Exposure measures

The selection of exposure variables included maternal, psychosocial and cultural factors, demographic, child and birth factors, and maternal health factors that have been shown or hypothesised to be important based on current literature.

Participants were asked to recall how they intended to feed their baby from birth selecting from pre-specified options ranging from breastfeeding only, combined breastfeeding with water or juice or formula, to formula feed only or other. Participants also indicated their intended duration of any breastfeeding from given options of at least 6, 12, 24 months or planned to breastfeed for as long as possible or didn’t think about how long.

Awareness of the recommended infant feeding practices was assessed by two questions. The definition of EBF was provided and participants were asked what they thought the recommended duration of EBF was, and what the recommended age to introduce solids was. Participants could answer in weeks, months or don’t know.

Breastfeeding attitudes, control and perceptions of breastfeeding norms were measured using a modified version of the Breastfeeding Attrition Prediction Tool (BAPT), first developed and validated by Janke (24) for use in mainly white mothers. The original tool included 52 items on a 6-point Likert type scale and four subscales: Negative Breastfeeding Sentiment, Positive Breastfeeding Sentiment, Social and Professional Support, and Breastfeeding Control. A higher score indicates better attitude, greater perceived social pressure to breastfeed, and higher confidence to breastfeed.

Several items were modified to suit the Chinese Australian context. In the Social and Professional subscale, the questions asked the mother to rate how much she perceived certain social and professional contacts want her to breastfeed from ‘definitely not breastfeed’ to ‘definitely breastfeed’ or ‘not applicable’. The options were modified to suit the Chinese Australian context and included family members (baby’s father, my mother, my mother in law, my sister, other relatives), friends (my closest female friend), and health professionals (maternal and child health nurse, my doctor, my midwife). During analysis, the responses for this subscale was recoded to − 3 to 3 including zero to indicate non-applicability for ease of interpretation. The positive breastfeeding sentiment subscale was excluded as in the original study there was a general tendency for women to agree with the items resulting in decreased variability and poor predictive ability [24]. The modified BAPT tool consisted of 15 items for negative breastfeeding sentiment (disadvantages of breastfeeding and advantages of formula feeding), 9 items measuring subjective norm (mother’s perception on whether health professionals and their social contacts believed she should breastfeed or not), and 10 control items (perception of ease or difficulty associated with breast and formula feeding).

Confirmatory factor analysis was conducted on the modified BAPT tool to check if these constructs were adequately represented in Chinese Australian mothers. The results of the factor analysis suggested several modifications were necessary. An item (“I didn’t need help to breastfeed”) in the control scale was excluded due to poor factor loading (0.27). The subjective norm construct was a better fit as two factors, namely family and health professionals. Two items (“my sister” and “other relatives”) in the subjective norm family scale were excluded due to the high proportion of non-applicable responses.

The Cronbach’s alpha coefficient for the modified BAPT tool in this study with 289 participants were 0.93 for the attitude scale, 0.71 for the subjective norm family members, 0.92 for the subjective norm health professionals, and 0.91 for the control scale, indicating high internal consistency. This is comparable to the internal consistency of the original tool of 0.83, 0.85, and 0.81 respectively.

Participants were asked if they followed the traditional Chinese postnatal confinement practice and whether they thought this traditional practice was important in helping a mother to breastfeed using a 5-point scale ranging from ‘extremely important’ to ‘not at all important’. Acculturation level was measured with a single question stating what language do you prefer? with five response options (Chinese only, mostly Chinese and some English, Chinese and English equally well, mostly English and some Chinese, English only) [25].

Participants were asked to rate the relative importance on a 5-point scale (extremely important to not at all important or not applicable if they did not receive any advice) of various sources of information in influencing their feeding decisions. Sources included the woman’s partner, her mother, mother in law, other relatives, friends, other mothers in the community, health professionals (maternal child health nurses, midwives, doctors, other hospital staff, lactation consultant, antenatal class), breastfeeding support hotlines, and the internet (website, blogs, and apps).

Basic maternal and child demographical details were collected including date of birth, child’s gender, maternal marital status (re-categorised to married and not married with the latter category consisting of de facto, divorced, separated, never married, and widowed), maternal country of birth, maternal employment status (full time or part time including self-employed, or not working or studying including those who were unemployed or caring for children full time), and indicators of socioeconomic status including maternal education levels, family annual income, and postcode (categorised according to socioeconomic indexes for areas). Additional details important to breastfeeding asked were infant’s age when mothers returned to work, the length of residence in Australia, if they co-resided with either maternal or paternal grandparents at any time during the child’s first year of life, and the number of children (re-categorised to primipara and multipara).

Information related to maternal health known to influence breastfeeding were also collected. These included maternal smoking status during pregnancy, presence of diabetes (pre-existing and gestational diabetes), pre-pregnancy weight and height (converted to body mass index and categorised into healthy, overweight, or obese according to World Health Organization’s cut-off values for Asian populations [26]), and self-report diagnosis of postnatal depression or anxiety (yes/no/unsure). Similarly, relevant obstetric factors such as child’s birth weight, and mode of delivery were collected.