Wearables for research: Fitbit Charge HR, Withings Activite Steel and Apple Watch I
For a couple of weeks, I had my hands on these three wearables (one may say a smart watch and two activity trackers but it is OK). I paired all with the iPhone 7 and tested considering 3 fundamental elements in a research (at least for me in my case):
1. Data collection
2. Reliability of data
3. Reliability of hardware and software
Before go any further, let me briefly tell about specs of these devices:
Fitbit Charge HR is capable of tracking step counts, sleep and activity detection, heart rate measurement. It has 5 days battery life. Apple Watch (1st gen) can stand a day, and do all the things that Fitbit does, with a larger and colorful screen (yet it might night be tracking your sleep because you are possibly be recharging it at night). Withings Activite Steel is the simplest one, it can keep step counts, sleep and activity detection. It has analog screen with only activity percentage display (which is also analog). The plus side is Withings can last longer as 8 months (as they say), and using standard watch batteries (which you need to replace it not charge it!). All has silent (vibration) alarms, but since Apple Wacth has mic and speaker, it may go un-silent.
1. Data collection
They all showed their potential for data collection in terms of basics: steps count (so calories and distance), sleep and activity detection. Fitbit and Apple Watch were able to capture heart rate. All have iOS apps to follow your progress and enter more information about yourself (nutrition, health and activity details...). Withings app has also a feature to measure heart rate with phone camera and flashlight (it is mostly consistent with Fitbit). It is possible to export data in CSV, TXT or XLS format from Fitbit and Withings web interface. If you need to go further to get detailed data (such as, heart rate for each 5 secs), I suggest you to check dev.fitbit.com for customizing your way to collect data.
2. Reliability of data
In short, I cannot say these devices perfectly accurate. Yet, in the research, instead of getting the exact step count or heart beat, I look for consistency, which in return may help to get a reliable ratio. For instance, if heart beat elevated from 90 to 100, but I got reading from 87 to 97, that is enough evidence to statistically approve my hypothesis.
In further investigation, I compared their step counts and heart rate measures (except Withings). I used in every combination possible for each watch (same arm together, cross arms...) to see if there will be change in data among the devices. Result, no significant change was observed. Fitbit and Apple is close in terms of step counting, but they are mostly overcounting comparing to Withings. The ratio is roughly 20%. Here is some raw daily data for Fitbit/ withings: 15147/ 12751, 6489/5194, 7084/5271.
Sleep detection and duration is closely measured by all of the devices (max diff. 20 mins) and activity detection (max diff. 5 mins)
3.Reliability of hardware and software
Among all, Fitbit looks more durable, its hard silicon body looks sound. and apple less.
--Withings is slower in connecting and sync.
--First impression: Apple seems to be more complicated, easier to get external damages and iOS failures, and may be unable for sleep time monitoring (due to charging at night). Apple is better for programming the prompts and interventions.
Bonus: Comfortability & engagement
We need our participants to be comfortable and engaging in a research, right? We provide cozy environment, snacks and beverages during an interview. But for a longitudinal research, it may be hard to achieve. Thus, we need to be sure data collection continuously occurs. In our case, comfortable wearables and the less user involvement would achieve our goal. Comfort mainly assessed by the physical sense (how well it fits to your arm without any extra weight, discomfort or irritation), and also by the sociological (how well it fits to you in terms of your outfit, life style and environment...). So, from the physical sense, I can say all 3 wearables are quite comfortable.
For the engagement, the less user involvement in data collection is priority. If possible, the device and app should be synced continuously, and the researcher should be able to access and retrieve the data, without asking each day to the participant.
But in case of user involvement for engaging to the study, I believe smart phone will do the same effect as smart watch. Thus, for instance, just to show text reminders, using Apple watch may be a trade off for sleep tracking, which may not be tolerable for your research (it is your call).
--They both need manual initiation of apps in order to sync with the smart phone. Participants may be notified once in a day to sync (mornings maybe for data of previous day including sleeping activities) - Or a script may be used to auto-initiate the app
--Once fitbit is fully charged, the app shows push notification on iPhone
--In activities that is not including walking or running ( like pingpong), you need to specify what activity you were doing in the app.
Need more apps with your wearables?
Gyroscope, Life cycle, Moves, Journeys are some of the apps that I have tried. They provides graphical feedback about your activities. But if you need more out of your raw activity data, I would suggest to use less popular apps from individual developers. Such as, HascLogger can help you to collect accelerometer, gyroscope, magnometer and location data with a given frequency. You can do detailed analysis on your daily activities. Downside of these apps are that there are no promises to secure your data from being shared with third parties. So be careful about that.
If you need any details about this topic, let me know and I can edit this post to add some more.
Thanks for reading.
Do you like it? So, do not miss this post: Waldo: Patient with wearables