Device-measured sedentary behavior and physical activity in older adults differ by demographic and health-related factors

Physical activity (PA) is essential for healthy aging. Older adults who are physically active have lower rates of all-cause mortality and many chronic diseases, a lower risk of falling, better physical function and better cognitive function [1]. The robust evidence on the health benefits of PA has led to public health PA guidelines recommending that adults ≥65 years should accumulate at least 150 min/week of moderate-to-vigorous intensity PA (MVPA), i.e. ≥3 metabolic equivalents (METs) [1], with brisk walking often used as a primary example of an appropriate activity. In addition, muscle strengthening activities should be performed two or more days per week. Yet, adherence to PA recommendations is low in older adults, with studies relying on self-reported data indicating that 55 to 70% do not achieve the minimum level of MVPA [2, 3].

Additionally, older people may spend up to 80% of their waking day sedentary [4‐9] and the sedentary time seems to increase with age. Growing scientific evidence has recognized that sedentary behavior is a distinct risk factor for poor health, including increased risk of mortality, type 2 diabetes, metabolic syndrome, cardiovascular disease and cancer [10‐12]. Sedentary behavior is measured in varied ways but consistently defined as any waking behavior characterized by an energy expenditure ≤1.5METs, while in a sitting, reclining or lying posture [13]. Research suggests that not only the total time spent sedentary is of importance, but also the pattern of how the sedentary time is accumulated, i.e., prolonged vs shorter sedentary bouts or how frequently sedentary time is interrupted. Interruptions to sedentary time (breaks) play an important role for cardio-metabolic health [14, 15], and may also be associated with muscle mass and physical function in older adults [16‐18].

It is known that self-reported data have low accuracy due to recall bias or social desirability. Sedentary time is often underestimated and MVPA is often overestimated, while light-intensity PA (LPA), which is common among older adults, is almost impossible to recall or report accurately [19‐21]. It is therefore highly relevant to use device-based methods, which more accurately can capture the whole PA intensity spectrum, as well as other aspects of the daily movement pattern, for example sedentary time in bouts. Body-worn activity monitors, such as accelerometers, are shown to be sensitive and feasible for measuring PA and sedentary behavior in older adults, although it has been sparsely used in the oldest age groups. Studies that have used accelerometry to assess PA and sedentary behavior in individuals over 70 years have found that patterns differ according to gender, age group, education, and body mass index (BMI) [4, 6‐9, 22].

To date, most evidence on associations between device-based sedentary behavior and health has derived from accelerometers defining sedentary behavior as lack of movement, in contrast to measures derived from actual sitting posture. This can lead to misclassification of low-intensity non-sedentary behavior, such as standing [23]. The thigh-mounted activPAL3 is an accelerometer that directly measures the postural aspect of sedentary behavior, i.e. sitting/lying vs upright position, and thereby also more accurately can capture breaks in sedentary time by registering sit-to-stand transitions [23]. Richer knowledge of older adults’ pattern of sedentary time accumulation and PA in both light and moderate-to-vigorous intensity can guide development of health promoting efforts tailored for older people. Therefore, our aims were to describe and explore older adults’ device-measured sedentary behavior and PA pattern by sex, age, education, marital status, BMI and physical function; and to compare agreement regarding fulfillment of PA recommendations between device-measured and self-reported PA.

In this study aiming to describe and explore older adults’ device-measured pattern of sedentary time and PA by subgroups, we found that women spent about 30 min less sedentary and equivalent more time activities of light intensity, such as household chores or slow walking. These sex differences are in line with similar populations in Europe [4, 8, 9] and may be of importance for health. Previous research has shown that substituting 30 min of sedentary time to LPA may reduce the risk of all-cause mortality with 11% and risk of cardiovascular disease with 24% [37]. Interestingly, there were only small sex differences in pattern of sedentary time accumulation. Total sedentary time was higher with increasing age. Pattern of sedentary time accumulation varied by age group, but the differences in number of sit-to-stand transitions and usual sedentary bout duration were quite small, with no differences in longest sedentary bout. Sedentary behavior did not vary by education or marital status except for total sedentary time, which was 20 min/day more in those who were living alone. Obese participants spent 40 min/day more sedentary than those with normal weight on average and they also had a less favorable sedentary accumulation pattern. These differences were also seen for individuals with impaired physical function.

We did not find statistically significant sex differences in time spent in MVPA or number of daily steps. This is in contrast to many other studies, both those using self-reports and accelerometers, which found males to be significantly more active than females [38]. It is known that many PA questionnaires give examples of MVPA activities that traditionally are mostly performed by men, especially in older generations, and therefore are likely to underestimate MVPA in older women. However, recent device-based studies have also found that MVPA is higher in men [4, 8, 39, 40]. In our study, on the contrary, we found that women were more likely to fulfil the PA recommendations of 150 min/week of MVPA than men in the adjusted models. Similar findings have previously been reported in two Norwegian studies [9, 41], and the discrepancies might depend on cultural and sociodemographic differences between older adults in the Nordic countries and other study populations.

As reported in previous research [4, 8], overweight and obese participants spent significantly less time in MVPA and had lower fulfillment of MVPA recommendations, compared to normal weight participants. Surprisingly, our underweight participants had a high level of MVPA. Underweight in older adults is often related to frailty and a lower muscle mass, which would be expected to result in lower MVPA. However, it is likely that our finding is a consequence of selection bias, and that our small sample of underweight participants, 24 woman and 1 man, consists of well-functioning, active older adults. Education level is another known correlate of PA [42], and this is consistent with what we found for MVPA, but not for sedentary behavior or LPA.

In our study sample, 26% of the participants were not able to do five chair raises without help from their arms. These older adults were significantly less active than those who could perform the test, with over 9 h/day sedentary and only 15 min/day in MVPA, which is clearly under the recommended PA level. Importantly, we also found that they had a lower number of sit-to-stand transitions and longer usual sedentary bout, suggesting that difficulty in raising from a chair may lead to a more sedentary behavior. Even though studies have found that breaks in sedentary time are associated with better health outcomes [16, 18], evidence is still inconclusive and more research, including prospective studies, is needed to gain a deeper understanding of these associations. Considering the negative health effects of prolonged sedentary behavior and a low PA level, it should be highlighted that this is a group of older adults that may benefit from interventions aiming to decrease sedentary time and increase MVPA. For example, a replacement of just 10 min of sedentary time with MVPA could result in a 34% reduction of cardiovascular mortality risk [37].

In agreement with other studies [8, 43], we found a wide variation in time in MVPA and steps/day within all age groups, indicating that high levels of PA may be sustained even among the oldest old. Two participants in the ≥90 group had a daily average of > 10,000 steps. However, when considering the data from study participants in this oldest age group, it is important to acknowledge that the results reflect those who have successfully aged and are able and willing to participate, and thereby represent a selective sample. Still, there is limited data on PA and sedentary behavior in the oldest old, and as the number of nonagenarians and centenarians are increasing worldwide, our results may contribute with valuable information about this group.

Walking is the most popular and most commonly reported form of physical activity in older adults, and to provide generalized guidance of the desired daily activity to meet the recommendations of 150 min/week of MVPA, a level of approximately 7000 steps/day has been suggested by the American college of Sports Medicine [33]. However, recognizing that the normative range for number of steps/day for older adults is 2000–9000 [32] and that many older adults may be limited in their everyday activities due to mobility limitations and chronic illness, we chose also to compare with 5000 steps/day as the threshold.

The age-related prevalence of meeting the PA recommendations of 150 min/week of MVPA ranged from 75% in the 66 years group to 23% in the ≥90 group, with very similar results if the 7000 steps/day definition was used. The few previous studies that have assessed device-measured PA in older adults divided by age groups have also shown that PA decreased progressively with age for both men and women [9, 38]. If the 5000 steps/day level was used, 95% in the 66 years group met the goal and 61% in the ≥90 group. This suggests that many of the oldest old, may find the recommendations challenging, and highlights the importance of a broad perspective when counselling older sedentary individuals [44]. A focus on the message to reduce sedentary time and increase light activities may be more realistic and pave the way to physical activity with higher intensity. The dose-response relation between physical activity and health clearly shows that there are benefits from any level of PA, and more is better [10].

The kappa agreement between self-reported and device-based MVPA was fair overall [36], which is in line with other studies. Previous validations of one or two PA questions using accelerometry assessed MVPA have found correlations of rho = 0.31 [45] and rho = 0.38 [46], and K = 0.18 [46]. Even more comprehensive PA questionnaires, and thereby possibly more accurate, have limited correlations in older populations ranging from rho = 0.11–0.65 [19].

Validation studies of self-report measures of sedentary time (usually ‘sitting’) have shown it is difficult for older people to accurately recall their behavior [19, 20]. Wearing devices can give people an accurate estimate of their behavior and therefore shape how they choose to respond to messages to sit less (as in the physical activity guidelines of many countries) and to ascertain the effectiveness of any changes to reduce their sedentary behavior. The same is true for PA. If we can develop better methods of self-report or cheaper devices that accurately measure these behaviors, then individuals can choose how to respond to that information. Our results add to the current knowledge that older adults sit for long periods and have low levels of PA, which are both suitable targets for change.

This study has several strengths. One important strength is the use of the activPAL3 accelerometer, which is a posture-based activity monitor. By using this device, we could investigate pattern of sedentary time accumulation in a more reliable way than studies using hip or wrist worn accelerometers, which estimate sedentary time based on limited acceleration rather than posture. The activPAL3 has also shown good accuracy in measuring steps in older adults who walk slowly and use an assistive device [47]. It might be worth noticing that the sedentary time derived from the activPAL3 can include reclining or lying during waking hours; however, these positions are included in the definition of sedentary time [13]. Another important strength is the population-based study sample, including 133 participants over 84 years. Further, a sample size of at least 50 participants has been suggested for validation studies on PA questionnaires [48]. The sample size in our study was over 650 participants, which made it possible to draw stronger conclusions, and to perform stratified analyses for subgroups. We used Cohen’s kappa for reporting results on the validity of the PA questions which is a considered a better choice of method for categorical measures than the often-used Spearman’s rho [19, 35].

A limitation of our study is that the design did not allow us to include participants with cognitive impairment or mobility limitations, leading to a study sample healthier than the general older population. It is also a limitation that the cadence of ≥100 steps/min used to define MVPA has not been validated for older adults, and some misclassification may have occurred. Ideally PA intensity cut points should be individually calibrated; nevertheless, this is not feasible in large study samples, a limitation shared with similar research. As in many other accelerometer studies, activities such as cycling, strength training, water-based activities, upper body movements and the effort of carrying loads were not captured. Additionally, even though we made a large effort to estimate wear time correctly, it must be acknowledged that the longest sedentary bout variable is sensitive to miscalculation of wear time. The overall wear time was high but is still likely to have missed some of all activities, since a 24-h wear protocol was not used, and participants did not always attach the devices immediately upon rising or immediately before bed. Further, our self-reported data provided information regarding the intensity and frequency of PA, but not the duration of the activity. Still, it is hard to accurately recall how many minutes that are spent in different activities in a day or week, especially unstructured activity interspersed over the day. Regular activities at a moderate intensity with longer durations are easier to recall and it is reasonable to assume that people reporting being active every day or several times per week also achieve the recommended PA level. The device-measured PA was collected over the entire year for different individuals and would thereby not be affected by season on a group level. Still, it is a limitation to the study that device-based PA was assessed over one week, while self-reported PA was asked about for the past 12 months. It should also be noted that, even though it is included in the PA recommendations [1], we did not investigate muscle strengthening physical activity. Finally, due to the cross-sectional design of this study we cannot draw any causal conclusions from the results.

We found significant sex differences in daily sedentary behavior and LPA in older adults, but not in MVPA, in contrast to previous findings in self-reports. Women spent on average about 30 min/day less sedentary and equivalent more time performing activities of light intensity than men. Sedentary time increased with older age, with small differences in pattern of sedentary time accumulation. Individuals with limitations in physical function spent over 9 h/day sedentary and only 15 min/day in MVPA. A majority of the participants over 80 years did not meet the PA recommendations of 150 min/week of MVPA. Age differences were much more pronounced with objective measures than by self-report. The limited agreement between device-measured and self-reported fulfillment of PA recommendations illustrate the consequences of using different methodological approaches in PA surveillance. Device-based measurement adds value to studies, providing richer and different data than self-report, such as more detailed amounts and sedentary behavior. Given the strong relationships between sedentary behavior, PA and health in older adults, programs are needed to address these behaviors. A better understanding of sex and age differences in sedentary behavior and PA in older adults can inform interventions and guide public health messages.