1. BACKGROUND
Measuring behavioral rhythmicity and specific patterns of activities using wearable devices such as an actigraphy watch (wrist‐worn actigraphy) is receiving more attention in ageing and dementia research because these devices are capable of objectively and longitudinally capturing behavioral changes, 1 , 2 , 3 Certain circadian rhythmicity changes are potential markers of ageing or progression of dementia, 1 , 4 and linked to the development of behavioral and psychological symptoms of dementia (BPSD). 5 Specific BPSD, such as agitation and aggression (A/A), are associated with the severity of dementia, 6 heavily burdensome to caregivers and society, and detrimental to the patients themselves, while difficult to treat. 7 , 8
Although using actigraphy as an objective tool to assess behavioral patterns in healthy older adults 1 and early stages of dementia without agitation appears feasible and acceptable in some studies, 9 , 10 its application in late‐stage agitated dementia patients is still largely unknown. A recent scoping review provided an informative roadmap, where most of the nine studies included were proof of concept in design, aiming to characterize and/or monitor agitated behaviors in dementia patients. 11 Only two small studies (n = 1 and n = 5, respectively) 12 , 13 recruited patients with well‐defined significant agitation at baseline, and no study reported acceptance and compliance rates in detail, leaving the feasibility and acceptability of wearable devices over an extended period questionable. A recent systematic review found poor acceptance of assistive technology including wearable devices in older people living in long‐term care facilities, 14 and that acceptance and compliance could worsen with the severity of cognitive impairment. Another review suggested that a personalized strategy is needed for both the patients and their caregivers, because with worsening cognitive function, increasing support to wear the watch from caregivers may also be needed. 15 We therefore investigated acceptance (whether the participant agrees implicitly or explicitly to wear the watch) and compliance (time spent wearing the watch) in late‐stage dementia patients with significant agitation over 4 weeks, and explored from both patients and caregivers the potential reasons and solutions for non‐acceptance and non‐compliance.
2. METHODS
2.1. Participants and study procedures
We tested the use of wrist‐worn actigraphy within the Sativex for the treatment of the Agitation & Aggression in Alzheimer’s Dementia (STAND) trial. The full protocol of the STAND trial has been published earlier. 16 Briefly, the STAND trial recruited care home residents aged between 55 and 95 years old with a probable Alzheimer’s Disease diagnosis 17 and clinically significant A/A symptoms, as defined by the Cohen‐Mansfield Agitation Inventory (CMAI) ≥ 45 18 and/or Neuropsychiatric Inventory Nursing Home Version (NPI‐NH) agitation sub‐score ≥ 4. 19 There was no restriction on mobility status; however, care home residents with severe, unstable or poorly controlled medical illness as deemed by the study doctor or the principal investigators, were excluded.
The first participant was recruited on the October 19 2021, and the recruitment continued until July 2022 through the National Institute for Health Research (NIHR) Maudsley Biomedical Research Centre Care Home Research Network (CHRN), and from caregivers or family members responding to trial adverts. After receiving written and witnessed informed consent from either the participant (if deemed to have mental capacity), a personal legal representative, or a professional legal representative. The research team arranged a screening visit to record the demographic and clinical characteristics and to confirm eligibility. The participant was randomized to either the Sativex or placebo arm after passing the screening, and watch‐wearing compliance was examined for 4 consecutive weeks. The team remained blinded throughout the study, and all participants were asked to wear the actigraphy watch on their non‐dominant wrist. Participants and care home staff were informed about the purpose of the watch and instructed to avoid removing or interacting with it. Each participant was visited three times during the trial period (baseline, midpoint, and 4‐weeks trial end) to mount, change, and retrieve the watch, and to obtain trial‐related information.
RESEARCH IN CONTEXT
- Systematic review: We searched PubMed for studies applied wearable device to monitor behavioral symptoms in late‐stage dementia patients and found most of the previous studies were in proof‐of‐concept stage, with very small sample sizes and/or short device‐wearing duration. Many studies didn’t recruit patients with pre‐defined behavioral symptoms and didn’t report detailed compliance data.
- Interpretation: Our study, recruiting the most late‐stage dementia patients with significant pre‐defined agitation at baseline, found that wrist‐worn actigraphy could be a feasible behavior‐monitoring tool for an extended period of time. Unlike in early‐stage dementia patients, we found late‐stage dementia patients’ compliance appears to be better along with the worsening of cognitive function. Clinical setting and sex may also contribute to compliance, but not age or behavioral symptomatology.
- Future directions: The findings would encourage the application of wrist‐worn actigraphy in late‐stage agitated dementia patients of larger scales, to explore the interaction between their behavioral symptoms, circadian rhythm, sleep, and cognitive decline.
2.2. Wrist‐worn actigraphy
The “GeneActiv Original,” is one of the most widely‐used research grade wearable tri‐axial accelerometers for the assessment of physical activity and sedentary behavior. 20 It is worn with a wrist strap but has a blank face that does not indicate the time. Its technical reliability and capacity to classify activity levels have been validated and described in detail previously. 21 The sampling frequency of the watch was set at 50 Hz. In addition to movement, a light sensor captures environmental light data that can be used for further analysis (Figure S1).
2.3. Questionnaires
In late‐stage dementia where verbal communication is impaired, standard cognitive tests are of limited usefulness. We therefore used three questionnaires that are suitable in severe dementia to assess (1) functional capacity (Functional Assessment Staging Tool, FAST 22 ), which has validated correlation to cognitive function in late‐stage dementia 23 ; (2) BPSD symptoms and their severity (Neuropsychiatric Inventory‐Nursing Home version, NPI‐NH) 19 ; and (3) a specific scale for the severity of agitation/aggression (Cohen‐Mansfield Agitation Inventory, CMAI). 18
2.4. Interviews and feedback
At each visit, participants and their caregivers were briefly interviewed to collect qualitative feedback on their feelings and attitude toward the watch‐wearing. As most of the participants were expected to have severe dementia and may not have been able to respond verbally, their non‐verbal responses such as hand gestures, facial expressions and other body languages were also noted.
At baseline, the initial response to wearing the actigraphy watch was recorded. Previous watch‐wearing habit was checked with both the participant and caregiver. At the midpoint and the trial end (fourth week), participants were interviewed again for their feelings and experiences wearing the watch. Their primary caregivers were interviewed separately for their opinions on assisting participants to wear the watch and for any observable behavior indicating non‐acceptance or non‐compliance (including watch damage or removal). If any significant non‐acceptance or non‐compliance was observed, the caregiver was additionally asked for the context, potential reasons, and suggestions for improvement.
2.5. Data analysis
The R markdown package developed by Activinsight 21 was used to generate a daily personal activity report to visually inspect specific shift periods of non‐wear (i.e., whether the non‐wear fell within day shift 06:01‐14:00, evening shift 14:01‐22:00, or night shift 22:01‐06:00). The non‐wear duration was calculated using a well‐validated, open‐sourced R package GGIR (version 2.8‐2) 24 , 25 and compared with the visually inspected results.
In the study, acceptance rate was defined as the percentage of participants who explicitly or implicitly (e.g., did not try to remove) accepted the watch. The compliance rate of each participant was calculated as the percentage of total wear time (trial duration—nonwear time/ trial duration × 100). Correlations between demographic and clinical parameters at baseline (including age, FAST score, NPI and CMAI total scores and sub‐scores; FAST score 6a to 6e and 7a to 7f were converted to 6.0 to 6.8 and 7.0 to 8.0 for ordinal regression) and watch wear compliance were calculated with Spearman’s correlation coefficients. Because this is an exploratory study interested in the overall pattern of findings rather than their statistical significance, we did not correct for multiple comparisons when correlating clinical and demographic measures with watch measures. Separate mixed model repeated measure analyses of variance (ANOVAs) were used to investigate the effect of shift (morning, evening, and night shift) and sex (male and female) on non‐compliance across the 4 weeks of the study. Greenhouse‐Geisser correction was used for violations of sphericity. Basic R packages and the SPSS statistical software version 26 (IBM SPSS) were used. P < 0.05 was used as the threshold for reporting.