Data collection
This study was a cross-sectional study, which is a type of observational epidemiological study. The study protocol was approved by the ethics committee of the University of Tsukuba School of Medicine (approval No. 1065). Written informed consent was obtained from all participants.
Data collection for the SLEPT Study (SLeep Epidemiology Project at the University of Tsukuba) was conducted between August 2016 and November 2017 at 4 workplaces—a national university, a national institute, an institute of a company at the Ibaraki Prefecture located near Tokyo, and a health care company in Tokyo—as well as from some workers introduced by the staff or study participants. Participants were enrolled only through open recruitment using flyers, posters, workplace group e-mails, and online workplace bulletin boards. There was no individual solicitation in principle, and we did not ask the participants about sleep status prior to enrollment.
Participants
In this study, 785 individuals were recruited. Among them, 4 participants withdrew their consent, and 110 participants were excluded as they did not fulfill the study criteria [e.g., lack of data on sleep measurements, Pittsburgh Sleep Quality Index (PSQI), or questionnaire responses related to housing or bedroom] or selected “others” for “type of housing” (n = 13). In total, 671 participants (298 men and 373 women; mean age ± standard deviation: 43.3 ± 11.2 years, range 22–68 years) who completed the sleep measurements, responded to all relevant questions in the questionnaire, and did not withdraw consent were included in the analysis.
Questionnaire
The participants were requested to complete a self-administrated questionnaire including questions about type of housing (apartment, detached house, or others); style of bedroom (western/Japanese); use of wood, for example, structure of housing (reinforced concrete, steel‐frame, or wood), amount of wood used for interiors, furniture, and door(s) in the bedroom (“How much wood is used in your bedroom, including interiors, furniture, and door(s)?”: 1: large, 2: rather, 3: rare, 4: not at all), use of wood on floor (yes/no), walls (yes/no), and ceiling (yes/no) of the bedroom; uncomfortableness of noises including snoring in the bedroom (noisy/quiet); health-related parameters such as height, weight, and lifestyle factors such as smoking status, nightcap (defined as drinking within 2 h before sleeping once a week or more), and habitual exercise.
The questionnaire also included items regarding comfort in the bedroom (“Do you feel mental comfort or serenity in the bedroom?”; 1: very, 2: rather, 3: neutral, 4: rather not, 5: not at all) and PSQI, which is one of the most frequently used indices for evaluating self-rated sleep quality in sleep medicine [12, 13]. PSQI, which can assess sleep quality and disturbances over 1 month, consists of the following seven components: subjective sleep quality, sleep latency, sleep duration, habitual sleep efficiency, sleep disturbance, use of sleeping medication, and daytime dysfunction [12]. The Athens Insomnia Scale (AIS) was developed based on ICD-10 (10th revision of the International Statistical Classification of Diseases and Related Health Problems) by the World Health Organization [14,15,16]. AIS consists of eight items; for example, one of the eight items is awakenings during the night, which can be graded as 0: no problem, 1: minor problem, 2: considerable problem, or 3: serious problem or did not sleep at all [14]. AIS was added to the questionnaire of the national university from June 2017. The available data for AIS were obtained from 530 participants.
Sleep measurements
Participants were required to wear an actigraphy device on the waist (MTI-210, ACOS Co., Ltd., Nagano, Japan) for 24 h a day, except during bath time, for a week to estimate actual sleep time and sleep efficiency and to record their sleeping and waking times in a sleep diary [17, 18]. The actigraphy data were analyzed using Sleep Sign Act software (Kissei Comtec Company Inc., Matsumoto City, Japan) [18]. We included only the longest sleep time during 24 h (from noon to noon on the next day) but not naps, because it seemed to be highly probable that only the longest sleep time would be spent in the bedroom.
Statistical analysis
The four categories for the amount of wood used in the bedroom interior were modified into three categories in this study by combining the two lower categories (“rare” and “not at all”) into one category, “not used”, because the number of participants who responded “not at all” was only 16. Comfort was defined by the two higher categories, “very” and “rather”. Poor sleep quality was defined as PSQI ≥ 6 [12, 13]. Suspicion of insomnia was defined as AIS ≥ 6 [15]. As no appropriate actigraphy-based cut-off point of sleep efficiency is known at present, we defined low sleep efficiency as sleep efficiency of < 70% based on the available actigraphy data.
Overweight was defined as body mass index (BMI) ≥ 25.0 kg/m2. BMI was calculated based on self-assessed height and weight. Exercise habit was defined as exercise activities performed for 30 min or more per session, twice a week or more, and continued for at least 1 year, as same as that defined by National Health and Nutrition Survey by conducted by Ministry of Health, Labour and Welfare, Japan [19].
The percentage difference between the groups was calculated using the Chi-square test. The associations between the ordinal variables were analyzed using the Mantel–Haenszel test for trend.
Since the participants had several different backgrounds and circumstances, multivariate analysis was conducted to adjust for various factors. Two multivariate analyses are commonly used in epidemiology. One is multiple regression analysis, which is performed when the dependent variable is a continuous variable. The other is logistic regression analysis, which is performed when the dependent variable is a binary one, such as presence or absence of a disease. This study adopted several binary variables as outcomes, e.g., presence or absence of the suspicion of insomnia. Therefore, logistic regression analysis was extensively used.
We used two models in the logistic regression analysis. The results of logistic regression analysis might differ according to the set independent variables; therefore, more than one analysis model, which have different independent variables, are often used. Model 1 was adjusted by considering possible confounders without wood-related factors. Model 2 was adjusted by considering possible confounders with the structure of housing for targeting only the wood use in the bedroom.
To assess the relevant factors related to comfort in the bedroom, in Model 1 the dependent variable was comfort in the bedroom (yes/no), whereas the independent variables were sex, age (5 categories: 20, 30, 40, 50, and 60s), type of housing (apartment/detached house), age of the building (4 categories: < 10 years, 10–19 years, 20–29 years, and ≥ 30 years), style of the bedroom, area of the bedroom (3 categories: < 6 Jo, 6–8 Jo, and ≥ 8.5 Jo), noise in bedroom (noisy/quiet), and each one of the wood-related factors in housing or bedrooms (5 factors): structure of housing, wood floor, wood wall, wood ceiling, and amount of wood used in the bedroom interior (3 categories). “Jo” indicates the unit of Tatami (approximately 1800 mm × 900 mm). In Model 2, the independent variable, structure of housing (3 categories), was added to the independent variables in Model 1, for all calculations.
To evaluate the relevant factors of sleep conditions, characteristics of participants were additionally adjusted in logistic regression analysis. The dependent variable was each one of the three sleep-related factors such as poor sleep quality (PSQI ≥ 6), suspicion of insomnia (AIS ≥ 6), and low sleep efficiency (< 70%), whereas the independent variables were sex, age (5 categories), BMI, habitual exercise (yes/no), smoking (current smokers/other responses), nightcap (yes/no), shift worker (current shift worker/other responses), type of housing, age of building, style of the bedroom, area of the bedroom, structure of housing, noise in the bedroom, and each one of the four wood-related factors in bedrooms; wood floor, wood wall, wood ceiling, and amount of wood used in the bedroom interior.
The data set used was version 20191028. The significance level was set at 5%. IBM SPSS Statistics version 23 for Windows (IBM, Armonk, NY, USA) was used for the statistical analysis.