Data and study participants
We used anonymized data from the 2013 Comprehensive Survey of Living Conditions (CSLC) conducted by the Ministry of Health, Labour and Welfare of Japan [2]. The CSLC is a nationally representative sample of the Japanese population, and the 2013 CSLC is the latest data available at the end of May, 2021. The details of the 2013 CSLC are explained elsewhere [15]. Briefly, the 2013 CSLC targeted all households (approximately 300,000 households) and household members (approximately 740,000 persons) in 5530 districts stratified and randomly selected from the 2010 census ward. Anonymized data had various anonymization measures (i.e., resampling). For example, about one-sixth of the data from the original CSLC was re-extracted, the age was given a 5-year-old class code, and the prefecture information was deleted. In the 2013 CSLC, people aged 19 and younger did not need to answer questions about drinking or smoking habits, and people in need of nursing care and those in a hospital/facility were exempt from answering about their health status and lifestyle. Therefore, among 97,345 anonymized data, we excluded 30,032 persons from our analyses because of being aged 20 or younger, being in a hospital/facility, being in school, having received long-term care certification, and missing data on age, hospital admission, working status, and/or self-reported LBP. The final number of participants included in this study was 67,313 persons (31,443 men and 35,870 women) (Fig. 1). The survey date was 6 June, 2013, but the work situation was as at May 2013.
Measurements
Self-reported low back pain
The CSLC asked the respondents about their health status, “Have you been feeling sick (subjective symptoms) due to illness or injury in the last few days?” For this question, the respondents chose either “yes” or “no”. Those who answered that they had subjective symptoms selected all the applicable symptoms from 42 options. One of these 42 options included low back pain (LBP). In this study, those who chose LBP as a subjective symptom were defined as persons with self-reported LBP.
Occupational class variables
With reference to prior Japanese studies [16, 17], we adopted working status, occupation, employment status, and company size as occupational class variables.
Working status in this study was evaluated using the answer to the question, “Did you have any paid work during May of 2013?” A respondent who answered “Yes” was considered to be working. On the other hand, a respondent who answered “No” was considered to be non-working.
The CSLC asked persons with paid work about the type of job (hereafter, occupation) and whether they were employed or self-employed. Then, those who answered “employees” were asked about the labor force status as classified by the employer and the number of employees in their business establishment (hereafter, company size). Employed persons were asked to answer questions about their employment contract with seven response options (regular employees, part-timers, casual staff, temporary employees, contract staff, contract-based workers, fixed-term employees, and others); and company size with nine response options (1–4, 5–29, 30–99, 100–299, 300–499, 500–999, 1000–4999, 5000 or more, or public offices).
Occupation was separated into 12 groups based on the definition of the major classification of the Japanese Standard Occupational Classification (JSCO) [18]: managers; professionals and technicians; clerks; sales workers; services workers; security/protective workers; agricultural/forestry/fishery workers; manufacturing workers; transportation/machine workers; construction/mining workers; carrying/cleaning/packing workers; and other unclassified occupations. The JSCO has an adequate validity as a theory-based classification system suitable for evaluating Japan's occupation-related social status [19].
Employment status was classified into four groups: regular employees, part-timers and casual staff, other types of non-regular employees including temporary employees, contract staff, contract-based workers, and fixed-term employees, and the self-employed and others [16].
Company size was classified into six groups: 1–4, 5–29, 30–99, 100, or more, and public servants. The CSLC did not ask self-employed people about the size of their establishment. According to a survey of 5000 self-employed people nationwide [20], 98.2% of self-employed people answered that the number of employees was 4 or less. Therefore, in this study, company size of self-employed people was classified into 1 to 4 employees.
Covariates
According to previous studies [4, 12, 21,22,23], the following variables were included as covariates that may be potential confounders of the association between LBP and occupational class: age, socio-economic status (SES), lifestyle habits, and chronic medical conditions. SES included marital status, family size, housing tenure, equivalent household expenditures, and education. Lifestyle habits included alcohol intake, smoking status, and sleep duration.
Marital status was categorized into married, never-married, and widowed/divorced. Family size was categorized into 1, 2, 3–4, and > 4. Housing tenure was dichotomized as owner-occupiers versus renters. Equivalent household expenditures (Japanese thousand yen per month) were divided into three groups by the tertiles (i.e., low, middle, and high), and this grouping by the tertiles was carried out by gender. Education (years of schooling) was categorized into < 10, 10–12, 13–15, and > 15. Alcohol intake (frequency of drinking) was categorized into none, several days a month, 1–4 days a week, and > 4 days a week. Smoking status was categorized into never-smokers, ex-smokers, and current smokers. Sleep duration (sleeping hours per day) was categorized into < 6, 6–7, and > 7. Chronic medical conditions were defined as persons with at least one disease under treatment for hypertension, diabetes mellitus, cerebrovascular disease, heart disease, and cancer.
Regarding the handling of missing values of the covariates, a group with missing values was created and included in the analysis subjects. By using this method, the influence of no answer on the covariates can be considered, and the number of analyzed participants can be maintained [24].
Statistical analysis
Data comparisons between men and women or between those with and without self-reported LBP were tested using the chi-squared test for categorical variables and the t-test for continuous variables.
To investigate the cross-sectional association between occupational class variables and self-reported LBP, we used the generalized estimating equations of the multivariable Poisson regression model. The independent variables were occupation, employment status, and/or company size. Using clerks, the regular employees, or the companies with ≥ 100 employees as a reference, a prevalence ratio (PR) and a 95% confidence interval (CI) for self-reported LBP was calculated for other groups of occupation, employment status, or company size, respectively. In model 1, the age-adjusted PR was calculated. In model 2, all covariates (i.e., age, SES, lifestyle habits, and chronic medical conditions) were simultaneously added and the multivariate-adjusted PR was calculated. In model 3, to assess the independent association of each occupational class variable with self-reported LBP, we conducted mutual adjustment for all three items of occupational class variables, in addition to adjustment for all covariates.
Since LBP prevalence and occupational class variables vary by gender [4, 11], we performed stratified analyses by gender. The level of significance was 0.05 (two-tailed test). Statistical analyses were performed using the IBM SPSS Statistics Ver. 27 for Windows (Armonk, New York).
Ethics
Based on Article 36 of the Statistics Act, we received approval of use for academic purposes from the Japanese Ministry of Health, Labour and Welfare (approval number 17003), and were provided with data without any information that would identify individuals.