This study is reported in accordance with the Strengthening the Reporting of Observational studies in Epidemiology guidelines [26].

Study design and setting

A cross-sectional telephone/online survey was conducted between October 2019 and April 2020 with women who had recently given birth in the Hunter New England Local Health District, New South Wales (NSW), Australia.

Sample and recruitmentParticipants

Women who had participated in a previous study during their pregnancy were invited to participate. Details of the previous study's methods are described elsewhere [27]. Women were eligible for inclusion in this postnatal survey if they: had consented to be followed up after birth in the previous study, were 18 years of age or older, had given birth between 8 and 21 weeks prior, had a level of English proficiency that enabled them to undertake the survey unaided and had not had an adverse pregnancy related outcome, including stillbirth and miscarriage. The survey's primary purpose was to assess women’s recollection of care received in the first 5 months following birth.

Recruitment procedure

Each week during the study period, 60 women from the eligible sample were mailed an information statement outlining the purpose of the study and inviting them to participate. The information statement included a toll free or freephone number that women could call to decline participation in the survey.

One week after the information statement was mailed, non-Aboriginal women were contacted via telephone and invited to participate in a Computer Assisted Telephone Interview (CATI). Women who declined to participate when called were offered the opportunity to complete the survey online.

As per advice received through cultural consultation, Aboriginal women were sent a text message following the mail out of the information statement, offering a choice to complete the survey via telephone, online or to opt out of participation. Aboriginal women who opted to complete the survey online were sent a link to the survey which was active for two weeks. For women who did not reply to the text message after three days, attempts were made to contact via telephone and invite them to participate in the survey.

Attempts to contact women were conducted over a two week period with up to 10 contact attempts made. Women could decline participation at any point during the CATI or online survey.

Data collection procedures

Both the telephone and online surveys were developed using REDCap (Research Electronic Data Capture) electronic data capture tools [28, 29]. The telephone survey was conducted by trained and experienced female interviewers and the survey was pilot tested before use. Aboriginal women were given the choice of undertaking the survey with an Aboriginal interviewer. Additional data was obtained directly from women’s electronic medical records and linked to individual survey responses.

MeasuresWomen’s demographic and pregnancy characteristics

Women’s residential postcode, age, pregnancy risk level, pregnancy outcome and the infants' date of birth were obtained via the woman’s electronic medical record. Women reported the following characteristics as part of the initial survey during pregnancy: marital status, highest level of education, whether this was their first pregnancy, and smoking status and alcohol consumption during pregnancy (using the validated Alcohol Use Disorders Identification Test Consumption (AUDIT-C) tool [30, 31]). The following characteristics of women and their most recent pregnancy were reported as part of this postnatal survey: Aboriginal and/or Torres Strait Islander origin, single or multiple pregnancy, gestational diabetes, pre-eclampsia, caesarean section, height, pre-pregnancy weight, current smoking status and current alcohol consumption (using the AUDIT-C).

Women’s intention to breastfeed and breastfeeding practices

Women were asked ‘Did you plan to breastfeed your baby?’ (yes; no; don’t know), followed by ‘Are you currently breastfeeding your baby?’ (yes – exclusively; yes – breastfeeding and formula feeding; yes – breastfeeding and commenced solid foods; no; don’t know). The following definition of exclusive breastfeeding was provided to women who asked for clarity on the definition of exclusive breastfeeding: ‘Exclusively breastfeeding means that the infant receives only breast milk. No other liquids or solids are given – not even water—with the exception of oral rehydration solution, or drops/syrups of vitamins, minerals or medicines.’ [32]. Women who indicated that they were not currently breastfeeding were asked ‘Did you try to breastfeed or breastfeed for a while?’ (yes; no; don’t know). Women who did try to breastfeed or breastfed for a while were asked ‘How long did you breastfeed for?’ (less than one month; one to two months; three or more months; don’t know).

Reasons for not initiating breastfeeding or breastfeeding cessation

To assess women’s reasons for not initiating breastfeeding, women were asked in an open-ended question: ‘We know there are many reasons for not breastfeeding. Why did you decide not to?’ Women who initiated breastfeeding but had ceased breastfeeding prior to participating in the survey were asked in an open ended question ‘We know there are many reasons for moving on from breastfeeding. Why did you move on?’. Responses were not prompted and women could nominate multiple reasons.

Analysis

Data analysis was conducted using SAS 9.3 (SAS Institute, Cary, NC) [33]. As per Table 2 condensed categories were created for maternal age, Aboriginal and Torres Strait Islander origin, highest level of education completed and reasons for breastfeeding cessation. Responses of 'don't know' were classified as 'no'. Categorisation of reasons for breastfeeding cessation were based on a similar study conducted by Rozga et al.,2015 [21]. The types of reasons in each category are listed in Table 1. For the purpose of assessing associations between maternal and pregnancy characteristics, age of the infant and exclusive breastfeeding at the time of survey completion, categories for the question ‘Are you currently breastfeeding your baby?’ were condensed to ‘Yes, exclusively’ and ‘No, not exclusively’ (including the response options of ‘yes – breastfeeding and formula feeding’, ‘yes – breastfeeding and commenced solid foods’ and ‘no’). Women’s height and pre-pregnancy weight was used to calculate pre-pregnancy body mass index (BMI) using the formula BMI = kg/m2. BMI was categorised at underweight (< 18.5 kg/m2), healthy weight (18.5–24.9 kg/m2), overweight (25.0–29.9 kg/m2) or obese (≥ 30 kg/m2). Model of antenatal care was used to indicate pregnancy risk level. Low risk was defined as hospital and community-based midwifery clinics, midwifery group practice continuity of care and multidisciplinary care for women with social vulnerabilities. High risk was defined as specialist medical clinics and multi-disciplinary care for women with complex medical needs. Aboriginal Maternal Infant Health Services (AMIHS) provide culturally appropriate maternity care to Aboriginal women for both low and high risk pregnancies. As such, women who were identified as receiving their antenatal care through an AMIHS were excluded from the pregnancy risk level analyses.

Table 1 Categorisation of survey responses for breastfeeding cessation

The risk of harm categories used for analysing alcohol consumption risk level are consistent with Australian national guidelines for pregnancy and have been applied for breastfeeding given the same guideline recommendation to not consume alcohol: ‘no risk of harm’ for an AUDIT-C score of 0, ‘low risk of harm’ for an AUDIT-C score of 1–2, ‘medium risk of harm’ for an AUDIT-C score of 3–4 and a ‘high risk of harm’ AUDIT-C score of 5 and above [34]. Women whose postcodes were ranked in the top 50% of NSW postcodes, based on the Australian Bureau of Statistics 2016 Socio-Economic Indexes For Australia [35], were categorised as ‘least disadvantaged’, while those in the lower 50% were categorised as ‘most disadvantaged’.

Descriptive statistics were used to describe maternal and pregnancy characteristics; the infants age at survey completion; women’s plans to breastfeed; breastfeeding initiation; current breastfeeding practice and reasons for not initiating breastfeeding. Simple logistic regression models were used to identify associations between each characteristic and the following outcomes: intention to breastfeed (13 models), breastfeeding initiation (13 models), and reasons for breastfeeding cessation (14 models). Additionally, simple and multivariable logistic regression models (15 models) were used to identify associations between all characteristics and whether women were exclusively breastfeeding at the time of survey completion. The ‘other’ category for breastfeeding cessation was excluded from logistic regression analyses due to a low response rate. An alpha level of 0.05 was set to denote statistical significance. Due to the exploratory nature of the study no further adjustment was made to the alpha level for multiple testing.

Assuming a sample of 500 women, 80% power and a significance level of p < 0.05, allowed a detection of 12.4% significant difference between characteristics for each dichotomous outcome.