Aggregated Data (not individual level data) were derived from the Comprehensive Survey of Living Conditions in Japan for 2019 . The survey is conducted by the Ministry of Health, Labour and Welfare on a yearly basis to obtain data on income status, savings, and households in Japan. In addition, health status is surveyed every 3 years. The subjects fill out self-reported questionnaires regarding their households. A total of 5530 districts (approximately 300,000 households) throughout Japan were selected through stratified random sampling, and all households in the selected districts (approximately 720,000 persons) were investigated in the survey . The inclusion criteria are all persons living in households in selected districts are surveyed. Those who work away from their families, migrant workers, long-term business travelers, those who study away from their home, residents of social welfare facilities, long-term inpatients, foster children left by their parents, inmates, and those who are separated from their households were excluded in the survey . In 2019, a total of 301,334 households became subjects for all forms of status, and the data of 218,332 households were gathered in the Ministry of Health, Labour, and Welfare . The responses from 217,179 households were finally aggregated in the data after removing responses that cannot be aggregated . The exact number of respondents for each question in the questionnaire was unknown. In addition, only those who were not hospitalized were included in the publicly available data. The estimated total number of persons in all of Japan corresponding to each response option of the questions in the survey was calculated using the prevalence of respondents for each response option and the Japanese population by the Ministry of Health, Labour, and Welfare, and this data are publicly available . This study used the data on the status of psychological distress, self-rated health, smoking status, alcohol drinking frequency, and status of participation for cancer screening. The flowchart of selecting study subjects is shown in Fig. 1.
Data on educational level were provided as elementary school or junior high school, high school, vocational school, junior college or technical college, university, and graduate school. For the study, it was classified into three levels, namely, low (elementary school or junior high school), middle (high school and vocational school), and high (junior college or technical college, university, and graduate school) as it was done in a previous study . Subjects with unknown educational level were not included in the analysis.
Psychological distress was assessed based on the scores of Kessler’s Psychological Distress Scale (K6), and subjects whose total score was ≥5 were classified as psychologically distressed. Regarding self-rated health, it was shown to be a valid indicator of overall health and an indicator of mortality in Japan . In the survey, self-rated health was assessed by one question: “What is your current health status?” It was divided into two statuses, namely, good (very good, good, and normal) and poor (bad and very bad). Smoking status was classified into two, namely, smoker (i.e., smoking every day or sometimes) and non-smoker (i.e., former smoker and non-smoker). Regarding alcohol drinking frequency, the question was “How many days do you drink alcoholic beverages in a week?” The response options were “every day,” “5–6 days per week,” “3–4 days per week,” “1–2 days per week,” “1–3 days per month,” “merely drink,” “stop drinking,” and “not drinking.” Therefore, drinking status was classified into two, namely, drinker (i.e., drinking every day, 5–6 days per week, 3–4 days per week, 1–2 days per week, or 1–3 days per month) and non-drinker (i.e., merely drink, stop drinking, or not drinking).
Regarding cancer screening, the data of participation status for stomach, lung, colorectal, breast, and uterine cancer were available. According to the guideline for implementation of cancer screening in Japan , stomach cancer screening is recommended once every 2 years for persons aged ≥50 years. Lung and colorectal cancer screening is recommended once every year for persons aged ≥40 years. Breast cancer screening is recommended once every 2 years for persons aged ≥40 years, and uterine cancer screening is recommended once every 2 years for persons aged ≥20 years. Therefore, the participation status data with the recommended frequency for each cancer type were used. In addition, the data of patients aged ≥50 years for stomach cancer, aged ≥40 for lung, colorectal, and breast cancer, and ages of 25 or above for uterine cancer were used.
Age groups in 5-year increments from 20 to 24 to 80–84 and the age group of over 84 years were available, and groups of 25–29 years or more were used because it is considered that many of the highly educated people had not graduated their university or graduate school yet by the age of 20–24 years. Therefore, the estimated number for each educational level does not contain those who are currently attending the corresponding educational level, and many people who are attending university or graduate school are not reflected in the data of the age group of 20–24. Prevalence of each health-related behaviors and statuses were calculated by age group, sex, and educational level.
In addition, age-standardized prevalence was calculated by sex and educational level using the sum of the estimated number of all the choices for each health-related behavior and status (estimated number of household persons in Japan) as the standard population. Specifically, we calculated number of household persons in Japan according to age group after removing number of persons whose responses (choices) for each outcome type were unknown. By multiplying the estimated number of household persons in Japan by the prevalence of outcomes for each age group and summing them by sex and educational level, we derived the expected number of persons with each health status or behavior in all of Japan for each sex and educational level. Then, by dividing the expected number of persons with each health status or behavior by the estimated total number of household persons in Japan, we derived the age-standardized prevalence of outcomes according to sex and educational level. Using this method, we were able to adjust for differences in age distribution depending on sex and educational level. We calculated age-standardized prevalence using the direct method previously described by Naing .
Moreover, Poisson regression analysis was conducted for evaluating an association of educational level with each health-related behavior and status using the data of each age group. In this analysis, educational level was used as factor variable (low educational level was used as reference), and each age group was adjusted as factor variable in the analysis. Lastly, statistical analysis was conducted using R3.6.3 .