This study aimed to determine the minimum number of days of activPAL3 monitoring required to reliably examine SB and ATMB in NH residents and the best time blocks in which to process the data. Our findings indicate that a minimum of 3 consecutive days wearing the activPAL3 device is required to achieve high reliability for those residents with capacity to stand and walk, and 6 consecutive days for those residents unable to stand and walk across a range of SB and ATMB variables. We also found that processing data from midnight to midnight was more reliable than processing from midday to midday.

The minimum number of days required to adequately measure SB and ATMB in NH residents may be dependent on a number of factors, including the monitor used and outcome measures explored, and the actual levels and variability between individuals in daily patterns of SB and ATMB.

Edwarson et al. noted that a person-oriented approach to analysis of SB and ATMB data can be behaviourally relevant, but that in community-dwelling adults and older adults person-oriented day durations (from one wake time to the next) are not always 24 h long [45]. In other words, people may wake up at different times each day, which may be accentuated in working age adults by differences between week (working) and weekend (non-work) days. Farias-Aguilar et al. found in a community-dwelling working-age adult population, differences in intra-individual variability between weekdays and weekend days, in that there was slightly less variability in SB and activity behaviours on weekends compared to weekdays [56]. However, person-oriented day duration and intra-individual and inter-individual variability between days of the week for NH residents are likely to be different from that of the working adult population. This is because NH residents spend most of their day lying or sitting during their daily activities, particularly related to mobility and feeding [37]. In most NHs the daily routines are set by the institution management and there is a high level of control by staff, shaping residents' daily routines [37, 57]. In line with our findings, Buckley et al. and Airlie et al. found that there was no difference in most outcome measures between weekdays and weekend days in NH residents [58, 59]. Therefore, for this population, it may not be necessary to include a weekend day in the assessment period, and older NH residents are not bound to a typical weekday/weekend week structure. Considering that the daily routines of NH residents are strongly conditioned, we suggest that their person-oriented day durations are likely to approximate to 24 h long (similar wake times each day) without significant differences between weekdays and weekends.

Most recent studies using the activPAL3 device report the choice of a 24-h wear protocol, meaning the monitor is worn continuously including overnight, but there is a lack of consensus on many other protocol decisions such as timing of starting to wear the monitor, what time blocks, and reporting on data processing decisions such as when data is processed [33, 46, 60]. This lack of consistency can potentially lead to discrepancies in data interpretation and make comparisons between studies difficult. For example, Reid et al. measured activity patterns among older adults in residential aged care using the activPAL3, they reported using a 24-h protocol for 7 days, but do not report the starting hour of the monitoring period or if the data of the first and last days were composed from partial days or were whole days [33]. Bootsman et al. also reported, using a 24-h wear protocol, with the activPAL3 in older adults living in residential aged care facilities for five days. They reported that the participant started wearing the monitor during the day, but that measurement only started at midnight, which was to minimize potential differences in movement behaviours during the first few hours of wear [60]. In a study measuring SB of community-dwelling older adults with the activPAL3, Dall et al. reported programming the monitor to start recording immediately with a recording duration of 14 days and that the monitor was then put on the participant at an unspecified later date and time [46]. The protocol specified the devices were taken off from the ninth day of wear onwards at an unspecified time. Data was then processed into midnight-to-midnight time blocks to extract 7 days of data each 24-h long and starting at the same time. Our study supports their choice of midnight time blocks for analysis, as they offer greater sensitivity and showed no statistically significant differences between days. This was perhaps surprising, as the analysis covers the same activity of each individual, and the only difference between time-blocks is that the midnight time block more usually represents a single person-oriented day (e.g. going to sleep before midnight on the day that you woke up on), than a composite from two person-oriented days (midday to sleep time on one day and the wake time to midday of the following day). It is unclear whether analysing a true person-oriented approach (wake time to wake time) would also represent a reliable method, or how these results would translate to a community-dwelling population, where more diversity of the timing of wake and sleep would be likely.

To the best of the authors' knowledge, the study by Reid et al. is the only one that examines the minimum requirements for obtaining reliable estimates in older adults residing in care homes in Australia using the activPAL3 device for measurement [33]. Reid et al. explored three standard outcome measures, and found it would take 5 to 11 days to estimate sitting time, 5 to 10 days for standing time, and 7 to 15 days for stepping time to achieve an ICC of 0.8 to 0.9. In contrast, for those three outcome measures, our study required a minimum of 2 to 4 days. Even across the full range of outcome measures we explored, the minimum days required were 3 for residents with the capacity to stand and walk and 6 for those who were unable to stand and walk. These are considerably shorter than the minimum days required by Reid et al. [33]. Participants in the two studies appear to be reasonably similar, in terms of mean age, 84.2 and 85.8, although some of the potentially more frail NH residents (individuals with pacemakers, behavioural issues, uncommunicable deafness or diagnosed severe dementia) were excluded from Reid’s study, but included in ours. Therefore, the main reason for differences in minimum number of days required between the two studies might be due to different sample sizes (n = 31, in Reid et al. vs n = 95 in the current study). Indeed, when we calculated the reliability stratified into two walking capacity groups, compared with the entire sample, the reduction in sample size resulted in a loss of heterogeneity, which affected the reliability of the results. Consequently, more days were needed to ensure an ICC of 0.8 or 0.9, to reach more accurate and reliable measure of the variable being studied.

Buckley et al., in 2020, explored the minimum reliable days of device wearing for walking activity in a sample of 257 NH residents in New Zealand, measured using a different device (Axivity AX3 device, worn on the lower back) [58]. Data was recorded across 8 days, and divided into 7 days for analysis using the half days on the first and eighth day of measurement. Although the start time of the days was not reported, this is functionally similar to our midday time block. Buckley et al. focused on walking and assessed volume variables of total walk time, total steps and total number of walking bouts and pattern variables, including mean walking bout length. Results were presented for the whole group, number of days required ranged from 2 to 5 days across all variables, and stratified by level of care, the dementia level of care ranged from 1 to 3 days, the intermediate ranged from 2 to 7 days, and the high level of care ranged from 2 to 6 days. Although the device used and the variables explored were different, our results are in line with their range of minimum days. Also, in line with our findings, the number of days of measurement required for volume-based metrics was lower than those for pattern-based metrics. Buckley et al., also classified their sample in groups according to their level of care, whereas we grouped participants according to their physical capacity to stand and walk. The number of days required for measurement for those who were able to stand and walk in our study [2, 3] was similar to the dementia care group [1,2,3], whereas the days required for the group who could not mobilise in our study [2,3,4,5,6] was similar to both the intermediate care group [2,3,4,5,6,7] and the high care group [2,3,4,5,6]. However, when it comes to issues of dementia, level of care does not necessarily equate to ability to mobilise. Indeed, individuals in the high care group showed better cognition status according to the Montreal Cognitive Assessment (MoCA) than both the intermediate and the dementia level of care groups, but had worse physical function assessed using the Time Up and Go (TUG). Also, as a group, those with the dementia level of care had the best physical function. In comparison, our results showed that the more disabled group had both a higher physical and cognitive impairment (68% with severe cognitive impairment) compared with the able to stand and walk group (23% with severe cognitive impairment). This suggests that it is the influence of physical performance capacity, rather than the cognitive status, that determined how many days of measurement are required, potentially because all of the variables assessed in both studies are related to physical performance. Another possible reason why a more physically impaired group might need more days to guarantee the reliability of data is the high homogeneity of their daily routines and their sensitivity to PA. These residents spend almost all of their waking hours in SB and their PA bouts are typically limited to the same daily routine (e.g., toilet, hygiene, or feeding) and depend on the availability of NH staff and their assistance. Therefore, any PA bout outside of their daily routine would become unusual (e.g., if a resident requires assistance due to an incontinence event). This isolated PA bout would be enough to make a large difference in both the volume and pattern of PA between days, thus reducing the reliability of the data. Consequently, more days of assessment may be needed to reach higher reliability values.

In a similar manner, Airlie et al. (2022) determined the minimum number of days of wear and optimal wear time criteria required to assess PA and SB, measured using a different device (ActiGraph wGT3X + worn on the right hip), in a sample of 91 NH residents in the United Kingdom with preserved mental capacity [59]. Data were recorded over the course of 7 days. Although the starting time of recording was not reported, it was mentioned that the first monitoring day was excluded if the monitor was administered after 1 PM. This implies that they did take into consideration the starting time of recording with the device and excluded data from the first monitoring day if it began after 1 PM. Moreover, they used data from half-days, but they did not employ data from half-days with less than 4 h of wear time, similar to our midday time block. However, they did not follow a 24-h protocol, as the NH residents were instructed to remove the device before engaging in any water-based activities. Instead, they utilized a diary log for wear time, where residents were required to report the day, time, and reason if they removed the device. Furthermore, the procedure for completing the activity log was clarified to the staff, who were also requested to provide assistance where necessary. However, the method by which they obtained awake time and excluded night time data is not reported.

The results from Airlie et al. indicated that estimates of accelerometer outcomes as counts per day, counts per minute, PA time in minutes, and SB time in minutes, did not significantly differ by monitoring day (weekdays or weekend) like our results and Buckley’s., and the accelerometer outcomes were equivalent regardless of the employed minimum daily wear time criterion. This suggests that accelerometer outcomes are consistent and reliable in NH population. Additionally, the study examined the impact of the number of monitoring days on the reliability of accelerometer outcomes. Results showed that estimates of counts per minute were equivalent regardless of the number of monitoring days. However, for counts per day and PA time, only estimates based on at least 6 monitoring days were considered equivalent to estimates based on 7 monitoring days. These findings suggest that a 7-day monitoring protocol remains advisable to ensure reliable estimates of PA and SB.

The requirement for a 7-day monitoring protocol can be attributed to the fact that nearly half of the sample consisted of dependent individuals, and the analysis was conducted by considering both dependent and independent residents together. When comparing these results to our findings, the initial analysis of minimum days of reliability did not involve stratifying the sample, leading to the reporting of more than six days required in one variable. However, after stratifying the sample based on their capacity to stand and walk, the more impaired group indicated a greater need for assessment days across multiple variables when compared to those with the ability to stand and walk. These findings further support the previously suggested hypothesis that the influence of physical performance capacity, rather than cognitive status, determines the necessary number of measurement days.

The minimum number of days wearing the activPAL3 has been explored in a few other populations, including asymptomatic female adolescents, working adults, middle age women and adults and older adults receiving hemodialysis. Female adolescents (n = 195, with a mean age of 15.7 (SD = 0.9) years old), assessed using a 7-day monitoring period, required a minimum of 12 days to achieve a reliability of ≥ 0.8 for the variables time spent sitting or lying, standing time, light PA (LPA), and moderate-to-vigorous PA (MVPA), while 21 days were necessary for assessing the number of steps [61]. Working adults (n = 90, with a mean age of 39.1 (SD = 12.43) years old), assessed using a 7-day monitoring period, required a minimum of 5 days (with at least 1 weekend day included) to achieve a reliability of 0.8 for the variables sitting or lying time, standing time, and stepping time, while transitions to standing required at least 3 days [56]. Middle age women (n = 68, with a mean age of 52 (SD = 8) years old), assessed during a 7-day monitoring period, required a minimum of 4 days an ICC of 0.80 for the variables sitting or lying time and LPA and 9 days were needed for an ICC > 0.9 [62]. For adults and older adults receiving hemodialysis (n = 70, with a mean age of 55.9 (SD = 15.7) years old), assessed during a 7-day monitoring period, required a minimum of one dialysis day and two non-dialysis days for an ICC of 0.80 for the variables of waking hours, percentage of time spent sitting or lying, percentage of waking time spent standing, the number of transitions to standing per hour, number of steps taken per day, number of steps taken per minute and energy expenditure per minute [63]. In general, as the participants in these studies get older, the number of days required to wear the activPAL gets lower. However, it is likely that it is how age impacts daily routine, and thus intra- and inter-individual variability, that may be important [56, 63]. This is supported by the study of individuals receiving haemodialysis, where it was suggested that days with and without dialysis, which likely had very different patterns of activity were included in the measurement period. Prescott et al.'s also suggested that comorbidities, and lower levels of functional independence, can lead to lower inter- and intra-individual variability [63]. Our findings suggest that variability is affected mainly by the high level of schedule control exercised by NH staff in residents' daily routines and also by the participants' ability to stand and walk. Moreover, in those who are unable to stand and walk, the variability would also be affected by their dependency to the NH staff and their assistance for their daily routines [37, 57].

Our study has several limitations. One limitation of this study is sample size as data collection was stopped in March 2020 due to the covid-19 outbreak. However, the sample size of 95 in this study is larger than, or similar to, other studies exploring minimum number of days of wear [33, 56, 62, 63]. The final sample analysed was 51% of those invited to take part in the study, which may not be representative of all NH residents. Also, the sample is specific only to NHs in Catalonia (Spain), which have their own politics, characteristics and context, and may not be generalisable to NHs elsewhere. The study was based on 7 days of activPAL3 measurement, so we can only report on up until < 6 days of wear. We found a high reliability (ICC > 0.8) for all variables explored within 7 days, but a longer period of assessment would be required to explore the number of days required to attain a very high reliability (ICC > 0.9) for some of the variables, in particular those exploring the pattern of SB. The limited availability of devices in the project meant that assessment was started for residents on different days of the week. Another potential limitation was the inability to use a diary log to document waking times, bedtimes, and napping within the NH population. During the pilot study, the team attempted to instruct the residents on how to complete the diary correctly. However, at the end of the assessment, after seven days of wearing the activPAL device, none of the residents returned their completed diaries to the research team. Subsequently, we sought input from NH staff, but their responses, though provided, were general and inaccurate for all residents (e.g. everyone getting up at 9:00 AM and going to bed at 10:00 PM). Due to the lack of compliance among NH residents, we decided to discontinue the use of diaries and instead focused on analyzing heat maps within the activPAL software to determine waking and bedtime patterns. Because of this decision and the absence of contextual information from the objective measurements provided by the activPAL, we were unable to identify any instances of napping or sleeping during waking hours for NH residents, especially those who were bedridden. In conclusion, we included all data from waking times to bedtimes, acknowledging that in certain cases, the sedentary behavior data might be somewhat inflated. Finally, we excluded variables related to the number of steps taken, due to the risk of the activPAL3 not recording steps in residents with a very low gait speed (< 1.5 km/h) [64]. On the other hand, our study has several strengths. Firstly, the results can help improve compliance with wearable devices among the NH population by reducing the required wearing time, thereby avoiding loss of both the device and data [33, 58]. To our knowledge, this is the first study to assess data eligibility based on time block distribution using the activPAL3 device, and to stratify the sample by residents' capacity to stand and walk in NHs. Additionally, our study provides information on the minimum number of days required for each variable individually, allowing researchers to choose and select variables according to their specific needs. Finally, our study offers pragmatic solutions for researchers working with the gold standard activPAL3 device and those seeking to evaluate interventions aimed at reducing prolonged sedentary bouts and promoting PA among NH residents.