Ethic statement

Nurse managers at the NICUs and the mothers of the preterm infants were informed that participation was voluntary. A written consent was signed by nurse managers from each participating NICU and another by each participating mother. The study was conducted in accordance with the Declaration of Helsinki [32] and approved by The Danish Data Protection Agency (Journal number2012-58-0004). It was pointed out to the participating neonatal nurses that breastfeeding support should focus on supporting mothers in fulfilling their own goals for breastfeeding, since it will always be the mother’s individual decision whether she will breastfeed and for how long.

Definition of terminology

Exclusive breastfeeding was defined as an infant solely feeding directly at and from the breast [33] on the day of discharge and could include medication and vitamins. This definition does not count for what or how the infant has been fed before exclusive breastfeeding was established. Partial breastfeeding included other feeding methods in addition to breastfeeding (i.e. bottle, cup, lact-aid) regardless of content. Breastmilk feeding included breastfeeding and breastmilk given by other feeding methods.

Discharge was defined as the end of the total period of care from the NICU. If the infant was in an Early Discharge Program, the discharge date was defined as the day the Early Discharge Program ended.

Postmenstrual age is gestational age (GA) plus time elapsed from birth.

All participating neonatal and paediatric wards provided respiratory support (nasal CPAP or mechanical ventilation) for new-born infants and are for practical reasons referred to as NICUs.

Design and settings

A multi-centre study using a quasi-experimental intervention design was conducted. From 2016, three phases were performed: 1) a pre-intervention phase with participation of preterm mother-infant dyads born from October 2016 to July 2017 serving as a control group, 2) an intervention phase including a training program for neonatal nurses, implementation of weekly breastfeeding meetings for neonatal nurses, and optimizing their clinical practices, and 3) a post-intervention phase with participation of preterm mother-infant dyads born from February to December 2018 serving as an intervention group.

In Denmark, health care is free of charge. Parents in Denmark have the right to paid parental leave for a total of 52 weeks, 18 weeks solely for the mother and 2 weeks solely for the partner. The reminding weeks can freely be shared between the parents [34]. About 97% of Danish infants initiate breastfeeding [8] and 61% are exclusive breastfed for four months [35]. Preterm infants less than 35 gestational weeks are always admitted to a NICU, whereas infants born ≥35 gestational weeks are only admitted to a NICU if they need additional treatment and care. Most of the Danish NICUs have an early discharge program with transition from tube-feeding to breastfeeding at home and with contact to the NICU twice a week. In general, preterm infants are hospitalized or in early discharge program until exclusive breastfeeding is established or given up and mixed or bottle-feeding established. Guidance of mothers in establishing breastfeeding is a core element of neonatal nursing in Denmark.

Participants

All 17 Danish NICUs and one children’s department, which as a routine cared for preterm infants during breastfeeding establishment, were invited. Ten NICUs participated in the whole study, among them all four high intensive NICUs (level 3B and 3C). Three NICUs refused to participate due to other ongoing intervention studies with breastfeeding as outcome, two NICUs refused because they had other priorities. One NICU participated only in the pre-intervention, as they could not allocate resources for the training program and the post-intervention, two NICUs participated only in the intervention phase and post-intervention phase as they had to complete other ongoing studies first. The 13 participating NICUs had a mean of 18.7 beds and 54 neonatal nurses (range 24–140), all NICUs had access to donor human milk, all but one had an early discharge program. No hospitals held a valid BFHI designation as the program was closed in Denmark in 2008. Rooming-in was possible in all NICUs, in eight the mother could room-in for the infant’s whole NICU stay, in three NICUs at least 50% of the infant’s NICU stay and in two NICUs at least the last night before infant discharge.

Inclusion criteria for participating in the control and intervention groups were preterm infants (<37 gestational weeks) admitted to a NICU within five days of birth. Exclusion criteria were infant discharge from NICU to maternity or paediatric wards other than neonatal, mothers of preterm infants not capable of reading the questionnaires in Danish with or without help from the family, or history of drug abuse, which precluded the recommendation to breastfeed. Mothers of preterm infants who were particularly vulnerable e.g. giving infant away for foster care, psychiatric problems, or too stressed by the infant’s severe conditions were not invited to participate in the study. In the control group, infants could be transferred to non-participating NICUs and the mothers were still sent a questionnaire at the time of infant’s discharge to home. In the intervention group, infants had to be admitted to a participating NICU within the first days and discharged to home from a participating NICU.

Intervention

Data collection

In the pre- and post-intervention phases two questionnaires adapted from our previous study of breastfeeding preterm infants were used with few revisions and pilot tested by one mother [7]. Content in first questionnaire: Demographic data on infant and mother (infant GA, birth weight, maternal age, education), experiences with and plans for breastfeeding, first breastfeeding attempt, breast milk expression, and skin-to-skin contact. Content in second questionnaire: Timing of and method at full oral feeding, feeding at discharge (exclusive breastfeeding, partial breastfeeding, or no breastfeeding), breastfeeding problems, reasons for not establishing exclusive breastfeeding, use of nipple shield, pacifier, and test-weighing, continued skin-to-skin contact and breast milk pumping, rooming-in, early discharge programs, breastfeeding support, and infant respiratory support (see questionnaires in full in S3 and S4 Files).

The questionnaires were available in an online format that automatically excluded or included more questions depending on the mothers’ answers. The first questionnaire was sent to the mother approximately one week after delivery, the second at the infant’s discharge to home. Mothers were reminded to answer the questionnaire alternately by text message and e-mail at least three times. To improve participation rate in the intervention group, most of the mothers who did not answer the questionnaire within three reminders were called by phone to complete the questionnaire as a telephone interview. The two questionnaires included a maximum of 43 and 62 questions, respectively.

To monitor the process of the intervention, the number of nurses participating in the training was registered for each NICU together with the total number of nurses working in the NICU. In each NICU a chart was filled out by the person leading the breastfeeding meeting and included the date, topics discussed, number of participants, and length in minutes. The concept was pilot tested in one NICU.

Two or more contact nurses were appointed in each participating NICU. They registered all preterm infants admitted to the NICU during the study periods, informed and included mother-infant dyads in the study, and performed and registered the breastfeeding meetings for neonatal nurses. Encouraging and supportive e-mails were sent to the contact nurses and nurse managers at two to three months intervals to improve adherence to the study. The contact nurses could contact the principal investigator with questions and need for support.

Process and outcome variables

The process variables were the proportion of neonatal nurses that have gone through the training program in each ward and registration of the breastfeeding meetings (frequency, length, number of participants, and topics).

The primary outcome was exclusive breastfeeding rates at discharge from the NICU to home and changes between control and intervention groups. Secondary outcomes were changes in postmenstrual age at establishment of breastfeeding and changes in the frequency of maternal self-reported use of the six clinical practices: nipple shield use, test-weighing, minimizing the use of pacifier, daily skin-to-skin contact after incubator care, early breast milk expression, and mother-infant rooming-in (staying overnight in same room) for the whole NICU stay.

Statistics

To detect an expected difference in exclusive breastfeeding rates from 68% to 76%, 500 infants should be included in the control group and 500 in the intervention group (p = 0.05, power 80%). Rates from a former study of exclusive breastfeeding in Danish NICUs were used [7]. Although the inclusion period for the control group was extended due to lost cases, the number of infants with information on exclusive breastfeeding was 421. Based on the exclusive breastfeeding rate in the control group (58% of 421 infants), a new power calculation was performed estimating that with a clinically significant increase of 9 percentage points the intervention group should include 488 infants (p = 0.05, power 80%).

Descriptive statistics (numbers and percentages) were used to describe characteristics of infants and mothers in the control and the intervention group, respectively. To detect differences in characteristics, primary, and secondary outcomes between groups, Pearson’s Chi-Square test for categorical variables and One-way ANOVA for continuous variables were applied. All infants in the control group were included in the analyses irrespective of the ward’s participation in the intervention, and all infants in the intervention group were included in the analyses irrespective of the wards’ implementation of breastfeeding meetings. Postmenstrual age at establishment of exclusive breastfeeding was calculated from the date the infant did not get any feeds with any methods other than breastfeeding.For the outcomes exclusive breastfeeding at discharge from the NICU and postmenstrual age at establishment of exclusive breastfeeding, comparisons were made between infants in the intervention and in the control group. Similarly, for the clinical practices; use of a nipple shield, minimizing use of pacifier during breastfeeding establishment, and daily skin-to-skin contact after incubator care, comparisons were made between infants, whereas when assessing changes in first breast milk expression < 6 hours after birth, rooming in of mothers and using test-weighing at most breastfeeds, comparisons were made between mothers. To test correlation between the proportions of neonatal nurses participating in the training in each NICU and the changes in breastfeeding rates from control to intervention group, a Scatterplot was performed, and Pearson’s correlation coefficient was calculated. The same approach was used to test correlation between changes in maternal self-reported use of clinical practices and how frequent the practices were discussed at breastfeeding meetings.

A number of sensitivity analyses were performed. To account for a potential dependency between observations all analyses were repeated with only one infant per mother. Furthermore, the analyses assessing clinical practice related to the mother were repeated with the number of infants instead of mothers. To ensure comparable values between NICUs of different sizes and different ways of performing breastfeeding meetings, number of participants were added for all breastfeeding meetings in each NICU and divided with number of neonatal nurses working in the NICU resulting in an attendance nurse ratio. Total duration of breastfeeding meetings was calculated per nurse working in each NICU.

A post-hoc analysis was performed after a significant difference was found between numbers of extremely preterm (<28 weeks) infants in the control and in the intervention group; to test the Linear-by-Linear Association in the exclusive breastfeeding rates among GA groups in the intervention and the control group exclusive breastfeeding rates at discharge were analysed between gestational age groups using the Mantel-Haenszel test of trend.

SPSS version 25 was used for relevant statistical analyses. Values of p <0.05 were considered statistically significant.

Exclusive breastfeeding improvements

Our findings indicate that improvements in exclusive breastfeeding outcomes can be achieved by training in evidence-based practice. Similarly, the training programs in the Baby-friendly Hospital Initiative (BFHI) include several of the six clinical practices; rooming-in, early breastfeeding (lactation), skin-to-skin contact, and pacifier use, and have earlier been shown to improve breastfeeding rates [36]. A systematic review of 58 studies and reports found that compliance with the BFHI Ten Steps had a positive impact on breastfeeding outcomes. They also found a dose-response relationship between the number of implemented BFHI steps and early breastfeeding initiation and exclusive breastfeeding at hospital discharge [36]. Also, breastfeeding (and breastmilk feeding) rates in NICUs has formerly been found to increase by implementing multicomponent breastfeeding programs in NICUs or by implementing BFHI in the maternity wards [37–39]. The improved breastfeeding rates in our study are additionally supported by a meta-analysis of 27 randomized controlled trials which found positive effect of breastfeeding promoting interventions on exclusive breastfeeding rates up to six months age, especially the multicomponent BFHI, other combined interventions, and interventions involving healthcare professionals [40]. However, in the present study it cannot be verified if or which of the three improved maternal self-reported practices: early breastmilk expression, continued skin-to-skin contact, or less use of nipple shield contributed most to increase the exclusive breastfeeding rate at discharge. An association between improved practices and improved breastfeeding has previously been found in other studies: A quality improvement study from the U.S. showed that improved counselling of mothers of very preterm infants decreased the time to first breast milk expression and increased the proportion of infants receiving exclusively maternal breast milk at 28 days of life. The study did not, however, investigate the association with exclusive breastfeeding at discharge [17]. Furthermore, a Brazilian pre-and post-intervention study showed significant increase in exclusive breastmilk feeding rates after implementation of Kangaroo mother care/skin-to-skin contact in a NICU, but they did not investigate exclusive direct breastfeeding [41]. To our knowledge, no studies have been performed on how to reduce the use of nipple shields in preterm infants, nor whether a reduction in use would influence exclusive breastfeeding at discharge. Our study suggests an association between training nurses in breastfeeding-supportive practices, improved maternal self-reported use of the practices in the NICU, and improved exclusive breastfeeding in preterm infants.

Rooming-in and minimizing use of pacifier, which have previously been associated with earlier establishment of exclusive breastfeeding [14], did not improve after the training. Even though all NICUs had some available beds for mothers, most did not increase the number of mothers staying overnight in the same room as their infant during the whole NICU stay. Continued daily skin-to-skin contact increased from 83% in the control group to 88% in the intervention group. As the rate of daily skin-to-skin contact was already high, this increase may not have had much impact on postmenstrual age at establishment of exclusive breastfeeding. We were not able to find a decrease in postmenstrual age at establishment of exclusive breastfeeding. This may partially be due to ceiling effect of skin-to-skin contact and may partially be influenced by the failure to improve minimization of pacifier use and the rate of rooming-in, and partly by the limitations of a quasi-experimental design.

Training of neonatal nurses

The training program in the present study consisted of 7 hours which was less than other intervention studies that have shown to improve breastfeeding rates: Nilsson et al. used 11 hours in their training program [42] and Lowson et al. 7 days [31]. We believe that actively involving the neonatal nurses in the training sessions and integrating their ideas from the brainstorms on how to improve the six clinical practices in their NICU strengthened their active participation and motivation to optimize guidance in the practices and was important to the increase in breastfeeding rates. Change in practice must grow from the existing environment to succeed, and staff reflection as part of an education program has previously shown sustainable success when implementing skin-to-skin contact [43]. A UK multicentre study in 18 neonatal units also used an approach tailored to the needs of the staff in the individual neonatal unit and succeeded in increasing skin-to-skin sessions and breast milk feeding [31]. The correlation between neonatal nurse participation in the breastfeeding training program and improvements in breastfeeding in NICUs points out the importance of including all health care staff in the training when implementing evidence-based practice. We found a tendency but not significant correlation between how frequently the six practices were discussed at the breastfeeding meetings and the improvement of the six practices. The lack of significance could be due to the small number of practices. A clinical effect could still be possible. We believe the reflection on own practice during breastfeeding meetings had an effect in adjusting and maintaining optimal practice.

Surprisingly, the exclusive breastfeeding rate at discharge in the control group was only 58%. Although we had not expected a decrease in exclusive breastfeeding in the Danish preterm population from 68% reported in 2009–2011, an even more profound decrease from 59% to 45% has been found in Sweden in the period 2004 to 2013 [7]. This reinforces that improving breastfeeding rates needs constant focus.

Strengths and limitations

The strengths of the present study are the high number of participants, the multicentre design, and the adherence to the pre-post-design: No mother-infant dyads in the control group were exposed to the intervention and all mother-infant dyads in the intervention group were from NICUs participating in the intervention phase. Another strength was the narrow definition of breastfeeding as exclusive and direct breastfeeding. Clinical practices such as nipple shield use and pacifier use could have an impact on the infant’s ability to establish exclusive breastfeeding whereas the mother’s milk supply might not be affected. As a result, exclusive breastfeeding was the primary outcome of interest rather than exclusive breastmilk feeding or any breastfeeding. The clinical practices were not measured according to whether nurses delivered information to mothers, but rather according to whether mothers adhered to the recommended practices. The intervention was not regarded as implemented until it resulted in optimal practice reported by mothers. Thus, compliance with the six clinical practices was only measured by mothers’ questionnaire responses.

In a quasi-experimental design, we cannot eliminate that other changes could cause the difference. However, the rather short distance between pre-and post-intervention phases increases the likelihood of the intervention to cause the increase in breastfeeding rates. Randomizing mothers-infant dyads to control or intervention would have been preferable but not possible, as a spillover effect could not be avoided. Improvement in breastfeeding rates between control and intervention groups could have been due to fewer extremely preterm infants in the intervention group who are known to be breastfed less [8] as we also found. However, subgroup analyses showed that exclusive breastfeeding rates increased in all four GA groups, why the improvement between control and intervention groups was not only explained by fewer extremely preterm infants. Not all NICUs participated in both pre- and post-interventions and differences in breastfeeding rates between NICUs could bias the results. Nevertheless, in the 2009–2011 cohort, the breastfeeding rate in Danish NICUs participating in the later pre-intervention was not significantly different from those not participating in the pre-intervention (66.4 vs. 69.4, p = 0.218), and the same applied for NICUs participating/not participating in the post-intervention (66.9 vs. 69.0, p = 0.405). We showed in this study that in a context with relatively high breastfeeding rates, exclusive breastfeeding rates were improved in the intervention group, while the rates of no breastfeeding were not improved. It is not known whether the present intervention would only improve exclusive breastfeeding or also no breastfeeding in contexts with lower breastfeeding rates.

Even though efforts were made to increase response rate, several participants did not complete the study. However, the response rates were similar between control and intervention groups (74% and 73%, respectively). Some participants who answered both questionnaires did not answer all questions and consequently numbers of participants included in each calculation differs. The question about test-weighing had the most missing values (10.8% missing, primarily from the second twin) but we have no reason to believe that the missing answers were systematic and would have affected the results substantially.