Research setting and subjects
The research setting is Company T, a large semiconductor manufacturer with 10,000 employees in Hsinchu Science Park in northern Taiwan, which is one of the world's most significant areas for semiconductor manufacturing, similar in importance to Silicon Valley. This company was selected because it was the first semiconductor manufacturer to provide lactation rooms and breast pumping breaks for working mothers, which was an uncommon practice at the time the survey was conducted. Secondly, about 50% of the company's employees are women, most of whom are of childbearing age. Thirdly, the female employees are office workers or fab workers or, so the influence of different working conditions on breastfeeding behavior could be observed.
Office workers had superior educational levels and higher compensation levels compared with the fab (fabricating) workers, and they worked about eight hours a day from 08:30 to 18:00. Their positions encompassed specific job responsibilities, and so their bonuses were largely tied to individual performances. By comparison, fab workers worked 12-hour shifts, from 07:00 to 19:00, or from 19:00 to 07:00. Their jobs were inconvenient and inflexible, as they have to remove and put back on their clean room suits when leaving and returning to their workstation, and a substitute had to take their place during their absence. Performance bonuses at most Taiwanese manufacturers are based on both group productivity and individual performance criteria. It is the group leader's responsibility to promote higher productivity as well as evaluate individual performances. If productivity is very high, a worker's annual bonus can usually exceed their base salary. Therefore, most workers will do their best to maximize productivity.
Data collection and variable definitions
This study was conducted between April and September 2003. Questionnaires were used to collect data on female employees' breastfeeding behaviour, child rearing and work status when raising their most recently born child (See Additional file 1). In developing the questionnaire, in-depth interviews were conducted with 25 female workers at Company T. Seven public health professionals also reviewed and revised the questionnaire. The preliminary questionnaire was further revised after it was used in a pilot study at another semiconductor manufacturer. The final questionnaire was then distributed in August and September 2003 to 1,326 female employees, whom had taken maternity leave between January 1999 and April 2003 as recorded by the manufacturer's human resource department. The company's occupational and environmental health nurses helped distribute and collect the questionnaires. To help clarify questionable answers, the questionnaire was not anonymous, but all personal information and answers were kept confidential by the researchers. Written consent was not required in questionnaire survey. Participation in this study is entirely voluntary. We described the aims of this study and invited the subjects to participate in this study, and oral consent was obtained from the participants. This study was reviewed and approved by the ethics committee of the College of Public Health, National Taiwan University.
The study used two dependent variables: breastfeeding initiation (whether or not the mother had ever breastfed the most recently born child) and continuation of breastfeeding after returning to work (yes or no). A mother initiated breastfeeding if she had ever breastfed her child. Mothers were defined as breastfeeding only during maternity leave (initiation and no continuation) if they completely stopped breastfeeding before, or within two weeks after returning to work. Mothers were defined as continuing breastfeeding if they continued for at least two weeks after returning to work.
Several employment characteristics were analyzed as independent variables, including worksite (fab/office), according to human resource department records), shift work ("Did you do shift work after you returned to work?" yes/no), flextime ("Did you need any substitute when you left your position during working hours? " yes/no), child's age (< 1 year, 1–3 years or >3 years) and length of employment (≤ 6 years, 7–9 years or ≥ 10 years). Other independent variables were an employee's knowledge of breastfeeding policies in the workplace, including awareness of lactation rooms ("Did you know about the lactation room in your company?" yes/no) and awareness of breast pumping breaks. ("Did you know about pumping breaks? " yes/no).
The effects of employment characteristics and manufacturer policies on breastfeeding behavior were estimated using chi-square tests and logistic regression. Preliminary associations between categorical variables were verified through chi-square tests on contingency tables. Logistic regression was used to analyze breastfeeding behavior with respect to the following variables: age, education, worksite, shift work, years of employment and policy awareness. Variables were culled after an evaluation of their collinearity. If two or more variables were highly correlated, separate models were run for each of the collinear variables, and the model with the strongest relation to the breastfeeding measures was selected. Statistical analyses were conducted using SAS Version 8.2.