Physical activity contributes to healthy aging, while it is increasingly recognized that sedentary behavior is also an independent determinant of health. The American College of Sports Medicine provides physical activity guidelines specific to older adults for functional ability, mortality, falls, mental and cognitive health, and the prevention and treatment of many chronic conditions [60]. They recommend that older adults accumulate 30 min of aerobic activity at a moderate level for at least 5 days/week [60], although more recent guidelines recommend 150 min/week [86, 90]. In contrast, sedentary behavior describes activities that are low in energy expenditure [≤1.5 metabolic equivalent of task (METs)]. These activities are done in a sitting or reclining position [76] and are associated with adverse health outcomes [85]. Accurately measuring time spent in both physical and sedentary activities is important for (1) investigating their dose–response influence on specific health outcomes, (2) informing appropriate interventions, and (3) testing the effectiveness of interventions aimed at increasing physical activity and/or decreasing sedentary time.

Physical activity and sedentary time are frequently assessed using self-report questionnaires, however, potential limitations include recall bias, socially desirable responses [74], and the influence of mood, depression, anxiety, cognition, and disability on responses [72]. This is especially important with questionnaires developed for younger adults but administered to older adults, as they may underestimate the physical activity level of older adults as they engage in different types of activities [92]. More recently, accelerometers are used to objectively describe activity patterns. They provide an objective measure that eliminates many of the challenges associated with self-report questionnaires and are appropriate for use in older adults [59].

The ActiGraph (Pensacola, FL, USA) accelerometer is a commonly used accelerometer for physical activity research. The monitor is usually worn at the waist. Early versions of the monitor used (cantilever beam) piezoelectric sensors to measure raw acceleration that is processed into activity counts with frequency filters. Since the introduction of the GT1M moniter, the sensor is a Micro Electro Mechanical Systems (MEMS) capacitative accelerometer [11]. Thresholds for the activity counts (cut-points) are determined from validation studies to classify activity intensity [11]. The monitors provide time and date-stamped information on activity intensity, categorized as sedentary (≤1.5 METs), light (1.6–2.9 METs) [64], or moderate-to-vigorous (≥3 METs) [23].

Accelerometry assumptions for the choice of cut-points and data analysis are not standardized across research protocols [52]. The majority of peer-reviewed literature informing accelerometry data analysis methods are from studies that included children and young adults [52, 53, 70, 88]. Literature on accelerometry in older adults is limited and many previous studies used validity studies completed on younger adults to determine cut-points for activity intensity. Taraldsen and colleagues [83] reviewed the use of accelerometers for physical activity monitoring in older adults and highlighted the wide variety of physical activity measures and called for the development of a consensus.

Therefore, our purpose was to: (1) undertake a systematic review of the literature to identify studies that used ActiGraph accelerometers to assess moderate-to-vigorous physical activity (MVPA) and sedentary behavior in older adults and (2) determine the effect of changing physical activity and sedentary cut-points on the results for older adults. The results of this study will provide an overview of literature objectively measuring physical activity and sedentary behavior patterns in older adults and will highlight current practice for accelerometry analysis for this age group.


In this two-part study, first, we completed a systematic search of the published peer-reviewed literature that used accelerometry to assess physical activity and sedentary behavior patterns of older adults and identified cut-points used to classify intensity of activities. Second, we used this information to analyze an accelerometry sample of older women’s activity patterns over 7 days to illustrate the effect of changing the different reported cut-points.

Phase 1: systematic review

Data sources and search strategy.

We completed an electronic search of the peer-reviewed literature for publications related to accelerometers, physical activity or sedentary behavior, and older adults. We reviewed published peer-reviewed literature from 1950 to July 4, 2012 from the following databases: AgeLine, CINAHL, EMBASE, OVID Medline, PubMed, and SPORTDiscus. We limited our search to adults aged 65 years and older using relevant Medical Subject Heading and keywords but included articles with older adults with a group mean age ≥60 years if they were found with our search strategy (Fig. 1). We completed the review in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-analyses guidelines [45].

Fig. 1
figure 1

Search strategy terms for accelerometers and physical activity or sedentary behavior in older adults

Study selection.

We included studies that used ActiGraph accelerometer (Pensacola, FL, USA) models CSA, MTI, 7164, GT1M, GT3X, and GT3X+ at the waist to objectively measure physical activity and sedentary behavior in free-living conditions. Specifically, we included studies with defined cut-points to determine activity intensity in community dwelling older adults (mean age of 60 years or older). We excluded studies that were targeted for special populations (e.g., older adults with stroke). Two of three reviewers (EG, HH, PY) independently reviewed each retrieved article based on title and abstract for relevance, and the additional reviewer resolved any discrepancies. Two of three reviewers (EG, HH, PY) then independently reviewed each full text article for inclusion and documented reasons for exclusion from the review. The final decision to include studies was decided by consensus with the third reviewer resolving any discrepancies. We did not rate the quality of the studies included in this review.

Data extraction.

We extracted the following information from the included studies: study population; accelerometry data collection methods and accelerometry analysis assumptions including valid day criteria; cut-points; and any other relevant assumptions reported. Two reviewers independently extracted data, and a third reviewer (EG, HH, PY) checked this information for accuracy.

Phase 2: accelerometry data analysis

Based on the cut-points reported in published literature identified and reviewed in Phase 1, we analyzed accelerometry data from active community-dwelling older women (65–75 years). Data were from a sample from a randomized controlled trial testing the effect of frequency of resistance training on cognition in older women [46]. This study was approved by the local university and hospital ethics review boards, and all eligible participants provided written informed consent. Participants wore the accelerometer (ActiGraph GT1M) at the waist during waking hours for 1 week in the winter. We set the accelerometers to collect data with 1-min epochs. To be included in the analysis, participants wore the accelerometer for at least 4 days and at least 10 h/day of valid wear time. We did not adjust for wear time. We analyzed the accelerometry data with the reported cut-points that we obtained from the systematic review. We used MeterPlus (Santech, Inc., La Jolla, CA) to analyze the accelerometer data files and excluded non-wear time (60 min or more of continuous zeros, based on the NHANES (National Health and Nutritional Examination Survey) criteria [87] and determined the time spent in sedentary, MVPA, and bouted MVPA (≥10 min of continuous MVPA with 1 to 2 min tolerance). For descriptive purposes, we included participants’ age, body mass index (BMI), 6-min walk test time (6MWT) [7] and Physical Activity Scale for the Elderly (PASE) self-reported physical activity score [93].

Statistical analysis

We used percent agreement to report consistency between reviewers for study inclusion in the systematic review. For data analysis, we summarized participants’ descriptive information with mean/standard deviation (SD) or median/interquartile range (IQR) (if data were skewed). We calculated the mean and SD, or median and IQR for the average daily time spent in sedentary and MVPA for each cut-point determined from the literature. For each cut-point, we calculated the percentage of participants who met 30 min of physical activity per day and the average percentage of the day spent in sedentary time. These values were compared with the most commonly used cut-point (1,952 counts/min for MVPA and 100 counts/min for sedentary time) following Bland-Altman’s method for assessing agreement between two measures [4]. To investigate differences between the most commonly used cut-point and other cut-points when used to measure group differences, we used logistic regression to calculate odds ratios for meeting 30 min of MVPA based on participant age and 6MWT result. For the 6MWT, we used a cut-point ≤490 m to indicate lower physical capacity [69]. For the Bland-Altman methods, we presented values of the mean differences, standard deviation (SD) of the differences, and the upper and lower limit of the Bland-Altman agreement levels (SD +/− 1.96). We used R (Foundation for Statistical Computing, Vienna, Austria) for the statistical analyses [18].


Phase 1

In this review, we identified 59 publications for inclusion (Fig. 2), and this represented 45 unique study data sets. Our two independent reviewers had 92 % agreement on the inclusion of articles in the study. The details of the 59 included publications, in particular, the activity monitor model, their accelerometry data collection, and analysis protocols are listed in Table 1. Within the 59 publications, there were: 34 papers that used the models 7164, 71256, MTI (7162); 23 that used the GT1M accelerometer; and only one study that used the GT3X monitor. A total of 53 publications included cut-points for MVPA and 36 for sedentary time. We located nine publications with the National Health and Nutritional Examination Survey (NHANES) 2003–2004 data and three that used the NHANES 2005–2006 data. The majority of studies (48 out of 59) included men and women, and the ActiGraph accelerometers were set at a 1-min epoch (40 of 59) for ≥7 days (52 of 59). Most publications (33 of 59) required participants to have at least four valid days with ≥10 wear hours/day to be included in their analysis.

Fig. 2
figure 2

Flow diagram of article inclusion for the literature search for accelerometers and physical activity and/or sedentary behavior in older adults

Table 1 Accelerometer-specific population, data collection, and analysis details for the 59 included articles from 45 unique data sources

We identified many cut-points for MVPA and sedentary time in the included studies. Eight different cut-points were identified for MVPA ranging from 574 to 3,250 counts/min. The most common cut-point used was 1,952 counts/min (19 of the 59 publications), and only 17 publications used a threshold lower than this. Of the 19 publications, 15 used the threshold of greater or equal to 1,952 counts/min, and four were greater than 1,952 counts/min. An additional four publications used 2,000 counts/min, as their analysis software was not precise enough to classify 1,952 counts/min. MVPA is usually classified as 3 METs or greater, but one study calculated it as greater than or equal to 4 METs and therefore used a larger cut-point of 3,250 counts/min. In addition, three studies included three cut-points for moderate activity to classify traditional MVPA and incorporate a lower value for moderate lifestyle activities. For sedentary time, five cut-points were identified, and they ranged from 50 to 500 counts/min with the most common cut-point being 100 counts/min. Twenty-one of the 59 publications used the cut-point of 100 counts/min, and of these, 19 used the threshold of less than 100, and two used less than or equal to 100 counts/min to classify sedentary time.

Phase 2

We analyzed the accelerometry data of 114 active community-dwelling older women using the cut-points identified from our search of the literature (eight MVPA cut-points and five sedentary time cut-points). The participants were healthy community-dwelling older women with an average age of 69.6 (2.9) years, average BMI of 26.6 (5.0), and achieved an average of 541.5 (75.03) m for the 6MWT (Table 2). Participants contributed a median 6 days of accelerometry data.

Table 2 Participant characteristics for the accelerometer data sample of older women (N = 114)


The eight cut-points yielded different amounts of MVPA ranging from a median of 4 to 80 min/day and between 0 and 24 min of MVPA accumulated in bouts of 10 min or greater/day (Table 3). Using Bland-Altman methods and the most common cut-point (≥1,952) as the reference, the observed differences ranged from −15 to 64 min/day. Between 4 and 95 % of participants met 30 min/day of MVPA depending on the cut-point (Table 3). Compared with the most common cut-point (≥1,952), the observed differences for accumulated bouts of MVPA ranged from −9 to 16 min/day (Table 3).

Table 3 Minutes per day of moderate to vigorous physical activity (MVPA) based on accelerometry, for the sample data of 114 older women analyzed with the different cut-points identified by a review of the available literature

Using logistic regression and age 65 to 68 years as the reference group, the odds ratios for not meeting 30 min/day of MVPA were highest for the 73 to 76 years participants and ranged from 4.32 to 5.88 depending on the cut-point used. For the 6MWT and >490 m as the reference group, the odds ratios for not meeting 30 min/day of MVPA ranged from 1.80 to 15.92 depending on the cut-point for the group who achieved less than 490 m (Table 4).

Table 4 Odds ratios of not meeting 30 min/day of MVPA using accelerometry and different cut-points, based on age and six-minute walk test (6MWT)

Sedentary behavior

For sedentary time, the amount of time determined from cut-points ranged from 475 to 665 min/day (Table 5). Using the reported cut-points, participants from the sample group averaged 62 % to 86 % of their day in sedentary behavior, and the most common cut-point (100 counts/min) resulted in an average of 68 % sedentary time per day. Using Bland-Altman methods and the most common cut-point (>100) as the reference, the observed differences ranged from −47 to 143 min/day (Table 5).

Table 5 Minutes per day of sedentary behavior based on accelerometry, for sample data of 114 older women with different cut-points based on a review of the available literature


In this systematic review, we identified 59 publications that used ActiGraph accelerometers at the waist to classify the activity patterns of older adults. Although the field is moving toward pattern recognition [67], accelerometry data analysis using cut-points remains the most common method used. We found eight cut-points used for classifying MVPA, ranging from 574 to 3,250 counts/min. We also identified five cut-points used for classifying sedentary time, ranging from 50 to 500 counts/min. This wide range of cut-points resulted in a correspondingly large range of minutes of MVPA (4 to 80 min/day) and sedentary time (62 % to 86 % of the day). We also noted that the most commonly used cut-point for MVPA (≥1,952) was able to best distinguish between participants who would likely meet the 30 min of MVPA/day based on age and results from the 6MWT. However, there were little differences between this value and cut-points within close proximity (≥1,566, ≥2,000, ≥2,020). We noted that some cut-points were distinctively different and could influence results by diminishing differences assumed to exist within groups. For example, by choosing a cut-point that is either too high or too low could either over or under estimate who meets the 30 min/day of MVPA depending on the group of older adults under investigation. Among the cut-points we analyzed, it appears that 1,566 to 2,020 counts/min (with 1,952 appearing to be the optimal) may provide the greatest differences across age groups and potentially physical capacity (6MWT). Therefore, different cut-points can substantially impact the classification of meeting recommended guidelines and the proportion of time spent in sedentary behaviors for a sample of healthy community-dwelling older adults.

Within our review of available evidence, we noted that more than half of the publications reported using the previous models [7164, 71256, and MTI (7162)] to acquire data. More recent evidence reports that the previous accelerometer models were more sensitive to movement. For example, in a 2013 article by Cain and colleagues [8], the authors noted significant differences between the older ActiGraph models and the more recent one (GT3X+). Specifically, the newer model (GT3X+) had significantly less daily step counts, more minutes of sedentary time and less light activity compared with the 7164. However, these differences were attenuated with the application of the low-frequency extension filter to the results from the newer models. Other recent studies support these findings [71, 91], and the understanding is that thresholds of the newer models were raised to overcome “noise” resulting from daily environmental vibrations. However, this higher threshold may be a limitation when measuring activity patterns of older adults who have slower gait speed and low activity patterns.

Although there is currently no consensus on the optimal cut-points for older adults, the majority of studies use the same cut-points for MVPA (1,952 counts/min) and sedentary time (100 counts/min), and this allows for comparison between the studies. However, these cut-points are not specific to older adults; the most commonly reported cut-point of 1,952 counts/min was validated in young adults [23]. Older adults may have a different capacity for activity, and their walking patterns may be altered [55] with increased energy expenditure [47, 51]. That is, for the same activity, older adults may expend more energy to complete a task compared with a fitter, younger adult. Therefore, a lower cut-point for MVPA than what is used in adult research may be appropriate due to the age related decline in fitness, if present. Ideally, individualized cut-points, which were used in one of the studies included in the review [68], would allow for the most accurate assessment of an individual’s activity level and reduce the risk of overestimating or underestimating physical activity. Individualized cut-points are not always feasible, and age specific cut-points may be an appropriate compromise for older adults, but the results will also depend on the physical capacity of the participants. For example, Copeland and colleagues’ [14] cut-point of 1,041 counts/min was developed specifically for older adults, and using this cut-point resulted in 64 % of the older women in our dataset meeting physical activity guidelines compared with 33 % meeting guidelines using the most commonly reported cut-point of 1,952 counts/min. Other investigations showed that different cut-point values result in statistically significant changes in the amount of MVPA [20] while Miller and colleagues [57] investigated the impact of age on the validity of ActiGraph accelerometers using a lab-based treadmill protocol. They found that there was no statistically significant difference across age groups for the absolute physical activity intensity. However, there was a significant difference in the relative physical activity intensity due to individual differences in cardiorespiratory and muscular fitness as a result of the aging process or physical inactivity.

The results from our analyses highlight that approximately one third of study participants were active for ≥30 min/day. This may reflect the fact that the study participants were active, community-dwelling older women and not representative of all older adults. In our review of the literature, we noted a range of physical activity results, including: a number of other studies that also reported high levels of MVPA for their study participants [30, 38] and population-based studies reporting very low values for MVPA [31]. Thus, the selection of cut-points may depend on target group and the purpose of the investigation. Using the same cut-points across all age groups may be appropriate for large epidemiological studies (that are interested in the absolute physical activity intensity), but age-specific cut-points may be necessary for physical activity prescription or when investigating the dose–response and effectiveness of interventions (where relative physical activity intensity is of interest). In addition, many recent studies have divided the MVPA category into two categories representing moderate lifestyle activity and moderate intensity physical activity. This allows for the comparison to studies using other cut-points.

We found that, for sedentary time, the range of cut-points resulted in a difference of 25 %, or over 3 h/day, which is important as older adults could spend a large proportion of their day in sedentary activities. It is promising that the majority of studies are using the same cut-points for sedentary time. However, to our knowledge, there are no studies that validated sedentary cut-points for older adults. This is an emerging area of research, and older adult specific validation is needed.

Another issue that can affect the accuracy of reported sedentary time is the ability to differentiate between non-wear time and sedentary time [94]. This is of particular concern for older adults’ accelerometry data because the large amount of time they spend in sedentary behaviors can potentially lead to the misclassification of sedentary time as non-wear time [94]. Of the included publications, 50 reported some assumptions for their data cleaning procedure to identify spurious data or non-wear time (Table 1). The most common criteria for non-wear time (24 of 50) was based on the NHANES recommended protocol for the removal of 60 min or more of continuous zeros with allowance of 1–2 min with counts between 1 and 100 [87]. More recent literature suggests 90 min may be more appropriate [12] for some older populations with limited mobility.

We note several limitations to this review. In Phase 1, we only included articles published in English and also note that with the scope of our search strategy, we may not have retrieved all studies that used accelerometry in older adults when it was not the primary outcome, or they used a different activity monitor. In addition, the included articles did not always state all of their accelerometry assumptions in their manuscript, thus making it difficult to fully assess their accelerometer protocols. The risk of bias for the included articles was not assessed as the scope of the review was specifically interested in only accelerometer related data collection and analysis, and this may not have been the primary outcome of the study. For Phase 2 of this study, we recognize that our participants were healthy, active, community-dwelling older women and did not represent all older adults, although some of the studies reviewed also included active older adults [30, 38]. Overall, the study participants were younger, and the results may not apply to older adults aged 80 years+. Due to the design of the RCT from which we sampled data, we did not include data from men in our analyses. Although we anticipate that our observed differences in outcomes based on cut-points would be similar for men, this is an area for future evaluation. Therefore, we are unable to show the implications of different cut-points for older adults of different activity or ability levels. In addition, for our analysis, we did not adjust for wear time, and this may influence the outcome, as participants who wore it longer may appear to accumulate more sedentary time. We also used 60 min of continuous zeros to determine “non-wear time.” Recent literature suggests that 90 min may be more appropriate for some older adults [12], and by limiting the sedentary time to 60 min at a given time, we may be underestimating sitting time. These are both important points and should be considered during the analysis process for the current way of data analysis. Finally, we acknowledge that we did not have a criterion measure of participants’ activity level to compare the results with and ultimately determine the optimal cut-points. However, we used the results from the 6MWT to calculate odds of not meeting 30 min/day of MVPA.

Future research in this area is promising as there are increasingly more studies being published using ActiGraph accelerometers in older adults. All of the publications identified in this review were published since 2004. Despite this growth in research utilizing accelerometers, many researchers are using a variety of different cut-points. Accelerometry is not only being used in healthy community dwelling older adults, as in the current study, but also in a variety of older specific populations (e.g., adults, older adults after stroke, or in hospital), and appropriate assumptions for these groups may need to be condition-specific. Furthermore, technological advances in device hardware and analysis software continue to evolve. With respect to analysis of data, this includes using artificial intelligence to identify activity types [67, 79]. Many of these advances are still under development, thus current practice remains in favor of using the manufacturer-provided software that allows for cut-point analysis. The monitors themselves continue to increase in memory capacity and battery life to allow for the collection of data in shorter epochs or in raw form for longer time periods as well as the addition of multiple axes.


In summary, our review highlights that there is not a standardized method to quantifying accelerometry-based physical activity and sedentary time in older adults. The assumptions used in data analysis of accelerometer data can produce markedly different results, and using too low or too high cut-points may obscure important group or treatment differences. For future studies, standard reporting should include specific data assumptions for analysis. Further research is needed to determine which assumptions are most appropriate for older adults, taking into account their physical capacity.