Breastfeeding self-efficacy and breastmilk feeding for moderate and late preterm infants in the Family Integrated Care trial: a mixed methods protocol

FICare is a model of care that actively supports families to participate in the care of their infant and was originally introduced to Canada at the Mount Sinai Hospital level III NICU in Toronto, Ontario [36]. The intervention originated from the Humane Neonatal Care model developed in Tallinn, Estonia, where parents actively participated in the care of their infants, while nurses and psychologists provided education and support [36]. Existing research regarding the effectiveness of FICare is limited to level III NICUs. A recent cRCT examining the impact of FICare in level III NICUs found that infants exposed to the FICare intervention were significantly more likely to be exclusively breastmilk feeding at discharge (279 [70%] of 396) compared to those receiving standard care (394 [63%] of 624; p = 0.016) [37]. As such, it is important to determine if the FICare model is similarly effective for improving breastmilk feeding outcomes in moderate and late preterm infants admitted to a level II NICU.

Families participating in the FICare study will be required to spend a minimum of six hours per day, or approximately three feeding times, at the NICU. Nurses will support and educate mothers and fathers in their parenting role with a focus on actively involving them in the care of their infants while in the NICU. Parents will share in the care of their infant(s) as soon as they are able, starting with simple tasks, such as skin-to-skin contact and diapering, and progressing to more complex tasks, such as feeding. FICare involves three main components (Fig. 2): (i) information sharing, (ii) parent education (including parent-education pathways and specially designed apps), and (iii) parent support. Information sharing is bidirectional and involves parents verbally reporting on their infants’ progress at daily bedside multidisciplinary rounds and actively contributing to discussions about the plan of care. Parent support will involve one-on-one discussions with veteran parents. Veteran parents (those who have previously had an infant in the level II NICU environment) will provide practical advice, guidance, and support to parents in the FICare study. Finally, parent education will include standardized, evidence informed parent education delivered individually or in group settings.

In Alberta, infants admitted to a level II NICU are generally born after 32 weeks, weighing more than 1500 g. Nutritionally, these infants tend to be on total parenteral nutrition, receiving gavage feeds, or full oral feeds, depending on their developmental maturity [7]. All 10 level II NICUs in Alberta formally support the provision of mother’s own milk as the first choice for infant feeding [38]. While there is no formalized breastfeeding education requirement [39] for practitioners working in NICUs, all mothers have access to International Board Certified Lactation Consultants. Visitation models vary among the 10 sites, with most NICUs allowing parents unrestricted 24-h access to their infant. There is a mix of open and closed ward models, with five of the NICUs planning to transition to single room wards during the timeline of the FICare study.

The FICare cRCT will include mothers of preterm infants born between 32^0/7 and 34^6/7 weeks GA, admitted to a level II NICU in Alberta, who speak, read, and write English. To ensure an adequate dose of the intervention, infants must have a minimum NICU stay of 5 days. Infants born greater than 35 weeks GA were excluded from the study as they would not meet the minimum 5-day requirement if discharged around 36 weeks GA. We will exclude mothers of infants with social risk that may interfere with their ability to engage in FICare, and infants with a severe congenital abnormality or chromosomal anomaly, or receiving palliative care.

We based sample size estimates on the primary outcome, length of stay. Due to an anticipated skewed distribution for length of stay, we used a natural logarithm transformation to calculate sample size [40]. To achieve a power of 0.80, we need to recruit 181 mothers into each group for the primary outcome of length of stay; and 211 to achieve a power of 0.9. In 2014, there were 1030 moderate and late preterm infants admitted to a level II NICU in Alberta. To account for a response rate of 80% [36], attrition, and infants with a length of stay of ≤5 days (6.08%), we will approach 824 potential participants over the 30-month recruitment period. This will also ensure that the sample size, 330 per group, is sufficient to assess secondary outcomes and to provide the qualitative sample for the study.

Within 72 h of admission to the level II NICU, nurses will inform mothers about the study. If interested, a FICare super-user (a nurse specially trained in the FICare model) will screen mothers for eligibility, answer questions about the study, obtain informed consent, and administer the baseline questionnaire. As this is a cRCT, infants of mothers at the intervention sites who do not wish to participate in the study will receive the same hospital care as infants whose mothers are participating in the study. The present study will be nested within the larger FICare clinical trial (Fig. 3).

Online questionnaires will be administered to mothers at two time points in the study (Table 1). The first set of questionnaires will be administered at enrolment; the second set about 24 h before the infant is discharged from the NICU.

Breastfeeding self-efficacy will be measured using the modified Breastfeeding Self-efficacy Scale – short form (BSES-SF; Table 1) for mothers of ill and preterm infants [20]. The BSES-SF will only be administered to mothers who are breastfeeding, expressing their own breastmilk, attempting to breastfeed, or are planning to breastfeed. Mothers who have weaned or are not planning to breastfeed will not complete the BSES-SF. However, infant feeding rates will be captured for all infants in the study. BSES-SF data will be assessed quarterly throughout the quantitative data collection period and will inform purposive sampling for the qualitative phase of the study. The remaining quantitative data will be assessed upon completion of quantitative data collection and completion of qualitative data analysis.

The Labbok and Krasovek classification system [41] will be used to assess infant feeding rates at admission and discharge (Table 1). Maternal recall is frequently used as an effective method to collect infant feeding data [42]. Maternal recall will be used to assess infant feeding over the previous 24 h and will be classified as:

The infant feeding questions will be predicated by i) is your baby receiving any human donor milk, and ii) is your baby receiving any additives to your breastmilk to help them grow?

We will collect data electronically and data will be stored on secure servers. Quantitative data will be managed as per Benzies et al. [30]. Upon completion of analysis, we will store data with the PolicyWise Secondary Analysis to Generate Evidence (SAGE; formerly the Child Data Centre of Alberta) database. This will help to facilitate data access by other qualified researchers.

Statistical analysis of the quantitative results will be performed using Stata Data Analysis and Statistical Software. The primary outcome of the statistical analysis is to determine if a difference exists in: i) breastfeeding self-efficacy scores (BSES-SF) and ii) breastmilk feeding rates, between the control and intervention groups. Results from the BSES-SF will be treated as continuous data whereas results from breastmilk feeding rates will be treated as categorical data [20, 43]. Characteristics of participants and scores on scales will be presented as descriptive statistics (means, frequencies, and percentages). We will use an omnibus test (Hotelling’s t-tests and Chi square) to assess for baseline differences between intervention and control groups on socio-demographic and health characteristics, breastfeeding self-efficacy, and breastmilk feeding rates.

To test if there is a difference in breastfeeding self-efficacy and breastmilk feeding rates between the intervention and control groups, we will use Hierarchical Linear Modeling (HLM; also known as multilevel modeling or mixed-effect modeling) and Hierarchical General Linear Modeling (HLGM), respectively. This study is nested in a larger cRCT and there is a potential for variance in care delivery at each of the NICU sites. As such, there are multiple levels of data that need to be considered to accurately assess the effectiveness of the intervention on breastfeeding self-efficacy and breastmilk feeding rates. By simultaneously investigating the relationships between the different levels of data, HLM and HGLM analysis can account for variance among variables at different levels [44]. The HLM approach will allow for the two observations (admission and discharge from the NICU) by treating each participant’s breastfeeding self-efficacy score and breastmilk feeding rate at each assessment point as single data points [44]. There will be two levels of analysis in the data: Level 1 will refer to the outcome variables of breastfeeding self-efficacy and breastmilk feeding rates; while Level 2 will be the maternal or subject effect. We will enter the subject effect into the model as a random effect to capture within and between-subject variation [44]. The group effect of intervention/control will be entered as fixed effect to determine if the intervention is effective.

Using a variation of purposive, maximum variation sampling [49, 50], we will select mothers who experience high positive mean differences in their breastfeeding self-efficacy scores between NICU admission and discharge. To capture experiences that may have worked to detract from maternal breastfeeding self-efficacy, we will also select mothers with high negative mean differences. Assuming a normal distribution, we will sample from the top and bottom 20% of the change score distribution and sampling will continue until data saturation is achieved [50].

We will conduct semi-structured telephone interviews (Additional file 2). The geographical dispersion of participants in the FICare study renders face-to-face interviews unfeasible, and telephone interviews will allow data collection to occur with minimal expenditure and time commitment. Telephone interviews are an effective and efficient approach to qualitative data collection and provide results similar to face-to-face interviews [51]. Telephone interviews may also help to reduce social desirability response bias and interviewer effects, which may be more prevalent in face-to-face interviews [51]. Interviews will be conducted until informational redundancy is achieved and no new topics or concepts are emerging with additional interviews, with a projected sample size of 15 [52]. We will allow participants to guide the conversation, with minimal probing and redirection to maintain the content of the interview within the context of the themes requiring verification.

We will conduct thematic analysis within a constructionist framework, focusing on the sociocultural context and structural conditions (such as NICU policies, physical space and relationships with practitioners) that inform infant feeding experiences [46]. Employing theoretically driven coding may not fully capture maternal experiences with infant feeding while in the NICU. As such, we will use an inductive approach to thematic analysis [46, 53] to enable themes to be developed that are linked to the data and not to a previously determined theory. The inductive technique of data analysis will allow for codes and sub-codes to describe themes as they are observed in the data [54]. Inductively driven coding is constructed from the raw data and is interpreted by the researcher [53]. This may result in the development of themes that are not directly related to the pre-determined qualitative research question [53]. Coding will occur through a three-step process. The first step will be to read through transcripts to find codable moments that emerge from patterns in the data. Once the pattern is identified, we will move on to the second step of classifying or encoding the pattern by giving it a label [53]. Finally, once the data has been sufficiently coded and we have reached saturation, we will interpret the themes using thematic networks [55]. Selecting mothers based on their mean difference breastfeeding self-efficacy scores will situate breastfeeding self-efficacy as the criterion reference [53]. We will be able to compare-and-contrast themes that emerge from infant feeding experiences and identify observable differences between mothers with low and high breastfeeding self-efficacy change scores [53]. We will bracket our assumptions of the breastfeeding self-efficacy theory [56] and allow for themes to emerge as they are related to the concept of maternal experiences with infant feeding in the NICU. Bracketing is a reflexive process that involves preparation, action, evaluation, and systematic feedback [57] regarding thought processes and data analysis techniques throughout the coding process. Bracketing will be conducted through acknowledging our assumptions of, and affinity for, the breastfeeding self-efficacy theory and conducting reflexive journaling [56].