Data sources and study population
This retrospective population-based cohort study used deidentified data from the Victorian Perinatal Data Collection (VPDC), a validated dataset  of routinely collected state-wide information for every birth in Victoria, Australia, of ≥20 weeks’ gestation (or ≥ 400 g birth weight if gestation is unknown). Variables include mode of birth, type of intrapartum analgesia, use of oxytocin infusions in labour, initiation of breastfeeding, use of infant formula in hospital and feeding status at discharge from hospital. We analysed all singleton term livebirths at ≥37 weeks gestation whose mother initiated breastfeeding in hospital between 1 January 2010 and 31 December 2018. Breastfeeding initiation refers to whether the baby was put to the breast at least once or whether any attempt was made to express breastmilk during the hospital stay.
A subset of this population, namely those term singleton livebirths born between 1 January 2015 and 31 December 2017, were linked to the Child Development Information System (CDIS) containing information on infant feeding status at 3-month and 6-months. These dates were chosen because of the availability of linked data within this timeframe. The CDIS is a centralised dataset reporting ongoing breastfeeding outcomes through the Maternal and Child Health (MCH) service. The MCH service is a universal primary health service that provides all Victorian families with 10 consultations with MCH nurses between birth and school age, with additional as required. At each consultation parents report their child’s feeding status as exclusively breastfeed, partially breastfed or formula-fed. The most recent total participation rates for the MCH service were shown to be 94.1% at 4-months and 85.8% at 8-months , when parents report their child’s feeding status at 3-months and 6-months, respectively. Almost all Local Government Areas reported their MCH information to CDIS by the study period, 2015–17, enabling linkage with births in 2015–17. The linkage was conducted by the Centre for Victorian Data Linkage based on a concordance of predetermined identifiers.
Outcomes, exposures and covariates
The primary short-term breastfeeding outcomes were formula supplementation in hospital, defined as breast-fed babies receiving any amount of formula during the initial hospital stay, and last feed at the breast, defined as having the last feed before hospital discharge directly and exclusively from the breast, and therefore excludes expressed breast milk. These were assessed using the total study population.
The primary long-term breastfeeding outcomes were breastfeeding status at 3-months and 6-months of age, which was recorded as either exclusive, any or no breastfeeding. Exclusive breastfeeding is defined as receiving only breastmilk as the source of nutrition and includes expressed breast milk. Any breastfeeding includes receiving any breastmilk, in addition to artificial milk sources, water-based drinks or solid food. Long-term breastfeeding outcomes were assessed using the linked VPDC-CDIS dataset sub-study population for births in 2015–17.
The exposure variables were four intrapartum interventions: pre-labour CS, in-labour CS, epidural analgesia, and synthetic oxytocin infusion. Method of birth was categorised as either pre-labour CS, in-labour CS, or vaginal birth. Epidural analgesia included epidural block, spinal block and combined spinal/epidural block used to relieve pain in labour. We did not assess epidural anaesthesia provided de novo to facilitate operative birth because of the brief fetal exposure. Synthetic oxytocin infusion includes that which was used for induction and/or augmentation of labour.
Several covariates were assessed. These included socioeconomic status, parity (primiparous, multiparous), hospital admissions status (public, private), birthweight (< 2500 g, 2500-3999 g, ≥4000 g), maternal region of birth, maternal age, maternal body mass index (BMI), smoking status during pregnancy, marital status (married, de facto, single), and gestation at birth. Socioeconomic status was defined using the Socio-Economic Indexes for Areas (SEIFA) indices . The indices are based on information from the five-yearly Census of Population and Housing and are divided into quintiles, with 1 being the most disadvantaged and 5 being the least disadvantaged. The SEIFA used here is based on the smallest available residential neighbourhood level (SA1). Maternal region of birth was classified by the Standard Australian Classification of Countries (SACC) . Marital status information was available for the total population but not for the linked VPDC-CDIS sub-population.
We first described the rates of intrapartum interventions and breastfeeding outcomes in our populations. Categorical variables were reported as absolute numbers and percentages and compared using Pearson’s Chi-square test. To assess the association between each intervention and breastfeeding outcomes, we then performed univariable and multivariable logistic regression models to obtain adjusted odds ratios (aOR) with 95% confidence intervals (CI), adjusting for available confounders. Given that the use of each intervention was not always mutually exclusive, each intrapartum intervention was first assessed regardless of whether other interventions were also used, then each was assessed when used in isolation, and finally cumulatively. A two-tailed P-value of < 0.05 was deemed statistically significant.
Missing data were excluded case-wise. Data were missing for 8676 (1.4%) cases regarding formula supplementation in hospital, 3808 (0.6%) cases regarding last feed at breast and 41 cases (< 0.1%) regarding method of birth. There was no missing data on epidural analgesia and synthetic oxytocin infusion. Data were missing for < 0.1% of mothers regarding parity, maternal age, sex of baby and hospital admission status, 53,029 (8.9%) cases regarding smoking status, 39,181 (6.5%) cases regarding maternal BMI, 7744 (1.3%) cases regarding marital status, 37,224 (6.2%) cases regarding SEIFA quintile, and 3681 (0.6%) cases regarding country of birth region. Within the linked population, feeding status information was missing for 16,269 (15.4%) of infants at 3-months and 10,072 (9.5%) of infants at 6-months. A ‘missing’ category was created for covariates with substantial missing data in order to include cases with valid data on other variables in multivariable analyses.
Statistical analyses were performed using Statistical Package for the Social Science Version 26 (SPSS, IBM Corp., Armonk, New York, USA) and Stata SE Version 16 (Stata, 2020, release 16, StataCorp, Texas, USA).