Physical activity and cognitive function : moment-to-moment and day-to-day associations

This is a self-archived version of an original article. This version may differ from the original in pagination and typographic details. Author(s): Title: Year: Version: Copyright: Rights: Rights url: Please cite the original version: CC BY 4.0 https://creativecommons.org/licenses/by/4.0/ Physical activity and cognitive function : moment-to-moment and day-to-day associations © 2023 the Authors Published version Kekäläinen, Tiia; Luchetti, Martina; Terracciano, Antonio; Gamaldo, Alyssa A.; Mogle, Jacqueline; Lovett, Hephzibah H.; Brown, Justin; Rantalainen, Timo; Sliwinski, Martin J.; Sutin, Angelina R. Kekäläinen, T., Luchetti, M., Terracciano, A., Gamaldo, A. A., Mogle, J., Lovett, H. H., Brown, J., Rantalainen, T., Sliwinski, M. J., & Sutin, A. R. (2023). Physical activity and cognitive function : moment-to-moment and day-to-day associations. International Journal of Behavioral Nutrition and Physical Activity, 20, Article 137. https://doi.org/10.1186/s12966-023-01536-9 2023 Kekäläinenetal. Int J Behav Nutr Phys Act (2023) 20:137 https://doi.org/10.1186/s12966-023-01536-9 RESEARCH Open Access © The Author(s) 2023. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. International Journal of Behavioral Nutrition and Physical Activity Physical activity andcognitive function: moment-to-moment andday-to-day associations Tiia Kekäläinen1* , Martina Luchetti2, Antonio Terracciano3, Alyssa A. Gamaldo4, Jacqueline Mogle5, Hephzibah H. Lovett2, Justin Brown2, Timo Rantalainen1, Martin J. Sliwinski4 and Angelina R. Sutin2 Abstract Background The beneficial effect of acute physical exercise on cognitive performance has been studied in laboratory settings and in long-term longitudinal studies. Less is known about these associations in everyday environment and on a momentary timeframe. This study investigated momentary and daily associations between physical activity and cognitive functioning in the context of everyday life. Methods Middle-aged adults (n = 291, aged 40–70) were asked to wear accelerometers and complete ecological momentary assessments for eight consecutive days. Processing speed and visual memory were assessed three times per day and self-rated evaluations of daily cognition (memory, thinking, and sharpness of mind) were collected each night. The number of minutes spent above the active threshold (active time) and the maximum vector magnitude counts (the highest intensity obtained) before each cognitive test and at a daily level were used as predictors of momentary cognitive performance and nightly subjective cognition. Analyses were done with multilevel linear models. The models were adjusted for temporal and contextual factors, age, sex, education, and race/ethnicity. Results When participants had a more active time or higher intensity than their average level within the 20 or 60 minutes prior to the cognitive test, they performed better on the processing speed task. On days when participants had more active time than their average day, they rated their memory in the evening better. Physical activity was not associated with visual memory or self-rated thinking and sharpness of mind. Conclusions This study provides novel evidence that outside of laboratory settings, even small increases in physical activity boost daily processing speed abilities and self-rated memory. The finding of temporary beneficial effects is consistent with long-term longitudinal research on the cognitive benefits of physical activity. Keywords Cognition, Accelerometry, Ambulatory assessment, Naturalistic setting, Ecological momentary assessment *Correspondence: Tiia Kekäläinen tiia.m.kek[email protected] Full list of author information is available at the end of the article Page 2 of 12 Kekäläinenetal. Int J Behav Nutr Phys Act (2023) 20:137 Introduction There is a consistent positive association between physical activity and cognitive functioning: More physically active people have better cognitive functioning and less cognitive decline with age than less active people [1, 2]. In addition, both structured long-term physical activity interventions [3] and acute exercise sessions [4–6] have positive effects on cognitive functioning; although not all interventions find this effect (e.g., [7]) and the effect may vary depending on the cognitive task [4, 5]. Behavioral, neurophysiological, and neurochemical mechanisms may explain the short- and long-term effects of physical activity on cognition [6, 8]. Some mechanisms, such as increased blood flow in the brain and stimulation of neurotransmitters, are activated immediately during a single bout of physical activity and diminish within hours, while other mechanisms, such as stress relief and improvement in positive mood, may last longer [6, 8]. To date, however, most of what is known about the relation between physical activity and cognitive function comes from laboratory settings or intervention-based studies. Less is known about the benefits of physical activity on cognitive functioning in the real-world, everyday environment. Both physical activity and cognitive performance are dynamic and fluctuate in daily life, even across the course of a single day. Although accelerometers have been available to capture such fluctuations in physical activity, only recent advances in technology have made it possible to embed cognitive tasks within ecological momentary assessments (EMA) to capture cognitive functioning in real-life contexts [9, 10]. The combination of EMA with accelerometer-assessed physical activity provides a unique and robust opportunity to combine these momentary assessments with detailed time-stamped information on physical activity [11]. Analyses of withinperson (intra-individual) associations reveal more nuanced dynamic associations between variables in daily life compared to between-person analyses [12]. Studies applying the EMA concept to investigate within-person effects of physical activity in daily life have shown that when individuals are more physically active than they typically are, they report better mood, more energetic feelings, and lower perceived stress compared to their own average level [11, 13–15]. The within-person association between physical activity and cognitive performance is less studied. Most studies on within-person associations have found physical activity to be related to executive function. A study among adults aged 50–74 (n = 90) found that across 2 weeks, days with greater accelerometer-based physical activity were associated with faster executive function but not verbal learning or recall [16]. The same participants also had lower executive function when they reported currently doing passive activities [17]. A study among adults aged 60+ (n = 51) found no statistically significant within-person association between daily accelerometer-based physical activity and cognitive function assessed on the same day, but previous day physical activity explained the withinperson variance for processing speed [18]. In a study on perceived cognitive ability, college students (n = 128) reported greater perceived cognitive ability on days when they were more physically active than usual [19]. In all these studies, each cognitive test was performed once a day either by smartphones with randomized prompts [16, 17], self-selected time on a web page [19], or at a local day center [18]. Thus, the momentary associations between physical activity and cognition were limited by the single cognitive assessment per day and the analyses focused mainly on day-to-day, not moment-to-moment, variation. Laboratory-based studies suggest that exercise sessions should be at least 20 minutes to have cognitive benefits and that the most significant improvement in cognitive performance is observed approximately 15 minutes after exercise, depending on the intensity [4, 5]. While light intensity exercise can provide immediate benefits, more intense exercise may be necessary to achieve delayed effects [5]. However, the temporal associations in daily life context may be different. Studies examining the momentary associations between physical activity and mood and affective states suggest that engaging in physical activity of any intensity, or even replacing sedentary time with standing, is associated with better mood and feelings of energy on a within-person level [15, 20, 21]. These associations appear to be similar in both 15- and 30-minute epochs [15, 20]. However, the beneficial effect of physical activity on mood in the previous 60 minutes decreases over a three-hour time window [21]. Similar temporal associations may exist between physical activity and cognitive performance. The purpose of the present study was to assess the momentary and daily association between physical activity assessed by accelerometers and cognitive function assessed with momentary performance three times a day and subjective evaluations nightly among middle-aged adults. We posed the following research questions: RQ1. Does physical activity during the preceding 20 or 60 minutes predict cognitive performance? (within-person, momentary-level). RQ2. Does daily physical activity predict self-rated cognition of the day? (within-person, day-level). We hypothesized a similar positive association between physical activity and subsequent cognitive performance in daily life, as previously found in laboratory settings [4, Page 3 of 12 Kekäläinenetal. Int J Behav Nutr Phys Act (2023) 20:137 5]. The selected time frames of 20 and 60 minutes were based on previous findings that the effects of exercise on cognition subside following over 20 minutes delay [5] and that engaging in physical activity is associated with a better mood in a one-hour time window but not in a threehour window [21]. We expected to replicate the finding of better perceived cognition on days with more physical activity than usual observed for younger samples [19] with our sample of middle-aged adults. Methods Participants andtheprocedure The data were from adults in the United States who participated in the Couples Healthy Aging Project (CHAP) (n= 308). Participants were recruited through social media advertisements, community events, and snowball sampling. The inclusion criteria were 1) both members of the couple were aged 40 to 70 years, 2) in a committed relationship for at least 1 year and cohabitating, and 3) both members of the couple were free of severe cognitive impairment (The modified Telephone Interview for Cognitive Status score > 6 [22, 23]) and willing to enroll in the study. All procedures and materials were approved by the Institutional Review Board of the Florida State University (ID: STUDY00000472). Eligible and interested participants were invited to an online meeting in which, after the informed consent process, participants completed a battery of cognitive tests. A study-provided smartphone and accelerometer were delivered to participants, and they were asked to wear the accelerometer and complete the ambulatory assessments for eight consecutive days. The smartphone alerted participants at three semi-random times to complete a brief assessment, including a battery of cognitive tests and a survey about their day each night (Fig.1). The beep windows varied based on reported wake-up times with six possible beep profiles. The morning window varied between 6 am to 12 pm (average 9:14 am), the mid-day window between 11 am and 5 pm (average 2:29 pm), and the afternoon window between 3 pm and 9 pm (average 6:20 pm). The end-of-day survey was beeped between 6 pm and 11 pm (average 9:09 pm). Participants were allowed self-initiation to make up for missed notifications and forced interruptions during a survey. The median time lag between a beep and the start of the response was 0.27 minutes (range 0.02–289.02). Of the 308 participants recruited, 98% (n = 303) had valid data from the EMA portion of the study (n = 5 data were lost due to technical problems with the phone) and 96% (n = 296) wore the accelerometer. The present study includes participants who had information on both EMA and physical activity from at least 1 day (n = 291). The analytic sample did not differ from the rest of the recruited participants in terms of sociodemographic factors, health status, or cognitive functioning (Additional File 1). The data were collected between February 2020 and October 2021. Because of the onset of the COVID-19 pandemic, data collection was temporally suspended and resumed in June 2020. Assessments and procedures were modified to allow participants to complete all study components remotely. Measures Ambulatory cognition Participants completed two cognitive tests validated for ambulatory assessment: the Symbol Search Task (SST) assessed processing speed and the Dot Memory test (DMT) assessed visual memory [24]. These tasks assess two fundamental and distinct cognitive functions [24]. In the SST, participants match symbol pairs as quickly as possible. Each test session comprised 12 trials and the mean response time of correct trials for each session was calculated. In the DMT, participants saw three dots in a 5 × 5 checkerboard for 3 seconds, and after an 8-second filler task, they were asked to indicate the location of the red dots. The average distance between the correct and the indicated location of the dots across each trial (two trials for each test session) was used in the analyses. In both cognitive tests, a larger value (slower reaction time, more errors) indicated worse performance. Fig. 1 The schedule for Ecological Momentary Assessments Page 4 of 12 Kekäläinenetal. Int J Behav Nutr Phys Act (2023) 20:137 Self‑rated cognition In each night survey, participants rated whether their mind was as sharp, their memory as good, and their thinking as fast as usual today [25, 26]. The response scale was from 0 to 100, with a higher value indicating better cognition. Physical activity ActiGraph (ActiGraph Corp., Pensacola, FL) wrist-worn tri-axial accelerometers were used to measure physical activity for 8 days. Participants were instructed to wear the device on their wrists for 24 hours per day for the same eight consecutive days that they took part in the ambulatory assessment. They were instructed to take the device off only for showers and water-based activities. ActiGraph data were analyzed using the ActiLife (Acti- Graph Manufacturing Technology Inc., FL) Software. A non-wear time was defined as 90 minutes of continuous zero vector magnitude counts (VMC) [27, 28]. Days with at least 10 hours of wear time were included in the analyses [29]. Wear time was full 24 hours for 83% of days. The data were analyzed in 60-s epochs and divided into sedentary and active time based on the cut-points of < 2303 VMC per minute (cpm) for the dominant wrist and < 1853 cpm for the non-dominant wrist [30]. Using the time stamps from the smartphones and accelerometers, the timing of EMA prompts was synchronized with the accelerometer data. Custom-written MATLAB (version R2019b, The MathWorks Inc., Natick MA, USA) scripts were used to extract the specific period (20 and 60 minutes) before each EMA assessment and the time between the last activity minute and the EMA assessment. Active time (minutes) and maximum counts were used as indicators of physical activity in the present study. Active time is an indicator of time spent other than sedentary activities, whereas maximum counts indicate the maximum intensity of physical activity reached. VMC and steps were reported for descriptive purposes. Demographics Age in years, sex (0 = male, 1 = female), race/ethnicity (0 = white, 1 = person of color), and education in (years) were asked in the main interview. Statistical analyses Statistical analyses were performed with IBM SPSS Statistics Version 28.0.1.1. (IBM Corp. in Armonk, NY) and R Version 4.2.1. (R Foundation for Statistical Computing, Vienna, Austria). Data were prepared for analysis and check for quality following recommendations [31]. Means, standard deviations, frequencies, and correlations were used for descriptive purposes. Within-person correlations were calculated using the R package misty [32]. The data were analyzed with multilevel models to account for the hierarchical structure of the data (days and moments nested within individuals). Level 1 repeated assessments of physical activity were personmean centered (i.e., each momentary value of physical activity minus the mean of physical activity across assessments; 0 represents the within-person mean for each participant) to estimate when participants were more or less physically active than their average. Level 2 between-person variables were grand-mean centered (i.e., person overall mean minus grand mean; 0 represents the mean for all participants). A third level (cognitive assessments nested within participants within couples) was also tested to account for participants recruited in pairs. The variance explained by between-couples was not significant after accounting for between-person variables (age, sex, education, race/ ethnicity) and thus, the results for the two-level analyses are reported. For each cognitive outcome, a null model without any predictors was estimated to separate the within- and between-person variance using intraclass correlation coefficients (ICCs). Next, between-person and withinperson predictors of interest were included in the model. The models were adjusted for between-person covariates of age (grand-mean centered), education (grand-mean centered), sex, and race/ethnicity. Temporal covariates included in the models were weekday (weekend = 0, weekday = 1) to account for weekly rhythm, day in the study (range 1–8) to account for practice effects, and time window (1 = morning, 2 = mid-day, 3 = afternoon) to account for time-dependent variation within days. Contextual covariates were location (0 = home, 1 = other) and company (0 = alone, 1 = presence of other person) at the time of the assessment to account for possible distractions. For day-level analysis, only the first two temporal covariates were included. Additionally, the accelerometer wear time for each day was included in the models assessing the associations between active time and self-rated cognition. Models were run separately for active time and maximum counts, and physical activity extracted 20 or 60 minutes before each EMA session. The Variance Components (VC) random covariance matrix was used in the analyses. The data were analyzed with restricted maximum likelihood (REML) estimation using all available data to estimate the model parameters (Additional File 1). Three sets of additional analyses were conducted and presented in Additional File 3. To account for betweenperson differences, the grand-mean centered physical Page 5 of 12 Kekäläinenetal. Int J Behav Nutr Phys Act (2023) 20:137 activity (i.e., person overall mean minus grand mean; 0 represents the mean for all participants) was included in the models. To account for the time lag between the latest physically active minute and the EMA assessment, a supplementary analysis was performed for cases having at least one physically active minute in the 20- (78.6% of cases) or 60-minute (94.2% of cases) epoch preceding the EMA assessment. To account for the response delay, a supplementary analysis was performed by excluding cases that took more than 15 minutes to respond to the prompt [33]. Results Descriptive statistics The participants’ characteristics are shown in Table1 for participants (N = 291) with at least 1 day of accelerometer data from the EMA days. Of these participants, all had information on self-rated cognition from at least 2 days (93% of participants had data from 6 or more days) and all except one on cognitive tests from at least seven sessions (82% of participants completed 20 or more sessions). The bi-variate correlations between study variables are in Supplementary TableS1 (Additional File 3). Across all participants over the 8 days, there were 6221 EMA assessments of processing speed and corresponding physical activity, 6045 EMA assessments of visual memory and corresponding physical activity, and 2012 nightly assessments of self-rated cognition and physical activity on the same day. Multilevel models The ICCs suggest that 65% of the total variance in processing speed was between participants and 35% was within people; corresponding values for visual memory were 27 and 73%, respectively. For the self-reported outcomes, 59% of the variance in memory and 58% of the variance in thinking and sharpness of mind were attributable to between-person and 41 and 42% within-person, respectively. The results from multilevel models are shown in Tables2, 3 and 4. At the momentary-level (RQ1), participants performed better on the processing speed task when they were more physically active than their usual (Table2). The same association was apparent in all four models including either maximum counts or activity minutes as a predictor and either 20 or 60 minutes before the cognitive assessment. For example, each one-minute increase in physical activity during the 20-min period before cognitive assessments was associated with 3.11 milliseconds faster processing speed (B = -3.11, SE = 0.70, p < .0.001) and every 1000 counts increase in maximum physical activity intensity during the 20-min period before cognitive assessments was associated with 0.5 milliseconds faster processing speed (B = -0.50, SE = 0.13, p < .0.001). Supplementary materials include models with between-person physical activity variables (Table S2). The within-person associations between physical activity and processing speed remained consistent after accounting for between-person differences in physical activity level. At the between-person level, participants with higher physical activity intensity (maximum counts) had faster processing speed (Table S2). The time lag between the latest physically active minute and cognitive test was not statistically significantly associated with processing speed (Table S4). The within-person association between physical activity and processing speed remained consistent after excluding cases with 15 minutes or longer response delay (Table S6). Physical activity was unrelated to visual memory at either the between-person or within-person level (Table3, Table 1 Descriptive statistics for the sample (n = 291) VMC Vector Magnitude Counts, SST The Symbol Search Task, mean response time, DMT The Dot Memory test, error mean (Euclidean distance), adata from 60 mins period before each EMA session, bdata from 20 mins period before each EMA session M/N SD/% Sex Women, % 159 54.6 Men, % 132 45.4 Race/ethnicity White, % 215 73.9 Other, % 76 26.1 Age, years 51.9 7.4 Education, years 16.6 3.3 Complete EMA night sessions 7.4 1.2 Complete EMA day sessions 21.7 2.9 Day level Self-rated sharpness of mind 71.7 16.5 Self-rated memory 71.9 16.2 Self-rated thinking 72.0 16.3 Wear time per day, hours 22.9 0.9 Active time per day, mins 354.7 104.3 Maximum counts daily average, 10312.7 3.2 VMC per day, 1031980.5 614.9 Steps per day 4077.9 2312.6 EMA level SST, ms 1621.1 443.9 DMT, Euclidean distance 1.5 0.7 Active time 60mina, mins 24.5 15.5 Active time 20minb, mins 8.4 6.3 Maximum counts 60mina, 10313.7 9.2 Maximum counts 20minb, 1034.7 3.5 Page 6 of 12 Kekäläinenetal. Int J Behav Nutr Phys Act (2023) 20:137 Table 2 Associations between physical activity and processing speed analyzed by multilevel modelling (n = 6221) a Grand-mean centered, bPerson-mean centered, Reference categories female (sex), weekday (weekend), white (Race/ethnicity), home (location), with others (company) Model: Max Counts 60 min Model: Max Counts 20 min Model: Active time 60 min Model: Active time 20 min B SE p B SE p B SE p B SE p Fixed effects Intercept 1849.93 54.71 <.001 1843.76 54.61 <.001 1823.62 54.48 <.001 1825.36 54.47 <.001 EMA day −27.49 1.78 <.001 −27.44 1.78 <.001 −27.52 1.78 <.001 −27.46 1.78 <.001 EMA session number 4.28 4.87 .380 4.00 4.87 .412 5.10 4.88 .296 4.64 4.87 .341 Weekend −8.78 9.43 .352 −9.35 9.43 .321 −8.70 9.43 .356 −9.04 9.43 .338 Company −23.27 9.05 .010 −23.86 9.06 .008 −24.36 9.06 .007 −24.70 9.07 .006 Location −21.83 9.30 .019 −24.31 9.30 .009 −21.01 9.31 .024 −23.36 9.29 .012 Agea23.48 3.22 <.001 23.56 3.22 <.001 23.66 3.22 <.001 23.65 3.22 <.001 Educationa−1.67 7.06 .813 −1.61 7.06 .820 −1.57 7.07 .824 −1.57 7.07 .824 Sex 123.86 47.20 .009 124.58 47.23 .009 127.00 47.27 .008 127.09 47.27 .008 Race/ethnicity 176.92 54.05 .001 176.81 54.09 .001 178.12 54.14 .001 178.02 54.14 .001 Physical activityb−.20 .05 <.001 −.50 .13 <.001 −1.26 .30 <.001 −3.11 .70 <.001 Variance components Estimate SE Wald Z p Estimate SE Wald Z p Estimate SE Wald Z p Estimate SE Wald Z p Residual 95,534.17 1764.47 54.14 <.001 95,531.31 1764.41 54.14 <.001 9549.35 1763.66 54.14 <.001 99,671.15 1841.34 54.13 <.001 Random intercept 149,485.96 12,793.43 11.68 <.001 149,708.39 12,811.99 11.69 <.001 149,977.38 12,834.40 11.69 <.001 15,128.72 12,959.85 11.67 <.001 Page 7 of 12 Kekäläinenetal. Int J Behav Nutr Phys Act (2023) 20:137 Table 3 Associations between physical activity and visual memory analyzed by multilevel modelling (n = 6045) a Grand-mean centered, bPerson-mean centered, Reference categories female (sex), weekday (weekend), white (Race/ethnicity), home (location), with others (company) Model: Max Counts 60 min Model: Max Counts 20 min Model: Active time 60 min Model: Active time 20 min B SE p B SE p B SE p B SE p Fixed effects Intercept 1.94 .11 <.001 1.95 .11 <.001 1.94 .11 <.001 1.94 .11 <.001 EMA day −.05 .01 <.001 −.05 .01 <.001 −.05 .01 <.001 −.05 .01 <.001 EMA session .08 .02 <.001 .08 .02 <.001 .08 .02 <.001 .08 .02 <.001 Weekend .00 .04 .943 .00 .04 .951 .00 .04 .950 .00 .04 .965 Company −.05 .03 .129 −.05 .03 .119 −.05 .03 .124 −.05 .03 .107 Location .03 .03 .422 .03 .03 .431 .03 .03 .413 .03 .03 .423 Agea.03 .01 <.001 .03 .01 <.001 .03 .01 <.001 .03 .01 <.001 Educationa−.03 .01 .015 −.03 .01 .015 −.03 .01 .015 −.03 .01 .015 Sex −.35 .08 <.001 −.35 .08 <.001 −.35 .08 <.001 −.35 .08 <.001 Race/ethnicity .33 .09 <.001 .33 .09 <.001 .33 .09 <.001 .33 .09 <.001 Physical activityb.00 .00 .972 .00 .00 .535 .00 .00 .757 .00 .00 .233 Variance components Estimate SE Wald Z p Estimate SE Wald Z p Estimate SE Wald Z p Estimate SE Wald Z p Residual 1.30 .02 53.29 <.001 1.30 .02 53.29 <.001 1.30 .02 53.29 <.001 1.30 .02 53.29 <.001 Random intercept .40 .04 1.17 <.001 .40 .04 1.17 <.001 .40 .04 1.17 <.001 .40 .04 1.17 <.001 Page 8 of 12 Kekäläinenetal. Int J Behav Nutr Phys Act (2023) 20:137 Table 4 Associations between daily physical activity and self-rated cognition analyzed by multilevel modelling (n = 2021) a Grand-mean centered, bPerson-mean centered day-level variable. *All Wald Z -tests p < 0.001. Reference categories female (sex), weekday (weekend), white (race/ethnicity) Memory Thinking Sharpness of mind Model 1: Max counts Model 2: Active time Model 1: Max counts Model 2: Active time Model 1: Max counts Model 2: Active time B SE p B SE p B SE p B SE p B SE p B SE p Fixed effects Intercept 66.03 1.55 <.001 71.84 3.13 <.001 65.08 1.56 <.001 66.56 3.15 <.001 64.61 1.57 <.001 68.38 3.21 <.001 Day .53 .14 <.001 .56 .16 <.001 .85 .14 <.001 .88 .16 <.001 .71 .14 <.001 .70 .16 <.001 Weekend 2.27 .67 .001 2.25 .68 .001 1.37 .67 .040 1.42 .69 .038 1.77 .68 .010 1.70 .70 .016 Agea.16 .13 .228 .16 .13 .217 .11 .13 .405 .11 .13 .402 .15 .13 .249 .16 .13 .241 Educationa.11 .29 .700 .11 .29 .706 .13 .29 .646 .13 .29 .649 .22 .29 .446 .22 .29 .448 Sex 3.86 1.92 .046 3.84 1.92 .047 4.68 1.94 .016 4.68 1.94 .017 5.27 1.94 .007 5.26 1.94 .007 Race/ethnicity 4.83 2.19 .028 4.76 2.19 .030 3.18 2.21 .151 3.17 2.21 .153 4.56 2.22 .041 4.52 2.21 .042 Physical activityb.01 .08 .889 .01 .00 .007 −.03 .08 .663 .00 .00 .810 −.05 .08 .502 .01 .00 .088 Wear time .00 .00 .050 .00 .00 .593 .00 .00 .231 Variance components Est. SE Wald Z* Est. SE Wald Z* Est. SE Wald Z* Est. SE Wald Z* Est. SE Wald Z* Est. SE Wald Z* Residual 162.47 5.56 29.22 161.92 5.54 29.23 163.29 5.59 29.22 163.39 5.59 29.23 171.50 5.87 29.22 171.40 5.86 29.23 Random intercept 232.98 21.33 1.92 232.60 21.29 1.93 237.68 21.73 1.94 237.46 21.71 1.94 237.18 21.79 1.89 236.68 21.74 1.89