Study design and participants’ characteristics
This study enrolled the SCGs of individuals who visited the geriatric psychiatry clinic at Chungnam National University Hospital (in South Korea) from May 2020 to August 2023. The inclusion criteria for the study participants were as follows: (1) age between 55 and 90 years; (2) serving as the primary caregiver for the spouse; (3) capable of independent functioning; and (4) no diagnosis of dementia. A total of 104 SCGs were recruited for the study. Of these, 54 caregivers voluntarily agreed to wear a Fitbit device. Participation in Fitbit monitoring for more than two weeks was entirely voluntary. The remaining participants declined to wear Fitbit device due to discomfort with wearing the device, scheduling conflicts, or lack of access to a compatible smartphone for data syncing. Of the 54 SCGs, 30 were caring for spouses diagnosed with dementia, 19 were caring for spouses with MCI, and 5 were caring for spouses with normal cognitive function (CN).
Dementia was diagnosed using the DSM-IV criteria, while MCI met the core clinical criteria recommended by the National Institute on Aging and Alzheimer’s Association guidelines [37]. CN was defined as having a CDR score of 0 and a Mini-Mental State Examination (MMSE) score of 27 or higher [38] and was treated for anxiety disorder or insomnia. The SCGs underwent a comprehensive clinical assessment by experienced neuropsychologists and research nurses. This study did not involve any clinical trials. Clinical trial number: not applicable.
Circadian rhythm assessment
Sleep-wake cycle variable
Participants were instructed to wear the device daily for a minimum of two consecutive weeks. The following objective sleep-wake cycle parameters were obtained from wearable Fitbit: sleep duration, sleep efficiency, and sleep onset and offset times (with onset and offset times expressed in minutes from midnight).
The Korean version of the PSQI [39] was used to determine participants’ self-assessed sleep-wake cycle [22]. In the PSQI, each question is assigned a score ranging from 0 to 3, resulting in a total score between 0 and 21. Participants with scores above 5 were considered to have poor sleep quality. Component scores were derived for subjective sleep quality, sleep latency (i.e., time taken to fall asleep), sleep duration, habitual sleep efficiency (i.e., the ratio of time a person actually sleeps to the total time they spend in bed), sleep disturbances, use of sleep medications, and daytime dysfunction.
Circadian rhythms of heart rate parameters
Heart rate data were collected via photoplethysmography sensors embedded in the Fitbit device, summarized every 15 min, and averaged over the next valid day to generate participant-level estimates. If less than 75% of heart rate data was collected during the analysis time, the analysis was excluded. Data were retrieved using a custom Python script via the Fitbit Web API. The utility and feasibility of using Fitbit in clinical research has been previously reported and clinical outcomes have been documented in several studies [29, 31].
To evaluate CHR, we applied cosinor analysis via CosinorPy using the 2-day continuous heart rate data. Cosinor analysis can measure the following key circadian rhythm parameters: amplitude (half the difference between peak and trough of the fitted curve), Midline Estimating Statistic of Rhythm (MESOR; the rhythm-adjusted mean heart rate), acrophase(the timing of the peak in the fitted rhythm, expressed in hours), and goodness of fit (GoF). GoF was calculated as an R-squared value, indicating how well the observed data fit the cosinor model. Values closer to 1 indicate better model fit and stronger circadian rhythmicity.
Clinical assessments
Caregiving burden
Caregiving burden was assessed using the Korean version [40] of 22-item Zarit Caregiver Burden Interview (ZBI). Each item is scored on a five-point Likert scale, ranging from 0 (never) to 4 (nearly always), yielding a total score between 0 and 88 [41].
Other clinical variable assessments
Global cognition of each participant was assessed using the Korean version of the MMSE [38]. To evaluate the severity of depressive symptoms, we used the Korean version of the Geriatric Depression Scale [42]. Further, we used the Korean version of the International Physical Activity Questionnaire (IPAQ) [43], which utilizes the Metabolic Equivalent Task (MET) variable to determine physical activity categories. The total minutes spent engaged in physical activity of various intensities over the past seven days were identified, and responses were transformed into MET-minutes per week (MET-min/week) according to the IPAQ scoring protocol [44]. Average MET scores were calculated for each activity type. The following MET values were used: walking = 3.3 MET, moderate-intensity activity = 4.0 MET, vigorous-intensity activity = 8.0 MET. Total physical activity was calculated as the sum of the MET-min/week values derived from walking, moderate-intensity activity, and vigorous-intensity activity.
Statistical analysis
To investigate whether there were differences in the characteristics of the full sample (n = 104) and the Fitbit-wearing subgroup (n = 54, of whom 52 also completed the PSQI), Mann–Whitney U tests were used for continuous variables, and chi-square tests were used for categorical variables.
Multiple regression analyses were then conducted to examine the association between caregiving burden, as measured by the ZBI, and circadian rhythm related variables. Analyses were performed sequentially for: (1) Fitbit-derived sleep-wake cycle parameters (sleep duration, sleep efficiency, sleep onset time, and offset time); (2) PSQI scores (total and subdomains); and (3) CHR parameters (amplitude, MESOR, acrophase, and GoF). In all models, SCG’s age and sex, as well as the care recipient’s cognitive status (dementia vs. non-dementia), were included as covariates.
Additional moderation analyses were conducted to examine whether SCG sex or care recipient cognitive status moderated the relationship between caregiving burden and circadian rhythm outcomes. Care recipient cognitive status was categorized as dementia versus non-dementia for these analyses. To limit the number of exploratory tests, these were only performed on sleep or circadian variables that showed a trend-level association (p < 0.1) with caregiver ZBI score in the initial regression model. The interaction term ([moderator] × [ZBI]) was used as an independent variable; sex of the SCG, as well as the cognitive status of care recipients, were treated as covariates when appropriate, while circadian rhythm variables were treated as dependent variables.
All analyses were performed using SPSS version 21.0 (SPSS Inc., Chicago, IL). Statistical significance was set at p < 0.05.