The measurement of physical activity as an important health parameter is a widely-used procedure in public health. In studies, objective methods such as accelerometers are often favored over subjective methods (i.e. questionnaires) to assess physical activity because self-reports of physical activity have been shown to overrate true values by up to 28% in males and 40% in females [
1]. However, hip-worn accelerometers are not always worn reliably enough to accurately assess true levels of physical activity because of discomfort or simply because participants forget to wear them. This leads to false results or misinterpretation of the data. It has been shown that participants’ compliance expressed as a high wear-time in the long term, for example during intervention studies, is higher if the device is to be worn on the wrist instead of the hip due to the greater comfort and lower hindrance of the daily routine [
2]. A recent study showed that 98.6% of participants wearing wrist-worn devices but only 91.0% of participants wearing hip-worn devices fulfil the criteria of 600 min wearing time per day [
3].
Because they have a higher compliance, wrist-worn devices are increasingly used in research to assess physical activity in large-scale studies [
4,
5]. Measurement of physical activity and energy expenditure by wrist-worn devices have been validated. However, as stated in the UK-Biobank study [
4], a limitation of wrist-worn devices is that the optimal method to identify non-wear time remains elusive. A recent systematic review, which included all publications between January 2010 and December 2015 that are listed in PubMed and Web of Science and used the ActiGraph GT3X+ in their studies, analyzed the device with regard to data collection and data processing [
6]. The authors provide some guidance on how activity data should be recorded and how activity should be categorized and analyzed. However, regarding non-wear time detection, the authors stated that
“There is a need to thoroughly test this criterion” (Migueles et al., 2017, p. 1823). The matter of a missing validation concerns both wrist-worn and hip-worn devices and a valid wear-time recognition is crucial to distinguish non-wear time from sedentary behavior. Especially sedentary behavior needs to be detected correctly in epidemiological studies because it is an important health risk factor. In addition, participants’ wear-time is an important criterion to determine if data should be included in analyses or if data have to be excluded because of participants’ non-compliance. Conventionally, wear times ≥10 h per day are considered compliant wear [
7]. According to a review [
8], 51% of large-scale studies (participants:
n > 400) used ActiGraph (Pensacola, United States) accelerometers which makes it one of the most-used tools to measure physical activity. In all ActiGraph devices, two different algorithms can be selected to identify wear and non-wear time. The algorithms provided by the corresponding software ActiLife (version 6.13.3) are “Troiano 2007” and the updated and modified Version “Choi 2011”. To our knowledge, the Troiano algorithm is not validated and not published in any peer-reviewed journal and is solely based on the 2003–2004 NHANES (National Health and Nutrition Examination Survey, National Cancer Institute) dataset. Choi and colleagues [
9] modified this algorithm and validated it in the laboratory, however, a free-living validation is missing to this day.
The development of the algorithms was based on data collected with the Actigraph 7164 Physical Activity Monitor (Troiano) and Actigraph GT1M (Choi); however, nowadays the ActiGraph wGT3X+ is predominantly used and it is unclear if both devices use an identical data processing method. In addition, the algorithm was only validated for placement of the device on the hip but not on the wrist, which is the increasingly used placement for physical activity assessment today. To justify an even wider and more frequent use of wrist-worn accelerometers, validation of wear-time recognition under free-living rather than laboratory conditions is most important because a simulation of free-living behavior is not always possible in a laboratory setting. The primary objective of this study was, therefore, to validate the automatic wear-time recognition of wrist-worn and hip-worn ActiGraph wGT3X+ devices against self-reported non-wear time according to diaries in a free-living setting in healthy adults. The second aim was to analyze the average duration of single non-wear episodes and total duration of all non-wear episodes throughout the day. These durations are relevant to define the ideal cut-off value for the “minimum length of non-wear period” in the automatic detection of non-wear time and thereby decrease type II errors.