OBSERVATIONAL STUDY OF ENVIRONMENTAL FACTORS THAT AFFECT BODY MOVEMENT DURING NORMAL SLEEP

. In this observational study, body movements during sleep were evaluated. Measurements were recorded in the subject's bedroom under the normal conditions. Healthy 8 males and 8 females aged 19-56 years old participated in the experiment. Data for 20 to 30 nights were obtained for each subject and the analysis was conducted using 325 valid data. As a result, there observed significant differences among subjects for body movement and environmental factors. Concerning body movements, the results showed significant differences in gender, age group, group of The General Health Questionnaire's value, and alcohol intake. The regression line between body movement and relative humidity has a negative slope, and 12 out of 15 subjects showed a positive correlation between radiation temperature and the frequency of the body movement. For relative humidity, 3 subjects showed significant correlations with negative slopes. The factors could be categorized into two groups: those based on individual differences and those based on the environment. It was also affected by the factors that varied from day to day. The degrees of mutual influence of these factors on body movement were examined. The results showed that the relative humidity had the highest standardized coefficient for overall sleep duration, followed by gender, radiant temperature of celling, and drinking habit.


Introduction
It has been pointed out that body movement may be an indicator of sleep stability or persistence [1]. In order to know factors influencing body movement, is necessary to examine optimal body movement conditions. Akiyama et al. [2] reported that the percentage of mid-wake, induced by rolling over when a subject slept in a thermoneutral zone, was 50%, while the percentage of mid-wake increased in case an environmental condition was out of a thermoneutral zone: the slightly hotter side of the environment to be rather acceptable. Imai et al. [3] showed that, 25°C and 28°C room temperatures were comfortable, and that the body movements increased at 30°C and 80% humidity. Morito et al. [4] showed that the number of body movements was significantly higher for 1 m/s airflow than that for 0.3 m/s airflow in the test room. These results indicate that the bedroom environment influences body movement, however they are not intended to elucidate the factors that influence body movement, and the effects of interrelationships and other conditions have not been examined in detail.
In a previous report [5], we reported that the frequency of body movements in the early stages of sleep in summer is positively and significantly correlated with radiant heat of the nearest wall, but the measurements were taken in a rental apartment room * Corresponding author: ota@j-michishirube.net and it has not been confirmed whether our findings could be also obtained in a normal bedroom.
In this report, we evaluate fluctuations of body movement during night sleep to identify the factors that cause body movement. The experiments were conducted in each subject's bedroom for the purpose of evaluating normal sleep. A mat sensor, which ensured non-contact and non-restrain, was used to measure body movement in order not to interrupt his/her normal sleep.

Summary
The experiment was conducted in each subject's bedroom from May 2019 to December 2020. The subjects were healthy 8 males and 8 females aged 19 to 56 years without sleep disorders. The results obtained from male subjects are denoted with A* (*: 1-8), while those from female are denoted with B* (*: 1-8). Each subject participated in the 20 -30 nights experiment. Data for a total of 325 nights were obtained without missing values. Table 1 shows the measured physiological and environmental parameters. We used a mat-type sensor ('aams' manufactured by biosilver) for the measurement of the physiological data. The sensor was placed on the bed under the torso of the subject. Fig.  1 shows a plan view of the measurement instruments in A1's room. The numbers indicate the place where the measurement equipment were placed.
Subjects were asked to answer the General Health Questionnaire (Japanese version, GHQ28) prior to the experiment. They were also given a questionnaire about their daily activities before going to bed and also asked to report their actual wake up time.

Method of calculating analysis factor
The mean and standard deviations for each night were calculated from the body movement and environmental factor data.
For body movement, data were recorded every 0.5 seconds. It was checked with a time interval of one minute if there was more than a single body movement or not; if so, it was considered that the body moved in that specific time interval. The value obtained by dividing the total number of minutes with body movement by the total sleep time was defined as the number of body movements (BM). BM is a dimensionless quantity because it is a percentage measure.
For environmental factors Tp, Hc and CO2, the average value (suffixed with ", Ave") were calculated.
The start of sleep was determined to be the time when no body movement was detected for continuous 5 minutes.

BM Ave / Environmental factors
The distributions of BM, Ave environmental factors Tp ,Ave, Hc ,Ave, and CO2 ,Ave for each subject are shown in     Fig.3. Fig. 2 shows that the mean value of BM ,Ave tends to be larger for male subjects, except for B7, For this subject, the effective data were collected for only 3 nights. Fig. 3 shows that Tp ,Ave varies among each bedroom; Hc ,Ave is between 40-70% with no extreme value, CO2 concentration levels were more than twice that of the external environment, except for A3.
The Kruskal-Wallis test was used to confirm the differences between bedrooms for each factor, and the differences between groups are shown in the upper part of Figs. 2 and 3. Significantly pairs were shown by " .x" . The BM ,Ave values of A5, A6, A7,A8 and B8 were larger, which indicate that the subjects can be categorized into groups based on the size of their BM ,Ave. In terms of Tp ,Ave, we can confirm that the distributions of Tp ,Ave were different among the bedrooms: higher Tp ,Ave for A3, A5, B3 and B8 and lower for B2 and B6. These differences might be due to seasonal temperature changes and insulation performance. In terms of Hc , Ave, B6 and A1 had the highest relative humidity, A7 and A8 have significantly lower relative humidity. In terms of CO2 ,Ave, A7, B3, and B6 had high values, which were significantly different from the other bedrooms, indicating the ventilation conditions were different among bedrooms. Fig. 4 shows the scatter plots of the mean values between BM ,Ave and environmental factors. Fig. 5 shows the scatter plots of the deviation from the mean value of each subject. Table 2 shows the slopes and coefficient of determination for each subject, obtained by a linear regression to the results. In Fig. 4, BM ,Ave and Hc , Ave have a significant negative correlation of P < 0.01. In Fig. 5 and Table 2 Based on the findings above, it is necessary to consider separately that the steady-state effects represented by the mean influence of the environmental factors and its deviation on BM , Ave.       Table 3 shows a result of Mann-Whitney U test about the difference of BM ,Ave on each item of a questionnaire for individual attributes and their daily lifestyle before going to bed. Subjects were categorized into two groups depending on the ages: under 20 and over 40. Score of GHQ were divided into two groups: over 7 or not, as most of the subjects are university students [7]. BMI values were divided into two: under 20 or not, because all subjects are inside the range of 18.5 to 25 which is considered to be the normal range of BMI as defined by Japan Society for the Study of Obesity [8]. Table 4 shows the data categorized into two groups depending on 1) whether the subject had alcohol, 2) whether the subject took a bath, 3) whether the subject used a fan, 4) whether the amount of the clothing of a subject were over 1.0 clo, and 5) whether the comforter was heavy. The items which had a high r value and significant difference are: Gender, Age, GHQ. BM ,Ave is higher for males than females and for younger age group. GHQ were higher in the case of high BM , Ave. It indicates that unhealthy mental states may cause higher BM ,Ave. The low r value for BMI is thought to be due to a normal BMI. Alcohol intake has a significant difference: higher BM ,Ave were obtained when a subject had alcohol. As for bathing, BM ,Ave are lower when taking a bath. This is considered to be due to the effects of bathing in promoting the lowering of core body temperature.

BM ,Ave / individual differences and lifestyle
Although Morito et al. [4], showed an airflow affects the body movement, the use of a fan was not significant. As there was no clear instruction for the usage of a fan it might have affected the result.

Analytical method
In the analysis, two factors will be examined separately: a factor of daily change and a factor of each subject. For environmental factors, the measured values are divided averages and deviations from the average. Intraclass correlations were determined for the similarity of environmental factors for each bedroom. In terms of environmental factors, the intraclass correlations are over 0.1 and significant (p < 0.01). Moreover the design effect values are over 2.0. Therefore, the data structure was assumed to be multilevel. Considering the response variable is BM, Ave and explanatory variables are environmental factors, the degree of influence for item of individual attributes and daily lifestyle is examined by standardization coefficients in the structural equation modelling. In Tables 3 and 4, partial correlations were not considered, so items that were not significant were also re-examined as explanatory variables. Gender, Age, GHQ, BMI, drinking habits and average of environmental factor in bedroom are treated as Between Level. Alcohol intake, bathing, use of fan, the amount of clothing and comforter, and deviations from the average of environmental factors in bedroom are treated as Within Level. Since the distribution of values obtained from the experiment is nonparametric, Bayesian estimation was performed using MCMC methods. Mplus Version 8.4 was used for the analysis, with 8 Markov chains and 40,000 MCMC simulations. Table 5 shows standardized coefficients and p-values for each explanatory variable for the model with the best fit. The posterior predictive p-value for this model is 0.224. The significant factors of Within Level are Tp ,Ave and the alcohol intake. High radiation temperature or drinking habit seems to influence BM ,Ave to be higher. The significant factors of Between Level are Hc ,Ave and Gender. Male subjects or having lower average of relative humidity in bedroom seems to influence BM , Ave to be higher. This could be due to the overall range of Hc , Ave in Fig. 4 is low.

Results/ Discussion
From the above, the standardized coefficient was highest at -0.567 for the relative humidity environment, however daily relative humidity changes were not significant, and the standardized coefficient was not large. This would increase the body movement due to the effects of daily dry conditions in a low-relative humidity bedroom environment. Standardized coefficient of daily change factors is highest at 0.133 for the radiation temperature. Daily radiant temperature of celling variations were over 3 degrees for the most of the room conditions. The fluctuation of the temperature of this range could well occur on a daily basis in a typical bedroom environment and is considered to be more influential than other environmental factors.
Standardized coefficient of alcohol intake is -0.121. BM ,Ave was recorded when alcohol was not taken. This is likely to be affected by the amount of alcohol consumed and should be handled with caution.

Conclusion
The observational study was conducted on subjects during sleep to determine the degree of influence between factors that affect the body movement during sleep. The findings are listed below. -Environmental factors determined by the subject's bedroom characteristics should be considered separately from those determined by daily change. -Daily change of the radiant temperature of celling significantly affected the body movement. The higher the change, the more body movement. -Alcohol intake affected significantly. Sleep after drinking reduced body movement. -Difference of relative humidity is significant and has the highest standardized coefficient: the more body movement in lower relative humidity. -Differences in gender were observed: for male subjects relatively higher frequency in body movement.
In this study, it was found that there are factors that have different effects on body movement depending on the bedroom environment.
As a future work, it is necessary to increase the sample size of this observational study, whether there remain other factors. In particular, it is essential to unveil the effects of the age of the subjects. In this research many of the subjects were relatively young, and the characteristics of middle-aged and elderly subjects are not yet well known. In this study, the number of body movements was considered to be linear, but in reality, it would be more appropriate to consider the ideal body movement during sleep: that the number of body movements is unlikely to be zero and would converges to some optimal number. This optimal number of body movements may also vary depending on the physical condition of the day, and future experiments should be conducted taking these factors into account.