A. Study population and design
The Institute of Occupational Safety and Health, a research institute under the Council of Labor Affairs, conducted the Survey of Perceptions of Safety and Health in the Work Environment in Taiwan to characterize employees' perception of safety and health in the work environment. This was done every three to four years as a
supplement to the Human Resources Survey organized by the Department of Statistics, who conduct large scaled nationwide surveys on a routine basis as references for policy setting. A representative sample of workers in the week of September 9, 2007 was obtained from a questionnaire survey using a two-stage stratified cluster sampling design. “Villages” and “Lis” (a unit of administrative district in urban areas, equivalent to “villages” in rural areas) were the primary sampling units, and were stratified into 23 strata according to the level of urbanization, industrial structure and educational
attainment in the first stage. The numbers of villages and lis to be sampled from a certain stratum were determined according to the total numbers of villages and lis in the stratum. Households were the sampling units in the second stage, and were randomly selected from the villages and lis sampled in the first stage according to the resident registration data of each villages and lis. Within each household selected, subjects who were in employment during the study period were identified. A total of 28,716 workers were identified among the sampled households. All the adults in each sampled
household met the inclusion criteria were asked to participate in the survey.
The questionnaire contained questions to obtain data on workers’ demographic characteristics, lifestyle characteristics, perceived exposure to physical, chemical and ergonomic factors in the work environment, and musculoskeletal discomfort in different bodily parts. The study questionnaire was distributed by trained interviewers to the
residents of the study candidates. All the interviewers had received a series of standard training and been given a standard procedure manual before the survey. The
door-to-door questionnaire administration was conducted at different times and on different days of the week to accommodate the availability of the study candidates.
Each sampled worker was asked to fill out the questionnaire, which was collected by the interviewer right after completion. If the participant was not able to complete the
questionnaire or had any problem with answering, the interviewer would offer instant help. If the worker was not at home upon the visit of the interviewer, the questionnaire would be left and appointment wound be made for it to be collected after it was
completed. The questionnaire was designed for the study subjects to complete in 15-20 minutes. Upon receiving the answered questionnaire, the interviewer would perform an error check to clarify confusions and correct possible errors immediately. Before visiting, a telephone call was made to make an appointment with the candidate. In addition, multiple attempts were made to reach the study candidate to ensure a high response rate. No monetary incentive was provided for participation.
B. Questionnaire
The interview was conducted by using a structured questionnaire containing only close-end questions. Two major sections were contained in the survey. In the first section, demographic data and employment information including age, gender, body weight, body height, average sleep hours on workdays and on holidays, work exposure, and exercise habit were inquired. The other section surveyed musculoskeletal
discomfort symptoms. The musculoskeletal discomfort symptoms were collected using the Chinese version of the standardized Nordic Musculoskeletal Questionnaire (NMQ).
The NMQ questionnaire asked whether the participants had suffered from
musculoskeletal discomfort in any body part including pain or soreness during the past year. Those recalled having musculoskeletal discomfort were asked to answer which of the body part is involved (neck, upper back, lower back and waist, shoulder, elbow, hand and wrist, hip and thigh, knee, and ankle). An illustration of the human body was provided on the questionnaire to indicate the body parts. If there were any pain or soreness in any of the body parts reported, the degree of discomfort was then asked. The participants were inquired to place the discomfort into the following severity categories, namely, “caused no effect on work performance”, “affected work performance without causing sickness absence”, “caused sickness absence for less than four days”, or
“caused sickness absence for more than four days” in the past year were recorded.
Those reported having musculoskeletal discomfort which “affected work performance without causing sickness absence” or more severe categories were considered to have positive outcomes.
C. Body mass index
Body mass index (BMI) was calculated from a person's weight and height. For adults 20 years old and older, BMI is interpreted using standard weight status categories that are the same for all ages and for both men and women. The standard weight status categories associated with BMI ranges for adults were divided into three groups
according to the suggestion of Health Promotion Administration, Ministry of Health and Welfare, Taiwan: (1) less than 24 as normal, (2) 24 or more and less than 27 as
overweight, and (3) more than 27 as obese.
D. Exercise habit
The questionnaire asked about workers’ exercise habit falling into three
categories: less than once per month (never), one to three times per month, and more than once per week.
E. Work exposure
Exposures to ergonomic factors included whole body vibration (vibration transmitted to the entire body from a vehicle seat or through the feet and legs from a vibrating), using vibrating hand tools (such as grinders, road breakers or drills),
repetitive hand movement (such as typing, repetitive reaching or repetitive assembling work), using heavy hand tools, lifting heavy objects, awkward body postures (such as body twisting, long-term standing, walking, kneeling or squatting), speedy work pace (such as assembly line jobs), lengthy computer use, and inappropriate desk height (including tables or chairs). The questionnaire asked about the proportion of working time that was occupied by these activities, with three possible answers: never,
sometimes, and often. Workers were considered having the above occupational physical exposure if he or she reported “often” exposed to questions related to any of the above factors. The results were analyzed with Factor Analysis in order to group the main factors that represent work exposure.
F. Sleep Hours
There were two indices referring to sleep quality in our study, namely, average sleep hours on workdays in the previous week, and average sleep hours on holidays in the previous week. The average duration of sleep per day on workdays and on holidays in the previous week were recorded in hours and minutes and were divided into three groups according to the average duration of sleep hours of men and women: (1) more
workdays, and (1) more than 8, (2) more than 7, less than 8, and (3) less than 7 for sleep hours on holidays.
G. Statistical analysis
The main aim of the study was to determine whether sleep hour predicted low back pain among male and female workers. The analysis included descriptive summary of demographics and exposure to ergonomic factors. Differences in prevalence were evaluated by chi-square tests (categorical covariates) or analysis of variance (continuous covariates: age, sleep hours on workdays, and sleep hours on holidays) at the two-tailed significant level of 0.05. Associations of discomfort outcomes with risk factors were characterized by odds ratios (OR) and associated 95% confidence intervals (95% CI), estimated by fitting logistic regression models. After an initial univariate analysis, only significant factors (at P<0.05) were included for the backward stepwise logistic
regression to determine the order of entry of the independent variables into the regression analyses to construct a best-fitted multivariate model. Only significant predictors were retained in the final regression model. To stay in the model, variables were required to be significant at P value < 0.05. Generalized linear modeling was used to calculate the adjusted relative risk for the variables that were statistically significant in the logistic regression analyses.
Population attributable risks (PARs) were calculated to estimate the contribution of various risk factors to low back pain. The PAR for any risk factor represents the preventable proportion of low back pain cases if workers were not exposed to the
specified factor. PAR was calculated using the formula P(RR - 1)/ [P(RR - 1) + 1] x 100, with P representing the proportion exposed in the population and RR presenting the relative risk due to the exposure [68]. Although the OR is often used as an estimation of
the RR, in our study this is not appropriate due to the high prevalence of LBP, which is against the rare disease assumption. Population-attributable risk (PAR) was calculated by the prevalence of each predictive factor and the relative risk of that predictive factor.
Statistics were performed by means of JMP software version 9.0 (SAS institute Inc., Cary, NC, USA.) for Mac.