themselves. Lastly, data evaluation was only
performed for a 17-h time period for the intensity data evaluation. However, this
is consistent with previous canine 3] and human studies 18] and since each time point is evaluated individually, this should not affect our results.