Globally, the percentage of individuals aged 65 years and older is estimated to increase from 9.3% in 2020 to 16.0% in 2050 (United Nations Department of Economic Social Affairs, 2020). Despite the increase in longevity, there remains a gap in healthspan, translating to high comorbidity and disability in later years of life which could be improved by lifestyle factors (Queen et al., 2020). Physical activity is essential for healthy aging, preventing and improving chronic illness and enabling independent living. Moderate to vigorous physical activity (MVPA) is advocated for older adults (Chodzko-Zajko et al., 2009), although age-specific factors need to be considered (Zaleski et al., 2016). A systematic review found no evidence of increased risk for MVPA in the setting of cardiovascular disease (Kraus et al., 2019). However, adherence to the recommended 150 to 300 min of weekly MVPA is low among older adults (Watson et al., 2016). Recent activity guidelines emphasize that any MVPA can provide health benefits (Physical Activity Guidelines Advisory Committee, 2018), even MVPA accumulated in bouts of less than 10 min (Jakicic et al., 2019). Feasible strategies to increase physical activity, especially MVPA, are needed in older adults.

Wrist-worn consumer accelerometry devices are broadly available now, providing an opportunity to promote higher intensity physical activity in older adults. Validation work examining various devices across different populations is robust, yet there is a paucity of research in older adults. Several important knowledge gaps need to be addressed given the unique gait dynamics and energy expenditure of older adults (Schrack et al., 2016). In comparison with research-level accelerometers, consumer accelerometry devices are well-validated to measure step counts in both younger (Evenson et al., 2015) and older adults (Farina and Lowry, 2018). However, findings are less consistent in measurement of MVPA, and are particularly scarce in older adults (Straiton et al., 2018). There are established definitions of MVPA based on acceleration thresholds measured by research-grade accelerometers, such as ActiGraph devices, and a standard wear location of the hip. The threshold, or cut-off value in counts/minute, may need to be lower in older adults (Barnett et al., 2016) than the general population (Troiano et al., 2008) given differences in oxygen consumption at different walking speeds across age groups (Hall et al., 2013; Barnett et al., 2016). Yet, the bearing on commercial activity devices, and thus clinical utility in free-living conditions, is not clear. Validations studies of consumer devices in older adults should consider different thresholds to define MVPA to allow for the lower resting metabolic rate. In addition, consumer devices are routinely worn on the wrist, raising the question of the impact device wear location which could be amplified in older adults with an average slower gait.

We previously conducted a pilot study in adults ≥65 years of age to explore the acceptability and usability of consumer activity wrist devices from six manufactures and found Garmin Vivosmart HR to score the highest and most preferable (Tocci et al., 2016). The purpose of this cross-sectional study was to determine the content validity of the Garmin Vivosmart HR compared to ActiGraph GT3X+ for the domains of daily step count and MVPA in adults ≥65 years of age. We performed pair-wise comparisons of data collected during free-living conditions between the wrist-worn Garmin device and two ActiGraph devices simultaneously worn on both hip and wrist.

Demographic characteristics and ambulatory function of the study population are summarized in Table 1. Although all participants completed the study and returned the devices, data was missing in some instances. GT3X+ hip data was incomplete (<10 h) in 4 participants; two forgot to reattach it after bathing the first morning, one had a broken waist clip, and one had low wear time. Among the remaining participants, the mean ±SD daily wear time was 17.6 ±4.9 h. GT3X+ wrist data was missing in one device. For the Garmin device, one participant had all data lost during download and another had missing intensity minutes.

Table 2 provides summary statistics of step count for each device as well as pair-wise test of differences. There was no significant difference in step count between the ActiGraph hip and Garmin devices. Daily step count collected by the ActiGraph wrist device was significantly higher than both ActiGraph hip device and the Garmin device (Figure 1). Similar results were found for the mean difference (bias) generated by Bland-Altman plots (Table 3). Correspondingly, the intraclass correlation (ICC) for daily step count showed good reliability between the Garmin and ActiGraph hip devices and moderate reliability in the other comparisons (Table 3).

Table 2 provides MVPA results for each device using different accelerometer cut-off values to define MVPA. Overall, the Garmin device overestimated MVPA compared with the ActiGraph hip device, the reference standard. However, this difference was small and not statistically significant when a MVPA cut-off value for older adults was used. Specifically, Garmin MVPA was only 18 min higher than ActiGraph MVPA defined as ≥1,013 counts/min and only 12 min higher than ActiGraph MVPA defined as >1,924 vector magnitude counts/min. In contrast, daily Garmin MVPA was 43 min higher than the ActiGraph MVPA defined as of ≥ 2,020 counts/min (p < 0.001). Different results were found when the Garmin device was compared to the ActiGraph wrist-worn device. The Garmin underestimated daily MVPA by 4-fold compared to ActiGraph wrist MVPA data for older adults (p values < 0.01) but was identical to ActiGraph wrist MVPA data device defined as ≥ 2,020 counts/min. From another perspective, MVPA was consistently higher for the ActiGraph wrist-worn device compared to both Garmin and hip-worn ActiGraph devices, regardless of the MVPA cut-off value (Figure 2).

The mean difference in MVPA between devices (bias) derived from Bland-Altman plots supported these findings (Figure 3). The smallest difference in MVPA minutes between the Garmin device and the ActiGraph hip device was the comparison using vector magnitude counts for older adults. Yet, despite this relatively small difference, there was a wide level of agreement [mean bias (LoA) min; −19.3 (−206.5,167.9)]. Similarly, the reliability for MVPA minutes for this comparison was poor (ICC = 0.38), although it was the best amongst all device comparisons and cut-off values (Table 3).

The objective of this study was to investigate the accuracy and reliability of the Garmin Vivosmart HR, a consumer-level activity wrist device, to measure MVPA in older adults in a free-living condition. The results contribute to the limited knowledge of estimating MVPA in older adults with consumer activity devices. When compared to the ActiGraph hip device, a research standard, the Garmin overestimates MVPA. However, this difference was small if compared to MVPA defined by age-specific thresholds, supporting the accuracy of the Garmin device in older adults to measure higher intensity physical activity. Research in more diverse geriatric patient populations is needed given our select group of men without impaired ambulation. Further, the low intraclass correlation coefficient of MVPA data collected by this Garmin device incites overall caution for use of commercial devices in the research setting. In contrast, the Garmin device's measure of daily step count is both accurate and reliable. Future research and development, targeted to older adults, is warranted to facilitate usage of consumer activity devices to increase MVPA.

Most of the work in consumer activity devices focuses on young, healthy adults. To date, the capability of consumer devices to measure MVPA remains lower than step counts. But interpretation is limited by differences in proprietary definitions of MVPA for each company. Ferguson and colleagues compared several consumer-level devices worn on either hip or wrist with the ActiGraph GT3X+ on the hip in 21 healthy adults (mean age 33 years)(Ferguson et al., 2015). The reliability for MVPA (ICC range 0.36–0.79) was less than step counts (all ICC > 0.90), although a Garmin device was not included. A similar study in younger adults also found a wide range in MVPA reliability (ICC 0.37- 0.66) with all the consumer devices worn on the wrist (Degroote et al., 2018). Differences in minutes of daily MVPA, defined by the Freedson Adult cut-point ≥1952 counts/min, were small compared to the ActiGraph device. Among the different manufacturers, which did not include Garmin, the Fitbit MVPA data had the best accuracy and reliability.

The use of consumer activity devices in older adults is understudied, especially MVPA. In free-living conditions of older adults, devices from the manufacturer Fitbit (Fitbit Inc., San Francisco CA) are the most widely tested (Straiton et al., 2018). Studies with various Fitbit models (One, Zip, Flex, Charge) contain older adults with comorbid conditions (Paul et al., 2015; Farina and Lowry, 2018) including significant cardiac or pulmonary disease (Alharbi et al., 2016; Boeselt et al., 2016; Thorup et al., 2017). Yet, only one study reports MVPA (Alharbi et al., 2016). Alharbi and colleagues recruited physically active participants in either a cardiac rehabilitation program (n = 38) or a community walking program (n = 10) to evaluate the Fitbit Flex against the ActiGraph GT3X+ (Alharbi et al., 2016). While not targeting older adults, the mean (range) age was 65.6 (52–85) years. Similar to our findings, the commercial activity device overestimated the daily MVPA using the AcitGraph cut-off of 2,020 counts/min and Fitbit website-derived values. While information on reliability was not provided, the study classified individuals by their compliance with ≥150 min of MVPA per week and showed the Fitbit to be sensitive, though not specific, compared to the ActiGraph. This raises the question that physical activity measured by commercial activity devices could be misclassified as light rather than MVPA in older adults, which could lead to misguided physical activity counseling and intensity classification efforts. Previous research has shown substantial differences in the metabolic cost of daily activities by disability status among older adults, with lower-functioning individuals expending more energy to complete tasks than higher-functioning individuals (Knaggs et al., 2011). Thus, the potential for MVPA to be classified as light intensity is greater in an older population.

We chose the Garmin Vivosmart HR device for our validation study based on the preference of participants in Gerofit, who represent the end-user group of older adults most likely to be willing, and capable, of engaging in MVPA (Tocci et al., 2016). A recent study by Tedesco and colleagues in 20 older (>65 years) healthy adults also studied the performance of the Garmin Vivosmart against an ActiGraph device using the definition of ≥2,020 count/minute for MVPA (Tedesco et al., 2019). Our slightly larger study with longer wear time collaborates their finding of overestimated MVPA minutes for Garmin Vivosmart using this cut-off for MVPA. However, we also included age-specific thresholds for MVPA in our analyses and found no significant differences in this case. Our novel data in older adults suggest that MVPA estimated by commercial devices may be more accurate in older adults than younger adults if we assume that a lower threshold for MVPA better classifies higher intensity activity. This supposition is supported by laboratory-based research which shows resting metabolic rate (RMR) is reduced 32% in older adults and leads to misclassification of activity intensities for 60% of walking conditions (Hall et al., 2013). Further, oxygen consumption at different walking speeds in older adults wearing an ActiGraph device supports lower cut-off values for MVPA(Barnett et al., 2016). Lastly, our results for vector magnitude (VM) counts and comparison with the ActiGraph device worn on the wrist supports the capacity of the tri-axial accelerometer in the Garmin Vivosmart to differentiate arm and bodymovement.

Despite these encouraging results on accuracy, the ongoing challenge of the reliability of consumer activity devices to measure MVPA remains. While our finding on the accuracy of the Garmin Vivosmart for MVPA in older adults is consistent with some other brands in younger adults (Degroote et al., 2018), the intra-person reliability remains moderate at best. Our ICC results for the Garmin Vivosmart falls within the 95% confidence interval of the report by Tedesco et al. (2019) (ICC (95% CI); 0.68 (0.18, 0.91). Our comparison of Bland-Altman plots of MVPA shows a wide limit of agreement regardless of MVPA definition and demonstrates variability in device error. These results highlight the problem with relying on consumer activity devices to measure MVPA in the research setting. For the clinician, it is unknown if variability or inconsistency in MVPA over time would be noticeable to the patient, and if so, adversely affect the utility of the activity device as a motivational tool. However, in a survey of elderly adults (mean age 87 years), participants rate accuracy to be as important as usability (Hergenroeder et al., 2019). Ultimately, randomized trials are needed to determine the efficacy of commercial devices to increase MVPA in older adults.

With regards to daily step counts, our results show that the Garmin Vivosmart and ActiGraph hip device provide similar values with good reliability. Our findings are consistent with validation studies in older adults for other commercial activity devices (Straiton et al., 2018) in addition to a recent study with the Garmin Vivosmart (Tedesco et al., 2019). Much of the research to date in older adults includes relatively healthy older adults or those participating in an exercise program, such as our study. Further research is needed in older adults who have impaired gait. In a treadmill study of 10 commercial activity devices in young adults, slower walking speed decreased device validity (Fokkema et al., 2017). Of note, these results support the Garmin Vivosmart, which performed better at slow and average walking speeds compared to other commercial devices. However, our findings cannot be generalized to all older adults given our participants had an average gait speed of 1.25 m/sec. Gait dynamics in the setting of assistive walking devices, especially walkers, presents additional challenges and warrants specific consideration (Hergenroeder et al., 2019).

Further discussion of the generalizability of our results is warranted. The study population was predominantly men and may not represent women of the same age group. Further, all participants were patients engaged in a clinical center-based exercise program. Although devices were not worn on days when exercise classes were attended, the median duration of 50 min of daily MVPA is higher than most older adults. Results may be different in sedentary individuals and those using an assist device or with poorly controlled medical problems. Lastly, because the algorithm to classify intensity of physical activity in commercial devices is not available from manufactures, future derivations may be different and make it impossible to compare results.

We present new findings on the accuracy of a commercial activity device to measure MVPA in older adults when lower, age-specific cut-off values are used. Results support the use of a commercial activity device to estimate MVPA in older adults, but further research is needed to test if adherence with MVPA guidelines can be improved. Reliability of measuring MVPA remains problematic and hopefully will be addressed with advances in technology. The goal to use these popular activity devices to increase the intensity of physical activity in older adults will be advanced if manufacturers disclose their proprietary definitions for MVPA.

The datasets presented in this article are not readily available because datasets generated and analyzed during the current study are not publicly available due to information that could compromise research participant consent and anonymity. Limited data sets are available from the corresponding author on reasonable request and subject to permission from the Veterans Health Administration (VHA). Requests to access the datasets should be directed to krisann.oursler@va.gov.

The studies involving human participants were reviewed and approved by Salem VAMC Institutional Review Board (IRB) and Research and Development Committees. The patients/participants provided their written informed consent to participate in this study.

BB, KO, and CJ analyzed the data and drafted the manuscript. KH, MM, and MCM provided critical revisions of this manuscript. All authors made substantial contributions in drafting or revising the manuscript and gave final approval of version submitted.

Gerofit was supported by the Veterans Health Administration (VHA) Office of Geriatrics and Extended Care Non-institutional Long-Term Care and Mentored Partnership program and the VHA Office of Rural Health. This study was supported in part by VHA RRD I01 RX000667. MCM and KH supported in part by the Duke OAIC NIH/NIA AG028716. KH was supported in part by VHA RRD RX003120. Support for open access publication fees was provided by the Research and Development Service, Salem VA Medical Center. The funders had no role in any aspect of the research.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.