Ethical approval for this study was obtained from the Vancouver Coastal Health Research Institute and the University of British Columbia’s Clinical Research Ethics Board (H14-01301). All participants provided written informed consent.
Participants
Data for this study were collected as part of the Sleep and Cognition Study, a cross-sectional study examining the associations between sleep quality and cognition among older adults (26). We recruited and collected data between August 27, 2014 and June 30, 2016. Details of the full study protocol can be found elsewhere (26). We recruited 153 older adults from Vancouver, British Columbia by advertisements placed in local community centres, newspapers, and word of mouth referrals. Interested individuals were initially pre-screened for eligibility criteria.
Participants were included if they: 1) were 55+ years of age living the Greater Vancouver area; 2) scored ≥24/30 on the Mini-Mental State Examination (MMSE; [27]), with scoring of attention being performed using serial sevens; and 3) were able to read, write, and speak English with acceptable visual and auditory acuity. Participants were excluded if they were: 1) diagnosed with dementia of any type; 2) diagnosed with another type of neurodegenerative or neurological condition; 3) taking medications that may negatively affect cognitive function; 4) planning to participate or were currently enrolled in a clinical drug trial; or 5) unable to speak as judged by an inability to communicate by phone. Individuals were not excluded based on use of medications which may affect sleep quality (either negatively or positively), nor were participants selected on the basis of their sleep.
Study Design and Measurement
At study entry, we ascertained general health, subjective sleep quality using the Pittsburgh Sleep Quality Index (PSQI; [28]), height to the nearest 0.1 cm using a stadiometer, weight to the nearest 0.1 kg using an electronic scale, demographics, socioeconomic status, and education by a questionnaire. Height and weight were used to calculate body mass index (BMI; kg/m2; [29]). Global cognitive function was assessed by the MMSE (27), and the Montreal Cognitive Assessment (MoCA; [30]). We categorized participants based on MCI status with a score of <26/30 on the MoCA indicating probable MCI, which has been found to have good internal consistency and test-retest reliability, and was able to correctly identify 90% of a large sample of MCI individuals from two different clinics (30).
Participants were then fitted with the MW8 and provided detailed information on its features (i.e., the light sensor, event marker button, and status indicator). Participants were instructed to press the event marker button each night when they started trying to sleep; and again each morning when they finished trying to sleep. Consistent with established protocol for wrist-worn actigraphy, participants wore the MW8 on the non-dominant wrist (24, 31).
Participants were also given the 9-item Consensus Sleep Diary (CSD; [32]) and asked to complete it each morning upon waking. The responses from the CSD were used to confirm sleep windows as determined by the time stamped event markers. In cases where the event marker and CSD entry disagreed for the start time of the sleep window, we used activity cessation and light sensor data from the MW8 to determine “lights out”. Similarly, when the event marker and CSD entry disagreed for the end of the sleep window, we used activity onset and “lights on” to determine the end of the sleep window. If responses from the CSD entry disagreed with the event markers entered by participants as the start of the day (i.e., finished trying to sleep and awake and out of bed), we used activity onset and light sensor data to determine the start of the day. Similarly, when the event marker and CSD entry disagreed for the end of day (i.e., time spent trying to sleep), we used activity cessation and light sensor data to determine the end of the day. In accordance with the currently established protocol for measuring sleep quality via actigraphy (24, 31), each participant was continuously monitored for a minimum of 14 nights. After collection, stored activity counts were downloaded and saved to an IBM compatible computer for subsequent data reduction and analysis.
MW8 Instrumentation and Data Reduction
We measured sleep using the MW8 actigraphy system (CamNtech; Cambridge, United Kingdom). The MW8 is a tri-axial accelerometer designed to observe acceleration ranging in magnitude from 0.01G to 8G, with a frequency of 3-11Hz. The filtered acceleration signal is digitized and the magnitude is summed over a user-specified time interval. At the end of each interval, the summed value or activity “count” is stored in memory and the integrator is reset. The MW8 is the updated version of the Actiwatch7, an actigraph with evidence of validity against polysomnography in healthy adults (Mean age: 30 ± 6 years; 45% female; [33]), and also adults with chronic insomnia (Mean age: 41 ± 12 years; 78% female; [34]). There is also initial evidence of validity against polysomnography for the MW8 among 1) 54 adults with suspected sleep disorders including obstructive sleep apnea, insomnia, hypersomnia, and Ehlers Danlos syndrome (Mean Age: 53 ± 16 years; 61% female); and 2) 19 healthy adults (Mean Age: 28 ± 5 years; 53% female; [35]). For the current study, we used 60 second epochs (24, 31).
Data were analyzed using MotionWare 1.0.27 (camntech) to estimate different sleep indices including: fragmentation index, sleep efficiency (time asleep expressed as a percentage of time in bed), sleep duration (total time spent sleeping), sleep latency (time between “lights out” and falling asleep), and wake after sleep onset (time spent awake after sleep has been initiated and before final awakening). Fragmentation index is a description of restlessness while sleeping and is defined by MotionWare as the sum of 1) the total time spent sleeping categorized as mobile in the epoch-by-epoch mobile/immobile categorization expressed as a percentage of the time spent asleep; and 2) the number of immobile bouts which were equal to 1 minute in length expressed as a percentage of the total number of immobile bouts during time spent sleeping. Only minutes categorized as asleep were included in the calculation of fragmentation index.
Statistical Analyses
We performed all of our statistical analyses using R version 3.3.1 using the psych, Hmisc, and ICC packages. Our statistical code can be found in Supplementary Material S1. Two participants did not complete the MoCA. These individuals were removed from analyses such that our final sample was 151 participants.
Participant characteristics based on probable MCI status
We calculated means and standard deviations for all variables of interest based upon probable MCI status (i.e., MoCA score <26/30). We determined demographic differences in probable MCI status using independent sample t-tests for continuous variables and chi-square tests for categorical variables, using probable MCI status (yes/no) as the grouping variable. Subsequently, we performed analyses of covariance (ANCOVA) to determine differences in estimates of sleep quality based on probable MCI status. We performed separate ANCOVA models for each of our measures estimating sleep quality (i.e., PSQI total score, PSQI sleep duration, and MW8 measured sleep duration, fragmentation index, sleep efficiency, sleep latency, and wake after sleep onset), wherein we controlled for age, sex, and sleep medication use while using probable MCI status as the grouping variable.
Reliability of the MW8 for estimating sleep quality based on probable MCI status
We then calculated between-day intraclass correlations (ICC) and 95% confidence intervals (CI) for 1, 4, 7 and 14 days of monitoring, and classified ICCs according to the criteria of Koo and Li (36). For single day ICC’s we used a single absolute intraclass correlation coefficient (ICC1,1) to determine single day expected reliability using the following formula (37):
Wherein, BMS is the between subject mean square, the EMS is the residual error, TMS is the trial mean square, k refers to the number of trials (in this case, one trial), and N is the number of participants. We used all 14 days of data to calculate our single day ICCs. For our analysis of multiple day reliability, we used average random raters (ICC2,k) using the same formula, wherein k was the number of days monitored. For our calculations of 4 and 7 day reliability, we only used data from the first 4 and 7 days, respectively.
We also calculated the required days of monitoring needed to achieve ICC’s of 0.70, 0.80 and 0.90 using the Spearman-Brown prophecy formula (38). Subsequently, we calculated separate ICCs and Spearman-Brown prophecies for 1) participants with probable MCI and 2) those without MCI. We then performed z tests to determine if ICC estimates for our different sleep parameters differed significantly by MCI status after 1, 4, 7, and 14 days of monitoring.