1.1 Exclusive breastfeeding

The World Health Organization (2021) defines exclusive breastfeeding as the infant only receiving breastmilk with no other solids or liquids (including water) and recommends that infants are exclusively breastfed the first 6 months of life. Infants who are exclusively breastfed have a 48% reduced risk of all-cause mortality compared to infants who are predominantly breastfed and a 184% reduced risk compared to infants who are partially breastfed (Sankar et al., 2015). Introduction of formula to neonates in hospital can cause statistically significant and prolonged disruptions in the infant gut microbiome, even at 3–4 months postpartum (Forbes et al., 2018). Exclusive human milk feeding in preterm infants demonstrates a positive association of fat-free mass deposits and resulting in improved long-term growth recovery compared to infants who receive formula (Cerasani et al., 2020). Additionally, exclusive breastmilk feeding is associated with a reduced risk of necrotizing enterocolitis (0.11 95% CI 0.02–0.83), compared to infants who receive any formula (Colaizy, 2016). Finally, exclusive breastfeeding is associated with reduced incidence of sepsis and bronchopulmonary dysplasia (Miller et al., 2018) compared to infants who receive preterm infant formula. Further, partial, rather than exclusive breastmilk feeding at discharge from the NICU is one of the strongest predictors of early breastmilk feeding cessation (Ericson et al., 2018).

1.2 Breastfeeding Self-Efficacy theory

BSE in mothers of full-term infants is associated with breastfeeding duration and exclusivity (Brockway et al., 2017). Constructed in Bandura's (1977) broader social cognitive theory, BSE refers to a mother's perceived ability to successfully breastfeed her infant (Dennis, 1999) and has been used as a foundation for many interventions related to breastfeeding (Tuthill et al., 2015). Based on this theory, breastfeeding initiation and duration are influenced by a mother's perceptions of her ability to breastfeed (Dennis, 1999). For example, women with high maternal BSE are more likely to be exclusively breastfeeding at 2 months postpartum (Brockway et al., 2017). The concept of BSE places priority on the mother's breastfeeding confidence and recognizes the importance of empowering and supporting a mother's breastmilk feeding efforts (Dennis, 1999). Dennis (1999) described BSE as being influenced by four main factors: (a) performance accomplishments (e.g. previous breastfeeding experiences), (b) vicarious experience (e.g. observing other women breastfeeding), (c) verbal persuasion (e.g. encouragement from influential others, such as: lactation consultants, family and friends) and (d) physiological and affective states (e.g. pain, stress, exhaustion, anxiety). While BSE theory has gained popularity as a guiding framework for improving breastmilk feeding rates among mothers of healthy, full-term infants (Brockway et al., 2017; Chipojola et al., 2020; Dennis, 1999), limited research has explored this relationship for mothers of preterm infants (Brockway et al., 2020).

Given the complex nature of breastfeeding and the important implications for premature infant health outcomes, further research is required to support the value of using BSE with the preterm population. If applicable, BSE can be modified through behavioural change strategies (e.g. breastfeeding education and supportive statements from healthcare providers [verbal persuasion], peer support while in the NICU [vicarious experience], highlighting and emphasizing breastfeeding successes [performance accomplishments]) to improve breastfeeding rates in mothers of preterm infants (Brockway et al., 2017). The objective of this study was to examine the association between BSE and exclusive breastmilk feeding at discharge in mothers of preterm infants born at 32^0/7–34^6/7 weeks gestational age. We hypothesized that mothers with higher BSE at admission would be more likely to be exclusively breastmilk feeding at discharge, compared to mothers with lower BSE at admission. Understanding the predictors of breastmilk feeding may ultimately lead to the development of evidence-based interventions and revised clinical practices that may improve breastmilk feeding rates for this unique population.

2.1 Setting and study design

This study was a secondary data analysis of the standard care group in the larger Family Integrated Care (FICare) cluster randomized controlled trial (cRCT; [Benzies et al., 2017]). The larger FICare cRCT was implemented in 10 level II NICUs (five standard care and five intervention sites) throughout Alberta, a province in Western Canada with the highest rate (9.2%) of preterm births among the Canadian provinces (Centre for Surveillance and Applied Research & Public Health Agency of Canada, 2020).

2.2 Participants

Participants were recruited between December 2015 and July 2018. Inclusion criteria for the FICare cRCT were: mothers and their preterm infants born 32^0/7–34^6/7 weeks gestational age admitted to a level II NICU or transferred to a level II NICU, in 72 hours of birth ([Benzies et al., 2017]). Preterm infants <32 weeks gestation are usually admitted to a level III NICU (Committee on Fetus and Newborn, 2012), whereas moderate and late preterm infants are usually admitted to a level II NICU. Mothers who gave birth to infants requiring palliative care or with serious congenital abnormalities, and those unable to read and write in English were excluded from the study ([Benzies et al., 2017]). To avoid influencing the primacy outcome of type of feeding at discharge, only mothers in the standard care group were included in this study to ensure that the sample was not contaminated by the FICare intervention. In the case of twins, only data from twin A were included in the dataset.

2.3 Measurement

Maternal BSE at admission to NICU was measured by the 18-item modified Breastfeeding Self-Efficacy Scale–Short Form (BSES-SF; Wheeler & Dennis, 2013). The BSES-SF assesses a mother's confidence in her ability to breastfeed and is validated for mothers of ill and/or preterm infants (Wheeler & Dennis, 2013). Cronbach's alpha for the scale in this study was 0.95, indicating excellent internal consistency. Mothers who were not intending to breastfeed were not assessed for BSE at admission; therefore, they were not included in this study.

Type of feeding at discharge from NICU was assessed by 24 hr maternal recall. The survey item and response options were as follows: In the last 24 hr was your baby fed: (a) Only breastmilk – 100% of feeds were breastmilk, no formula; (b) Mostly breastmilk – 75% of feeds were breastmilk, 25% of feeds were formula; (c) Half breastmilk – 50% of feeds were breastmilk, 50% of feeds were formula; (d) Minimal breastmilk – 25% of feeds were breastmilk, 75% of feeds were formula; and (3) No breastmilk – 100% of feeds were formula. For the purposes of this study, these feeding categories were collapsed into either “exclusive breastmilk feeding” (i.e. 100% of feeds were breastmilk) or “combination or formula-only feeding” (i.e. feeds were a combination of breastmilk and formula), or 100% formula.

Maternal socio-demographic characteristics (e.g. age, education, income) were collected from self-report surveys. Parity, relevant maternal diagnoses (i.e. diabetes [Type I, Type II, gestational] and pre-eclampsia) and infant characteristics (e.g. gestational age, birth weight) were extracted from infant hospital records. Parity (primiparous or multiparous) was used as a proxy for no previous breastfeeding experience or previous breastfeeding experience, respectively.

2.4 Procedures

Study nurses, who were employees in their respective NICUs, received specific training about the FICare study purpose, informed consent and data collection procedures ([Benzies et al., 2017]) recruited all mothers into the study. Most mothers were recruited within 72 hr of infant admission to the NICU. Mothers provided written informed consent for both themselves and their infant and completed surveys using a web-based platform (Fluid Surveys or Qualtrics) on an iPad, soon after infant admission and shortly before discharge.

2.5 Data analysis

Data analyses were conducted using IBM SPSS Statistics for Windows version 25, with statistical significance set at p < .05. Characteristics of the maternal and infant samples were described using means/standard deviations and frequencies/percentages. In cases where less than 20% of data were missing from a questionnaire, missing data were imputed with values equal to the score of non-missing items (Downie & King, 1998). Chi-square tests and Spearman correlations were conducted to examine bivariate associations between variables of interest and type of feeding at discharge. Variables that were statistically significantly associated with type of feeding at discharge using bivariate analyses were included in a logistic regression model. To examine whether the multivariable logistic regression model was a good fit overall, a Hosmer and Lemeshow goodness-of-fit chi-squared test was conducted.

2.6 Ethical considerations

This study was approved by Conjoint Health Research Ethics Board ID: REB15-0067.

3.1 Characteristics of mothers and infants

A total of 306 mothers were enrolled in the standard care group of the cRCT, with 221 participants included in these analyses. Eighty-five participants were removed because of missing data (e.g. did not complete the BSES-SF at admission or report type of feeding at discharge). To identify any meaningful group differences in characteristics between the mothers who were excluded because of pertinent missing data and those who were included, a chi-square test of independence was conducted. There were no statistically significant differences between mothers who were excluded and included based on age, education and income. Table 1 describes characteristics of mothers and infants included in the study. At discharge, 147 mothers were providing exclusive breastmilk to their infant, compared to 74 mothers who were providing a combination of breastmilk and formula or only formula.

3.2 Breastfeeding Self-Efficacy scores

Mothers who were exclusively breastmilk feeding their infants at discharge had statistically significantly higher mean BSES-SF scores at admission (68.4, SD = 13.7), compared to mothers who were providing a combination of breastmilk and formula or only formula (59.6, SD = 14.7; mean difference = 8.8, 95% CI [12.9, 4.6], p < .001).

3.3 Multivariable logistic regression

Bivariate analyses were conducted to examine associations between variables of interest and type of feeding at discharge (Table 1). Maternal BSES-SF score at admission, age, birth in Canada, and diabetes, and infant birth weight and gestational age, were statistically significantly associated with type of feeding at discharge. No association was observed between parity (proxy for previous breastfeeding experience) and type of feeding at discharge.

Multivariable modelling indicated that BSES-SF scores at admission, maternal birth in Canada and absence of diabetes were statistically significant predictors of exclusive breastmilk feeding at discharge (Table 2). Mothers with higher BSES-SF scores at admission were more likely to be providing exclusive breastmilk at discharge, compared to mothers with lower BSES-SF scores at admission, while controlling for other variables in the model (Table 2). For each one-point increase on the modified BSES-SF, the odds of exclusive breastmilk feeding at discharge increased by a factor of 1.04 (95% CI [1.02, 1.07], p < .001).

Maternal birth in Canada and absence of diabetes were also statistically significantly associated with provision of exclusive breastmilk at discharge. Mothers born outside of Canada were 67.4% less likely (OR 0.33, 95% CI [0.15, 0.69]) to be providing exclusive breastmilk to their infant at discharge, compared to mothers born in Canada. Further, mothers without diabetes were over three times more likely to be providing exclusive breastmilk at discharge (OR 3.83, 95% CI [1.57, 9.37]), compared to mothers with diabetes. The logistic regression model was statistically significant, p < .001. Since the Hosmer and Lemeshow goodness-of-fit chi-square test was not statistically significant, p = .449, the model was a satisfactory fit with the data.

4.1 Clinical practice considerations around breastfeeding Self-Efficacy

Dosani and Currie (2017) highlighted that there are many unique breastfeeding challenges with late preterm infants and reinforced that breastfeeding support is crucial to ensure success. Further, Gerhardsson et al. (2018) found that length of hospital stay was inversely correlated with BSE in mothers of late preterm infants. As such, NICU admission may contribute to reduced BSE and potential reductions in breastmilk feeding in mothers of preterm infants. Because breastfeeding self-efficacy at admission to the NICU is predictive of breastfeeding outcomes at discharge, it is important for practitioners to consider practices that will protect or even promote breastfeeding self-efficacy while a preterm infant is admitted to the NICU. Current support interventions for mothers of preterm infants are traditionally based on the medical model and may lower a mother's BSE while in hospital by reinforcing the medicalization of breastfeeding and breastfeeding pathologies (Torres, 2014). For example, breastfeeding mothers often believe they are incapable of providing adequate quantities of breastmilk for their infant (Torres, 2014). In NICUs, healthcare providers may reinforce these beliefs by focusing their breastfeeding efforts on pumped milk volumes and timed breastfeeding sessions. In addition, the “breast is best” messaging reinforces the biomedical discourse on breastfeeding and is perceived as unsupportive by mothers, which can be detrimental to the initiation and success of breastfeeding (Brockway et al., 2020; Debevec & Evanson, 2016).

Improving breastmilk feeding rates for preterm infants needs to involve more than simply educating mothers and their families about the benefits of breastmilk (Debevec & Evanson, 2016) and should focus on the sources of information in breastfeeding self-efficacy theory (Dennis, 1999). Encouraging healthcare providers to adopt more supportive approaches with breastfeeding mothers in the NICU is essential. For instance, offering commendations on successful components of a pumping or latching session while in the NICU may enhance a mother's self-efficacy through performance accomplishments. Further, verbal persuasion from NICU nurses have demonstrated a greater influence on maternal breastfeeding self-efficacy, compared to individuals who are perceived to lack credibility (Brockway et al., 2020; Dennis, 1999). Being honest, non-judgmental and honouring a mother's breastfeeding goals are ways that healthcare providers can be supportive when working with breastfeeding mothers (Debevec & Evanson, 2016). Finally, patient-centred care that values maternal needs in conjunction with the needs of the preterm infant can help to address maternal affective-physiologic status and should be considered the standard of care in the NICU setting (Debevec & Evanson, 2016).

Healthcare professionals working in level II NICUs should consider a more individualized approach to breastfeeding support that enhances collaboration with breastfeeding mothers, considers their bio-psycho-social experience and enables them to feel more central to the process. Several studies have demonstrated that when mothers felt supported, they experienced increased BSE, which resulted in an earlier initiation and longer duration of breastmilk feeding (Debevec & Evanson, 2016; Mulder, 2006; Wheeler & Dennis, 2013). Ultimately, a better understanding of maternal breastfeeding goals and perceptions about what effective breastfeeding support looks like is integral to enhancing maternal BSE and breastfeeding success.

4.2 Limitations and recommendations for future research

Previous breastfeeding experience, while an important variable influencing BSE and subsequent breastfeeding at discharge, was not measured in this study. As such, parity was used as a proxy for breastfeeding experience, which may not have accurately captured previous breastfeeding experience. Future research that specifically captures breastfeeding experience as a variable is recommended.

Participants in this study represented a relatively homogenous sample (e.g. postsecondary education, high income, most were married); and the results may not be generalizable to the entire population of mothers of preterm infants. Future research with heterogeneous samples (e.g. socio-economic diversity and lone parents) would be valuable in further understanding the association between BSE and breastmilk feeding at discharge. Finally, only mothers who completed the BSES-SF at admission and infant feeding outcomes at discharge were included in this analysis. As such, these inclusion criteria may have contributed to a sampling bias by omitting mothers who did not complete infant feeding assessments at discharge because they were not successful in breastfeeding and were no longer motivated to participate in the study.