Feasibility, Reliability, and Validity of the MotionWatch 8 to Evaluate Physical Activity Among Older Adults With and Without Cognitive Impairment in Assisted Living Settings

Abstract

This study tested the feasibility, reliability, and validity of the MotionWatch 8 among assisted living residents with and without cognitive impairment. Data from the Dissemination and Implementation of Function Focused Care in Assisted Living Using the Evidence Integration Triangle study were used. The sample included 781 individuals from 85 facilities with a mean age of 89.48 (SD = 7.43) years. The majority were female (71%), White (97%), and overall (44%) had cognitive impairment. A total of 70% were willing to wear the MotionWatch 8. Reliability was supported as there was no difference in time spent in activity across three consecutive wear days. Validity was based on hypothesis testing, and function was associated with counts of activity at baseline (p = .001) and 4 months (p = .001). Those with cognitive impairment engaged in less physical activity (p = .04). The MotionWatch 8 is a useful option for measuring physical activity in older adults with and without cognitive impairment.

Estimates derived from national data indicate that seven out of 10 residents in assisted living have some form of dementia, with 29% having mild impairment, 23% with moderate impairment, and 19% with severe impairment (Hawes & Phillips, 2000; Zimmerman, Sloane, & Reed, 2014). The majority of assisted living residents also require help with activities of daily living (ADLs) and instrumental activities of daily living (IADLs) (Assisted Living Federation of America, 2014; Resnick, 2011), are sedentary, and have limited opportunities to engage in physical activity (Resnick, 2011; Davies, Ellis, & Laker, 2000; Galik, Resnick, Hammersla, Brightwater, & Chloe, 2015). Individuals with dementia, including mild, moderate, or severe dementia, have repeatedly been noted to engage in less physical activity than those without dementia (Falck, Landry, Best, et al., 2017; Hartman, Karssemeijer, Van Diepen, Olde Rikkert, & Thijssen, 2018). Specifically, older adults with dementia wearing a wrist-based accelerometer were noted to spend more time in sedentary behavior and less time in light to moderate or moderate to vigorous intensity physical activity when compared with those without dementia (Hartman et al., 2018). This was true even among those with mild cognitive impairment (Falck, Landry, Best, et al., 2017).

Current recommendations (American College of Sports Medicine and the American Heart Association, 2013) are for older adults to engage in at least 150 min per week of moderate-level physical activity (i.e., ≥3 metabolic equivalents or including activities such as walking up a flight of stairs or walking at 100 steps per minute). Given the combined cognitive and functional impairments of these residents, innovative approaches are needed to help them optimize function and physical activity and remain in assisted living settings. To evaluate the efficacy of these interventions, it is important to reliably assess physical activity by utilizing approaches such as accelerometry (Kochersberger, McConnell, Kuchibhatla & Pieper, 1996).

Known Advantages and Challenges to Use of Different Types of Accelerometry With Older Adults With Dementia

There are many advantages to measuring physical activity using accelerometry, particularly with older adults with dementia who cannot respond to surveys about physical activity. Advantages of accelerometry include the ability to objectively capture minute by minute physical movement and activity, including very low-level activity or higher intensity levels. Conversely, there are many known challenges, such as the lack of contextualization of the activity (e.g., what is being done), the possibility of missing data due to device malfunction, or the device not picking up activity. Examples of potentially missed data can occur when the individual is walking but there is no wrist movement if the device is worn on the wrist. Additional challenges include having to manage large amounts of data if multiple days are collected and the lack of consensus as to what constitutes sedentary, moderate, or vigorous activity among older adults. Lastly, there is the challenge of adherence to wearing the device on the part of the participant. Although the cognitively impaired resident may, initially, be agreeable to wearing the accelerometer, later, they may not recall and/or comprehend its use or it may be slightly uncomfortable, causing the individual to remove and possibly lose the device during the course of data collection.

There are numerous options for measuring physical activity among older adults with cognitive impairment with the most commonly used devices being the ActiGraph Gt3X and ActiGraph 7164 (Pensicola, FL), the activPAL (Kolikkotle, Finland), and the MotionWatch 8 (Cambridgeshire, UK; Table 1). Accelerometers may be placed around the waist, on the thigh, around the ankle, or on the wrist (either dominant or nondominant hand). Placement on the hip has the advantage of evaluating activity that reflects the larger muscles of the lower extremities and correlates with energy expended by the individual (Matthews, Hagstromer, Pober, & Bowles, 2012). There are challenges to waist wear, particularly for individuals who do not like the external appearance of the belt or feel it is uncomfortable if placed under clothing and directly against the skin. Also, the belt must be removed with dressing and undressing and bathing, and this leads to increased risk of loss or removal. For obese individuals, waist wearing is also problematic both in terms of fit of the belt as well as issues with skin folds.

Table 1.

Commonly Used Accelerometry Devices With Older Adults

Device	Placement	Pros	Cons
ActiGraph	Waist/hip	Collects a large amount of data; Evidence of reliability and validity; Used extensively so easy to compare across studies; Identifies intensity of activity and also can count steps.	Default is the Freedson calculation for cut points and changing this requires programming; Discomfort and challenges associated with waist wearing/belt use and having to remove and replace this; Need to consider placement and type of activity collected (wrist will be more focused on upper-extremity movement).
ActivPAL	Thigh	Focuses on measurement of time spent in different activities—sitting, standing, lying; Can leave on for days without removal.	Does not provide information about intensity of activity; May cause skin irritation on the thigh; Older adults with dementia may remove this as there is easy access. Does not identify upper extremity-focused activity.
Actical	Waistline; hip, most commonly, but can be worn on the wrist or ankle.	Set cut points but algorithms and advanced analyses can be obtained.	Belt needs to be removed and replaced, which increases the risk of missing data and missing devices; Set cut points may not be relevant to older adults with dementia. Need to consider placement and type of activity collected (wrist will be more focused on upper-extremity movement).
MotionWatch 8	Worn on the wrist	Can easily individualize cut points; Comfortable, easy wear consistent with wearing a wristwatch.	The device is visible and older adults with dementia may remove it and lose it; Captures activity driven by hand movement and, thus, may miss some activities if hands are not moving.

Thigh-worn devices have the advantage of providing valid information about the time spent in sitting, standing, or ambulation. Findings when using thigh-worn devices in community-dwelling adults <70 years of age were consistent with those worn on the wrist (White et al., 2019). It is possible, however, that thigh-worn devices may not pick up all activities that are done exclusively with the upper extremities, activities that are more prevalent in older adults. Ankle-worn devices are another option and have the advantage of identifying ambulation-related activities with good accuracy but may not identify activities done when sedentary, such as some resistance exercises, piano playing, or knitting. Advantages of ankle devices for use with individuals with cognitive impairment include the fact that they generally are unable to remove this device due to functional status or because it is hidden under a sock. Conversely, because of the high incidence of venous insufficiency and congestive heart failure, there are many older adults with peripheral edema who experience constriction and/or discomfort with a device placed around the ankle (White et al., 2019).

Wrist accelerometry has increasingly been used as an option for assessment. There are several advantages to wrist assessments as they tend to be more acceptable, particularly to older adults who are accustomed to wearing a watch. Consequently, there is a tendency to be more adherent to wearing the watch over periods of time (Troiano, McClain, Brychta, & Chen, 2014). Wearing the device on the wrist is also more likely to capture the many activities that older adults engage in, including upper extremity exercises and wheelchair mobility. Traditionally, the recommendation has been to wear the wrist monitor on the nondominant hand. Multiples studies have shown, however, that there is no difference in results based on use of the dominant or nondominant wrist (Dieu et al., 2017; White et al., 2019). Conversely, one of the challenges of use of devices placed on the wrist is that they may not pick up walking accurately among older adults who use assistive devices as there is no arm movement associated with ambulation. This is particularly problematic when rolling walkers are used as the individual does not even need to lift and move the device. Similarly, normal gait changes among older adults, particularly those with dementia, result in decreased arm swing (Mirelman et al., 2015), which may impact the identification of time spent walking.

Interpretation of Findings From Accelerometers Used With Older Adults With Dementia

One of the strengths of accelerometry is that it provides very detailed objective reports of minute by minute physical activity. These reports are generally provided based on counts per minute (CPM) of activity. The device counts the number of times the waveform crosses zero for each time period being evaluated. Most of the current accelerometers used provide counts of activity that are calculated through the assessment of triaxial analyses of movement. Cut points based on counts are used to determine the intensity of the activity being performed regardless of the type of activity (e.g., walking, swimming). There is, however, no consensus on the cut points for older adults associated with any level of activity. Freedson et al. (1998) cut points are commonly used as the default in several accelerometers, such as the ActiGraph. These cut points are set at 0–99 CPM for sedentary activity; 100–760 CPM for light activity; 761–1951 CPM for lifestyle activity; 1,952–5,724 CPM moderate-level activity; 5,725–9,498 CPM for vigorous activity; and 9,499–∞ CPM for very vigorous activity. The Freedson cut points are more conservative than those established by Landry et al. (2015) based on testing in a group of community-living older adults. Landry cut points set sedentary behavior at ≤178.50 CPM; light activity was set between 178.51 and 562.49; moderate-level physical activity was set at ≥562.50 CPM and vigorous activity was set at ≥1,020 CPM. Depending on how cut points are set, there may be an overestimation or underestimation of activity.

Wear Time Recommendations

Numerous studies (Bai et al., 2016; Falck, Landry, Brazendale, & Liu-Ambrose, 2017; Hart, Swartz, Cashin, & Strath, 2011; Hutto et al., 2013; Sasaki et al., 2018) have reviewed the length of time adults need to wear the accelerometer to establish reliable daily assessments of physical activity. These studies suggested a range of 2–7 days being sufficient to capture reliable daily activity reports. Regardless of the number of days the device is worn by participants, researchers generally report the amount of activity done in a 24-hr period. This facilitates comparisons across sites and with physical activity guidelines. For older adults with dementia, the increased risk of device removal and loss and the potential for discomfort for any participant for multiple days of wear warrant further exploration of the fewest days of wear needed to assure reliable assessment of physical activity.

To gain a better understanding of best practices for obtaining reliable and valid accelerometer data, this study tested the feasibility, reliability, and validity of the MotionWatch 8 among a large group of assisted living residents. Feasibility was based on the willingness and ability of the resident to wear the MotionWatch 8. Reliability testing was based on evidence that there was consistency in measurement across the 3 days of wear. Specifically, it was hypothesized that among residents with mild to moderate cognitive impairment there would be no significant difference in activity across 3 days of wear. For validity, it was hypothesized that (a) controlling for age, gender, comorbidities, and cognition, there would be a significant association between function (i.e., ADL) and counts of physical activity based on the MotionWatch 8, and (b) those with cognitive impairment would engage in fewer counts of activity, more sedentary activity, and spend less time in moderate and vigorous activity when compared with those without cognitive impairment.

Methods

Design

This was a secondary data analysis using baseline and Month 4 data from the study Dissemination and Implementation of Function Focused Care in Assisted Living Using the Evidence Integration Triangle. Function Focused Care in Assisted Living Using the Evidence Integration Triangle includes a four-step intervention that is implemented by a Research Nurse Facilitator working with a facility-identified champion and stakeholder team over a 12-month period. The four steps include: (Step I) Environment and Policy Assessments, (Step II) Education, (Step III) Establishing Resident Function Focused Care Service Plans, and (Step IV) Mentoring and Motivating. A total of 85 sites participated in the study, with 32 in Maryland, 33 in Pennsylvania, and 19 in Massachusetts. To be eligible to participate, the sites had to (a) have at least 25 beds; (b) identify a nurse (a direct care worker, licensed practical nurse, or registered nurse) to be the champion and work with the study team in the implementation of Function Focused Care in Assisted Living Using the Evidence Integration Triangle; and (c) have access to e-mail and websites via a phone, tablet, or computer.

Sample

Residents were eligible to participate if they were (a) 65 years of age or older, (b) able to speak English, (c) living in a participating assisted living setting at the time of recruitment, and (d) able to recall at least one out of three words as per the Mini-Cog test (Borson, Scanlan, Chen, & Ganguli, 2003). Residents were excluded from the study if they were enrolled in hospice. Potentially eligible participants were identified by the staff in the assisted living setting and were randomly approached until 10 residents per setting were recruited. If the resident was able to answer all questions correctly on the Evaluation to Sign Consent (Resnick et al., 2007), he or she was able to self-consent. If not, the individual had to assent to participate either verbally or via written signature and the Legally Authorized Representative was approached to complete the consent process. The study was approved by the University of Maryland Institutional Review Board, and all participants (and/or the Legally Authorized Representative) provided written or verbal assent and/or consent to participate.

Overall, 1,345 residents were screened and, of these, 1,329 were eligible (16 not eligible due to age or receiving hospice care). Of the eligible residents, 806 residents consented (61%), 482 refused (36%), and 41 (3%) other residents could not assent or their Legally Authorized Representative could not be reached. Of those consented, 12 (1%) were not eligible due to cognitive status, which left us with 794 participants that were enrolled. Twelve individuals withdrew from the study prior to baseline data collection, and one individual moved out of the facility, which yielded 781 individuals with data who continued in the study. At 4 months, there were a total of 690 participants continuing in the study.

Procedure

Descriptive data were collected by research evaluators using chart data, information obtained from the nurse working with the resident on the day of testing, and from resident observations. At baseline and at 4 months, the MotionWatch 8 was placed on each participant for a 5-day period with the goal of obtaining 3 full days of data during Days 2, 3, and 4 of wear. Participants and the staff working with the participants were informed that the watch was to be worn at all times, including showering, bathing, swimming, and when sleeping. All 7 days of the week were included, with some participants wearing the MotionWatch 8 on the weekends as well as during the weekdays.

Measures

Descriptive information included age, gender, race, cognition, comorbidities, and physical function. Age, gender, and race were obtained from chart review. Cognition was based on a single item from the Mini-Cog test, which focused on the ability to recall three out of three words (Borson et al., 2003). Comorbidities were based on the Cumulative Illness Rating Scale for Geriatrics (Linn, Linn, & Gurel, 1968). Scoring was done by summing the number of comorbidities based on the 13 organs or systems included in the measure. Function was evaluated using the Barthel Index (Mahoney & Barthel, 1965), which includes 10 items evaluating the individual’s ability to perform basic ADL (e.g., bathing, dressing) alone or with some degree of assistance from others (fully dependent vs. semi-dependent). Items are weighted to account for the amount of assistance required. A score of 100 indicates complete independence. The Barthel Index was completed by asking the nurse or nursing assistant working with the patient on the day of testing how the resident performed with regard to each functional activity. Prior testing using the Barthel Index provided evidence of reliability based on alpha coefficients of .62–.80 and interrater reliability based on an intraclass correlation (ICC) of .89 (Resnick & Galik, 2007). Validity of the measure was based on a significant correlation (r = .97, p < .05) between the Barthel Index and the Functional Inventory Measure (Resnick & Galik, 2007).

The MotionWatch 8 is a compact, lightweight, water-resistant, body-worn activity monitoring device used to measure physical movement. The device captures movements during routine daily living, including time spent in sedentary, light, moderate, and vigorous activity. The MotionWatch 8 is placed on the wrist to obtain data related to movement that, then, estimates overall movement. The data are gathered at 50 Hz and processed into “epochs” of set periods of time, generally at 1-min intervals. The data are then stored in an internal nonvolatile memory and can be easily downloaded for analysis at the end of the assessment period.

In contrast to other accelerometers, the MotionWatch 8 has the advantage of allowing the user to set the cut points for counts of activity to establish the overall time spent in sedentary, light, moderate, or vigorous activity (CamNtech, 2016). Directions for setting cut points are described in Table 2. The majority of participants in our study were either physically unable to walk at a moderate level of activity for 5 min or cognitively unable to follow those directions, so individual reference levels could not be accurately calculated. Consequently, we used the previously established reference levels for the MotionWatch 8 for sedentary, moderate, and vigorous physical activity when worn by older adults as described by Landry et al. (2015). As noted earlier, these cut points were set at ≤178.50 CPM for sedentary activity; ≥562.50 CPM for moderate-level physical activity; and ≥1,020 CPM for vigorous activity. Prior testing of the MotionWatch 8 with older adults provided evidence for reliability based on 3 days of wear, and evidence of validity was based on a significant relationship between activity counts and a measure of perceived exertion on the part of the participant (Chakravarthy & Resnick, 2017). We used a score of at least 1,000 counts per day as providing evidence of wear for that day. Other researchers have defined nonwear time as 15-min blocks of 0 counts or periods of 180 min of 0 counts (Cabanas-Sanchez et al., 2020; Parry, Chow, Batchelor, & Fary, 2019). Our study participants often remained sedentary, accumulating longer periods of 0 counts of activity during wear time, and, thus, those nonwear time guidelines were not appropriate. Although 1,000 counts per day is an extremely low level of activity, it is consistent with a possible activity level among individuals in assisted living settings (Chakravarthy & Resnick, 2017).

Table 2.

Calibration of Cut Points to Establish Sedentary, Moderate, and Vigorous Levels of Physical Activity

MotionWatch 8 calibration of cut points for individual residents
Step 1: After placing the MotionWatch 8 on the resident, have him or her perform a brisk walk (3–4 mph) for at least 5 min. This should take place at the start or at the end of a recording to ensure that it is easily identifiable when looking at the analysis outcomes.
Step 2: The default will use an average of the activity counts found within that 5-min period to calculate sedentary, moderate, and vigorous levels of physical activity for that individual.
Alternative approach to calibration of cut points for individual residents
Cut points can be preset by the investigator to match previously established set points for older adults. Examples include set points by Freedson, Melanson, and Sirard (1998)—these cut points are set at 0–99 counts per minute for sedentary activity; 1,952–5,724 counts per minute for moderate-level activity; and 5,725–∞ counts per minutes for vigorous activity—or Landry, Falck, Beets, and Liu-Ambrose (2015)—≤178.50 counts per minute sedentary activity; ≥562.50 counts per minute for moderate-level physical activity; and vigorous activity was set at ≥1,020 counts per minute.

Data Analysis

Descriptive statistics were used to describe the sample and test for normality and MotionWatch 8 outcomes. Feasibility was based on at least 70% of the residents wearing the MotionWatch 8 for at least one full day (Ridgers, McNarry, & Mackintosh, 2016). Reliability of the device was based on ICCs over the 3-day period for counts of activity and time spent in sedentary, moderate, and vigorous activity at baseline recordings and at 4-month recordings. An ICC of .75–.90 or greater was considered indicative of good reliability (Koo & Li, 2016). Lastly, the validity of the MotionWatch 8 was tested using a linear regression analysis. Step 1 of the analysis controlled for age, gender, comorbidities, and cognition. Function was then entered into the model. Entry level of acceptance at each step was p = .05, and removal was set at p = .10. A multivariate analysis of covariance was done to test the second hypothesis. The Pillai–Bartlett Trace was used to determine multivariate significance. Levene’s test of equality of error variances was used to establish that the error variance of the dependent variable was equal across groups.

Results

As shown in Table 3, the mean age of the participants was 89.48 (SD = 7.43) years and the majority were female (71%) and White (97%). The participants had a mean score of 2.39 (SD = 0.76) on the three out of three item recall on the Mini-Cog test with close to half (44%) of the participants showing some evidence of impairment based on scores of less than three out of three recall. Participants had an average of five comorbidities (SD = 1.95). Overall, the sample was functionally independent with regard to basic ADL with a mean of 81.27 (SD = 21.75) on the Barthel Index. On a daily basis, they engaged in a mean of 131,549 (SD = 106,422) counts of activity, spent 1,198 (SD = 183) min in sedentary activity, 55.83 (SD = 84.96) min in moderate-level activity, and 12.11 (SD = 27.93) min in vigorous activity. At 4 months, a total of 690 residents remained in the study and function remained essentially the same with a mean of 81.92 (SD = 21.40). Daily at 4 months, the participants engaged in 113,189 counts (SD = 91,031) of activity, spent 1,279 (SD = 943) min in sedentary activity, 37.45 (SD = 62.78) min in moderate activity, and 8.31 (SD = 32.86) min in vigorous activity. Counts of activity and sedentary activity were normally distributed, although time spent in moderate and vigorous activity were skewed to the left.

Table 3.

Sample Description

	Baseline N = 781
Variable	Mean	SD	Month 4 N = 690
Age (years)	89.48	7.43
Function (Barthel index)	81.27	21.75	81.92	21.40
Comorbidities	5.08	1.95
Three out of three recall item on the Mini-Cog	2.39	0.76
MotionWatch 8: Total counts	131,549	106,422	113,189	91,031
MotionWatch 8: Minutes in sedentary activity	1,198	183	1,279	943
MotionWatch 8: Minutes in moderate activity	55.83	84.96	37.45	62.78
MotionWatch 8: Minutes in vigorous activity	12.11	27.93	8.31	32.86
	N	%
Gender
Male	233	29
Female	561	71
Race
White	771	97
Black	23	3
Word recall
One out of three recall	137	17
Two out of three recall	214	27
Three out of three recall	443	56

Feasibility

With regard to feasibility of the MotionWatch 8, among the 781 participants with baseline data, 550 participants (70%) wore the MotionWatch 8 for at least 1 day. A total of 499 participants out of the full recruited sample of 781 participants (64%) wore the MotionWatch 8 for the full 3 days. A total of 77 (10%) participants refused to wear the MotionWatch 8 when the evaluator was there to place the device. There were 134 (17%) individuals who did not wear the MotionWatch 8 because: (a) the staff felt it would agitate the resident or he or she would remove it and lose it, (b) the resident removed it after placement or asked for it to be removed, or (c) the evaluator was unable to find the resident at the time of placement (e.g., out at a provider visit, in the hospital, or visiting with family). Additionally, in 20 (3%) cases there were issues with downloading MotionWatch 8 data either due to the computer used for initializing the device or due to issues within the device (e.g., low battery). At baseline, there was no difference between those with cognitive impairment (Mini-Cog test score of one out of three recall) versus those without cognitive impairment (Mini-Cog test scores of two or three out of three recall) in terms of wearing the MotionWatch 8 for at least 1 day (χ² = 2.76, p = .74).

Feasibility findings at 4 months were similar to those at baseline. At 4 months, there were 690 participants remaining in the study and available for data collection. Of these, 487 (71%) wore the MotionWatch 8 for at least 1 day. A total of 471 participants out of the available sample of 690 participants (68%) wore the MotionWatch 8 for the full 3 days. Among all available participants, 83 (12%) refused to wear the MotionWatch 8. There were 95 (14%) that did not wear the MotionWatch 8 because the individual was unavailable on the day of testing or the staff felt that it might agitate the resident if he or she had to wear the watch or they might remove it and lose it. There were 25 cases (4%) in which there were errors with the MotionWatch 8 due to downloading issues related to the MotionWatch 8 or the computer used to initialize the MotionWatch 8. At 4 months, there was no difference between those with cognitive impairment (Mini-Cog test score of one out of three recall) versus those without cognitive impairment (Mini-Cog test scores of two or three out of three recall) in terms of wearing the MotionWatch 8 for at least 1 day (χ² = 0.51, p = .48).

Reliability

Table 4 provides the means and SDs for Days 2, 3, and 4 for each of the MotionWatch 8 measures considered as well as the ICC. At baseline, there was evidence of reliability based on an ICC for all outcomes from the MotionWatch 8. Specifically, the ICC across Days 2, 3, and 4 for counts of activity at baseline was .95 and at 4 months was .97. For Days 2, 3, and 4 for minutes of sedentary activity, the ICC was .89 at baseline and .74 at 4 months. For Days 2, 3, and 4 of minutes per day in moderate-level activity, the ICC was .96 at baseline and .98 at 4 months. Lastly, for Days 2, 3, and 4 minutes per day for vigorous-level activity, the ICC was .95 at baseline and .98 at 4 months.

Table 4.

Descriptive Findings for Days 2, 3, and 4 and Intraclass Correlations

Variable	Day 2 Mean (SD)	Day 3 Mean (SD)	Day 4 Mean (SD)	Intraclass correlation
Baseline measurement
Counts of physical activity	137,551 (107,242)	132,022 (100,859)	127,445 (101,651)	.95
Time in sedentary activity	1,189 (180)	1,197 (172)	1,178 (210)	.89
Time in moderate-level activity	58.54 (86.62)	56.09 (79.91)	52.44 (77.13)	.96
Time in vigorous-level activity	12.73 (28.63)	11.00 (24.09)	10.19 (23.26)	.95
4-month measurement
Counts of physical activity	113,189 (91,031)	112,169 (94,106)	105,250 (91,046)	.97
Time in sedentary activity	1,279 (943)	1,236 (192)	1,253 (648)	.74
Time in moderate-level activity	37.45 (62.78)	37.45 (62.78)	34.31 (61.34)	.98
Time in vigorous-level activity	8.31 (32.86)	7.80 (28.39)	7.32 (32.92)	.98

Validity

For the first hypothesis, as noted in Table 5, linear regression analysis at baseline showed that controlling for comorbidities, gender, cognition, and age, function was significantly associated with counts of activity, F(1, 549) = 23.95, p = .001, R² change = .05, with 7% of the variance explained by the model. Similarly, at 4 months, controlling for comorbidities, gender, cognition, and age, function was significantly associated with counts of activity, F(1, 456) = 46.08, p = .001, R² change = .30, with 14% of the variance explained by the model. For the second hypothesis, as shown in Table 6, controlling for age and gender, those with cognitive impairment engaged in fewer counts of activity, F(1, 372) = 8.24, p = .006, and spent more time in sedentary activity, F(1, 372) = 8.75, p = .005, and less time in moderate, F(1, 372) = 8.66, p = .006, and vigorous activity, F(1, 372) = 6.65, p = .012. There was overall multivariate significance, F(4, 369) =2.56, p = .04.

Table 5.

Linear Regression Results for Baseline and 4 Months

Variable	Beta	t (p)	R2 change	F (p)
Model 1: Baseline
Step 1: Control variables
Three out of three recall item	−0.04	−.86 (.39)	.02	3.11 (.02)
Gender	0.13	2.81 (.01)
Age	−0.08	−1.64 (.10)
Comorbidities	0.06	1.36 (.17)
Step 2: Function (Barthel index)
Function	0.21	4.89 (.001)	.05	23.95 (.001)
Model 2: 4 months
Step 1: Control variables
Three out of three recall item	−0.06	−1.26 (.21)	.06	6.14 (.001)
Gender	0.15	3.40 (.001)
Age	−0.15	−3.29 (.001)
Comorbidities	−0.17	3.90 (.001)
Step 2: Function (Barthel index)
Function	0.15	4.13 (.001)	.30	46.08 (.001)

Table 6.

MotionWatch 8 Outcomes Among Those With and Without Evidence of Cognitive Impairment

Variable	Cognitive status	N	Mean	SD	F (p)
Day 2: Total counts	Cognitively intact	101	159,451	133,014	8.24 (.006)
	Cognitively impaired	279	123,696	103,655
	Total	380	133,199	113,135
Day 2: Minutes in sedentary activity	Cognitively intact	101	1,138	212	8.75 (.005)
	Cognitively impaired	279	1,203	193
	Total	380	1,186	200
Day 2: Minutes in moderate activity	Cognitively intact	101	79.65	110.71	8.66 (.006)
	Cognitively impaired	279	50.73	80.26
	Total	380	58.42	90.13
Day 2: Minutes in vigorous activity	Cognitively intact	101	20.33	42.16	6.65 (.012)
	Cognitively impaired	279	11.22	26.15
	Total	380	13.64	31.41

Discussion

The current study provides support for the feasibility of use of the MotionWatch 8 with residents in assisted living, including those with cognitive impairment. At both baseline and 4-month follow-up, the majority of participants (70–71%) were willing to wear the MotionWatch 8 for at least 1 day. This is similar to findings in previous research using the ActiGraph in a sample of 200 residents in assisted living (Resnick, Galik, Gruber-Baldini, & Zimmerman, 2011). In that prior study, 75% of the assisted living residents were willing to wear the waist-worn ActiGraph, and at 4 months, 65% of the participants were willing to wear the device. In our current study using the MotionWatch 8, only a relatively small percentage of participants (10%) refused to even have the device placed on their wrist for evaluation, although there was an additional 9% at baseline and 2% at 4 months that removed the device or asked for the device to be removed prior to the end of the 5-day wear time. Most of the removals were after the first day of wear.

The biggest challenge with capturing MotionWatch 8 data was the practicality of being able to find the resident on the day of testing and place the watch. Residents were, for example, out of the facility at an appointment, in the hospital following an acute event, at an activity, out on a trip, or at the hairdresser, and, thus, the evaluator was not able place the watch. Some facilities in this study were several hours drive from the university or home settings of the evaluators, and, thus, multiple trips to the setting were not practical. Future research should certainly consider additional visits or alternative ways to get the MotionWatch 8 placed on an individual to capture a larger percentage of residents. There was only a small percentage of situations (3–4%) at both time points in which missing data were due to technology. Problems included those within the MotionWatch 8 (frequently due to battery, such as an unrecognized low battery) or related to the computer used for initializing data collection or downloading the data.

Unlike what has been suggested in prior studies (Bai et al., 2016; Falck, Landry, Brazendale, et al., 2017; Hart et al., 2011; Hutto et al., 2013; Sasaki et al., 2018) with 2–7 days of wear recommended to provide reliable data, based on ICC findings, the current study found that a single day of wear was sufficient for reliable accelerometer data. For all MotionWatch 8 outcomes, including counts, sedentary, moderate, and vigorous activity, the ICCs ranged from a low of .74 for time in sedentary activity at 4 months to .98 for time in moderate and vigorous levels of activity at 4 months. Evidence of the reliability of single day data collection is particularly important given that there were up to 9% of participants who seemed to be willing to wear the MotionWatch 8 for one day but then requested that it be removed on Day 2 or Day 3. It is also possible that some of the participants refusing to wear the device for the 5 days may have been willing to wear it for 1 day. The procedure for data collection was to place the MotionWatch 8 on for a 5-day period so that Day 1 and Day 5 were just used for placement and removal. For single-day data collection, we would recommend that participants have the MotionWatch 8 placed on Day 1, recordings be obtained for the full day of Day 2, and then the device be removed at some time on Day 3.

There was sufficient evidence for validity of the MotionWatch 8 data as after controlling for age, gender, comorbidities, and cognition, function was significantly associated with counts of physical activity. Furthermore, as hypothesized, there was a significant difference such that those with cognitive impairment consistently engaged in less physical activity and more sedentary activity. Prior research has noted that those with cognitive impairment engage in less physical activity than those who are cognitively intact (Chen & Lauderdale, 2019; Ferreira, Brandão, & Cardoso, 2020; Hartman et al., 2018; Yi-Pei, Yuan-Han, & Shih-Fen, 2019). Of note, only a small amount of the variance in counts of activity was explained by the variables in the model. Although function was significantly associated with counts of activity, other variables, such as mood, motivation, fear of falling, and facility and caregiver factors, may further influence physical activity (Fleiner, Gersie, Ghosh, Mellone, Zijlstra, & Haussermann, 2019; Laybourne, Biggs, & Martin, 2011).

Although there were differences in this study in physical activity between those with and without evidence of cognitive impairment, overall the participants in this study exceeded the requirements for physical activity for older adults (American College of Sports Medicine and the American Heart Association, 2013). It is difficult to compare our findings with those reported in other studies as each study utilized different cut scores. The cut scores set by Landry et al. (2015) are more liberal than those established by others, such as Freedson, and, thus, may have resulted in an overinflation of activity. Conversely, given the physical status of the participants, it is possible that when they did engage in activity, they may have been working at a moderate level of activity. Future research should continue to explore whether there are ways to individually set cut scores for participants using the MotionWatch 8 approach and compare those findings with cut scores by Landry et al.

Study Strengths and Limitations

The strength of this study was the large sample of assisted living residents included across a variety of settings and multiple states. The sample, however, was relatively homogenous, including mostly White women, and the data related to participation in moderate and vigorous activity were negatively skewed. Several limitations in design also should be noted. We did not capture diary data or survey data related to exact wear time or the specific activities performed throughout the day by the participant. This would be very helpful to assure that the device was truly worn for the full day and to evaluate the types of activities performed, particularly among those who were high in moderate and vigorous levels of activity. Furthermore, there was variability in the days of wear with some individuals wearing the MotionWatch 8 over the weekend and others just during weekdays. Given that there is little variability in terms of activities in these settings on weekdays versus weekends and our consistency in findings over the 3 days of wear, it is not anticipated that this had a major impact on outcomes.

Conclusion

Despite the noted limitations, the MotionWatch 8 provides a reliable and valid way to measure physical activity among older adults in assisted living, including those with cognitive impairment. To facilitate data collection, a single full day of wear is sufficient to obtain reliable data. This can best be obtained by placing the MotionWatch 8 on participants on Day 1, leaving it on for a full day of data collection on Day 2, and removing the device on Day 3. The MotionWatch 8 captures data based on movement of the upper extremity and, therefore, may not record lower extremity movement when done alone (e.g., when there is no arm swing during walking). As an advantage, the MotionWatch 8 allows the researcher to alter the cut points based on individual ability or use previously set cut points for older adults. The majority of older adults, with and without cognitive impairment, are willing to wear the device, particularly if it is just for 1 day. When working with older adults in assisted living settings, however, flexibility in terms of initial timing of the placement of the watch is needed.