Time spent in physical activity, sedentary behavior, and sleep: Associations with self-rated sleep quality in middle-aged and older adults

Objectives: We examined the associations of estimated allocations of time spent in physical activity, seden- tary behavior and sleep with self-rated sleep quality. Methods: Between 2011 and 2016, 1918 participants (mean age 71 § 9 years, 51% women) from the popula- tion-based Rotterdam Study were included. Durations of light physical activity, moderate-to-vigorous physical activity, sedentary behavior, and sleep were assessed by accelerometry, self-rated sleep quality with the Pittsburgh Sleep Quality Index. Associations were assessed with compositional isotemporal substitution analyses. Results: Spending 30 minutes more in sedentary behavior (adjusted mean difference in PSQI score: 0.21, 95% con ﬁ dence interval [0.15; 0.28] or in light physical activity (adjusted mean difference in PSQI score: 0.25 [0.03; 0.46], and 30 minutes less in sleep, was associated with poorer sleep quality. Conclusions: Our ﬁ ndings suggest reducing sedentary behavior and increasing sleep duration might be a potential intervention target to improve sleep quality in this population of middle-aged and older adults. Authors. Published Elsevier National Sleep an access the


Introduction
Perceived poor sleep quality is common in middle-aged and older adults and improving sleep quality has been suggested as an important factor in promoting general health. 1,2 Increasing physical activity and reducing sedentary behavior might improve sleep quality, 3,4 but these behaviors have mainly been studied as single exposures, not taking into account that the durations of physical activity, sedentary behavior and sleep occur in the context of the 24-hour day. [3][4][5] Consequently, an increase in time spent on one behavior must be offset by time spent on other behavior. To explain how this is related to perceived sleep quality and which allocations of time might help improve perceived sleep quality, we studied the full 24-hour composition of activity behaviors and sleep in a population of middle-aged and older adults.
Compositional isotemporal substitution analyses account for the 24-hour constrained nature of activity behaviors and sleep 5 and enable estimating the effect of substituting time spent on one behavior (eg, physical activity) by another (eg, sleep) on the outcome of interest. We used these methods to examine how estimated allocation of time spent on light physical activity, moderate-to-vigorous physical activity, sedentary behavior, and sleep were associated with self-rated sleep quality. These findings will inform which allocations of time should receive particular interest in promoting sleep quality.

Study design and population
This cross-sectional study was performed within the populationbased Rotterdam Study, a prospective cohort of participants aged 45 years and over. 6  who.int/ictrp/network/primary/en/) under shared catalogue number NTR6831. All participants provided written informed consent to participate in the study and to have their information obtained from treating physicians.

Measurements
Physical activity, sedentary behavior, and sleep duration Participants wore a triaxial accelerometer (GeneActiv; Activinsights Ltd, Kimbolton, UK) on the non-dominant wrist and simultaneously filled out a sleep diary details are published elsewhere. 7 Accelerometer data were processed using PAMPRO software in Python (2.6.6). Activity was categorized based on acceleration relative to gravity (g units; 1g = 9.81 m/s 2 ) into durations of sedentary time (<48 mg), light (48-154 mg), and moderate-to-vigorous (>154 mg) physical activity. 8 Night-time sleep duration was estimated using the validated GGIR algorithm, version 1.6-7. 9 Sleep duration was subtracted from total sedentary time to estimate sedentary behavior (in and out of bed). Sedentary behavior was additionally divided into sedentary time in bed (ie, time spent awake during the nocturnal sleep period) and out of bed based on the sleep diary.

Sleep quality
Sleep quality was measured using the 19-item PSQI, which measures sleep quality and disturbance over the past month. 10 The global score (range 0-21, higher scores indicate poorer sleep quality) is the sum of 7 subscale scores (range 0-3): subjective sleep quality, sleep latency, sleep duration, sleep efficiency, sleep disturbances, use of sleep medication, and daytime drowsiness.

Other variables
During a home interview, self-reported age and sex, living situation (living alone or living with partner), education (primary, lower, intermediate, or higher education according to the UNESCO classification), paid employment (currently employed or currently not employed), smoking (current, former, or never smoker), alcohol use (drinks per day), and depressive symptoms (Center for Epidemiological Studies Depression Scale 11 ) were assessed.
During a research center visit, height and weight were measured on calibrated scales to calculate body mass index. Information on several chronic diseases (history of cancer, coronary heart disease, stroke and diabetes )and medication use (psycholeptics and psychoanaleptics) was obtained from both self-report and by linking medical records from general practitioners and pharmacies in the study area.

Statistical analysis
Based on the average daily durations of light physical activity, moderate-to-vigorous physical activity, sedentary behavior, and sleep, a compositional variable was constructed using the "Compositions" package in R. This variable is expressed in isometric log ratio coordinates and represents the proportions of time spent in each activity. 5 Linear and logistic regression models were used to assess the associations of the composition of activity behaviors and sleep with the PSQI total score and subscales. 5,12 Subscales were analyzed as dichotomous outcomes (1 vs. 2), due to the ordinal nature and skewed distribution. The compositional isotemporal substitution models were used to estimate the difference in outcome when a fixed duration of 30 minutes of time spent on one behavior (eg, physical activity) was instead spent in another behavior (eg, sleep), while other behaviors remained constant. 5 Analyses were adjusted for age, sex, and covariates mentioned above. Multiple imputation (m = 5) was used to impute missing data on covariates (<2%). We included sedentary time in and out of bed as separate components of the compositions to assess whether time awake in bed was driving associations that included sedentary behavior. Data were handled and analyzed using SPSS Statistics version 24.0.0.1 (IBM Corp., Armonk, NY) and R version 3.5.1 (The R Foundation for Statistical Computing, Vienna, Austria), using the mice, compositions and deltacomp packages.  Consistent with other observational studies, 4,13 less sedentary time was associated with a better sleep quality, but only when that time was used for sleep, and not any other activities. The association of sedentary time with poorer sleep quality could not be explained merely by more sedentary time in bed, suggesting that sleep research and interventions should focus beyond reducing sleep onset latency and wake after sleep onset. In contrast, more sedentary time in bed was associated with better perceived sleep quality in our sample. This could be due to sedentary time in bed being used for relaxing activities, which might positively affect perceived sleep quality, 14 or due to more sedentary time in bed signaling that one has spent enough time in bed to acquire needed sleep (i.e., no sedentary time in bed may indicate that the sleep window is too short). Otherwise, we could speculate that this finding can be explained by using accelerometry, as it might denote periods as wake although these might not be experienced as such. 15 Additionally, although we used a high sensitivity threshold, awake time in bed may have been overestimated. 9 Contrary to previous literature, we did not find any associations between moderate-to-vigorous physical activity and self-rated sleep quality, 3 but previous studies typically did not take sleep duration into account. We could speculate that persons who perform more physical activity are generally more aware of their health and lifestyle, including sleep. Therefore, the effect of physical activity on sleep quality might have been overestimated and is likely not a target for improving sleep quality in this age group.
Several limitations should be considered, such as the cross-sectional design of the study, the underestimation of certain physical activity behaviors by accelerometry (eg, cycling), 16 and lack of information on other dimensions of behaviors, for example, context or Plots represent the estimated differences in the outcome (Pittsburgh Sleep Quality Index) when sleep time is reallocated by sedentary behavior (left graph), light physical activity (middle graph) or moderate-to-vigorous physical activity (right graph). The mean composition of the population was used as reference. PA, physical activity; Mod-vig, moderate-to-vigorous; CI, confidence interval; min, minutes.
timing. Also, accelerometry is not the gold standard for determining sleep.

Conclusions
Spending more time in sedentary behavior or light physical activity instead of sleep, but not instead of any other activities, is associated with a poorer sleep quality. Overall, this suggests that mainly sedentary behavior, as opposed to physical activity, is a potential intervention target in this population of middle-aged and older adults, but only if replaced by sleep.

Declaration of conflict of interest
The authors have no conflict of interest to disclose.

Funding
The Rotterdam Study is funded by Erasmus Medical Center and Erasmus University, Rotterdam, Netherlands Organization for the Health Research and Development (ZonMw), the Research Institute for Diseases in the Elderly (RIDE), the Ministry of Education, Culture and Science, the Ministry for Health, Welfare and Sports, the European Commission (DG XII), and the Municipality of Rotterdam.

Data availability statement
Data can be obtained upon request. Requests should be directed toward the management team of the Rotterdam Study (secretariat. epi@erasmusmc.nl), which has a protocol for approving data requests. Because of restrictions based on privacy regulations and informed consent of the participants, data cannot be made freely available in a public repository.