Dataset of fitness trackers and smartwatches to measuring physical activity in research

Accelerometer-based wrist-worn fitness trackers and smartwatches (wearables) appeared on the consumer market in 2011 [1]. Many wearable devices have been released since. In addition to being used for personal physical activity tracking, these wearables are also increasingly being used in research as tools to collect health data [2–4]. These wearables are especially designed for long-term usage and thus facilitates long-term recording with low participant burden.

The objective of this data paper is to describe a dataset containing most in-scope wearables (n = 423) released before July 2017. In a study published in 2018, where this data set was used, we showed how sensor support changed over time, and that (in addition to accelerometer) optical pulse sensor (i.e., photoplethysmograph [PPG]) and global positioning system (GPS) were the two most common sensors [1].

The dataset contains information on 423 wrist-worn consumer-based wearables, produced by 132 different companies. The dataset is stored at DataverseNO [5]. A ReadMe-file (Table 1, Data file 1) describes the content of the dataset. The data are stored in two files, containing the same information using different formats, where “data.csv” (Table 1, Dataset 1) is a comma-separated file and “data.txt” (Table 1, dataset 2) is a tab-separated file. See Table 1.

Data were collected in two steps. In step one, we identified wearables and extracted information from six online and offline databases: The Queen’s University’s Wearable Device Inventory [6, 7], The Vandrico Wearables database [8], GsmArena.com [9], Wearables.com [10], SpecBucket.com [11], and PrisGuide.no [12]. In step two, we visited websites for all companies/brands identified in step one to identify additional wearables and extract additional information for each identified wearable.

Conflicting information from the databases was resolved by visiting brand websites. If no website could be found, or the website did not contain the needed information, we used Google search and other online resources (e.g., Wikipedia).

We collected 12 variables for each wearable. Six variables were related to sensor support and six variables were background attributes (i.e., meta data). The six sensor variables are: (1) accelerometer, (2) magnetometer, (3) gyroscope, (4) altimeter or barometer, (5) GPS, and (6) PPG.

The six background attributes are: (1) wearable name, (2) company/brand name, (3) release year, (4) country of origin, (5) whether the wearable was crowd funded, and (6) form factor (fitness tracker or smartwatch). An overview of each variable with a short description is available in the 00_ReadMe-file (Table 1, Data file 1).

The search was conducted between May 15th and July 1st, 2017. Inclusion criteria were: (1) wrist-worn, (2) consumer-based, (3) equipped with accelerometer for physical activity monitoring, and (4) capable of transferring data to connected smartphone.

Despite a thorough search of wearables, the list is likely not complete. During the search we only used English or Norwegian sources, or sources that could be translated using Google Translate. It is therefore possible that among the wearables we did not identify, a predominance of non-western countries is represented. We used multiple sources to map wearable capabilities, sometimes with conflicting information. We visited official vendor websites to identify the correct information, but these were not always available, and some misclassifications may thus exist.

The data set contains variables released between 2011 and 2017, as the search was conducted in 2017. We invite other researchers to update this list with newer models released after 2017, using this data set as a starting point.

This work was supported by UiT The Arctic University of Norway thematic priority Grant “Personalized medicine for public health”.

AH, AZW, GH, LAH, and SG conceptualized the overall project. AH and MM contributed to the data collection. AH drafted the manuscript, with critical review from all authors. All authors read and approved the final manuscript.

Open access funding provided by UiT The Arctic University of Norway (incl University Hospital of North Norway). This work was supported by UiT The Arctic University of Norway thematic priority Grant “Personalized medicine for public health”.

The data described in this Data note can be freely and openly accessed on DataverseNO under https://doi.org/10.18710/6ZWC9Z↗. Please see Table 1 and references [5] for details and links to the data.