Design
This study was originally conducted as a randomized controlled trial (RCT) with longitudinal data collection that evaluated a proactive breastfeeding telephone support programme [11,12,13]. In this paper we have conducted secondary analyses of the longitudinal data.
Setting
The data in this study were collected between March 2013 and December 2016 as part of the original RCT study. Six neonatal units across Sweden participated in the study.
In Sweden, breastfeeding rates have decreased over the past 15 years. Nevertheless, the initiation rates for breastfeeding are relatively high from an international perspective. However, breastfeeding has been shown to decline rapidly over the first months after birth, especially exclusive breastfeeding [14]. In Sweden, all mothers receive parental benefits, which consist of 480 days per child, 90 of which are reserved for each parent and cannot be transferred. The parental benefit for 390 days is income-based to 80% of the parent’s wages, and for 90 days, the compensation is based on a minimal level [15].
In this study, exclusive breastfeeding was defined as feeding with mother’s milk only but could include medications, fortification and vitamins. Partial breastfeeding was defined as feeding with mother’s milk in combination with infant formula and/or solid food. No breastfeeding was defined as full formula feeding and/or solid food with no mother’s milk intake. The previous 24 h were used as the recall period. The term breastfeeding was used both for breastfeeding at the breast and for mother’s milk feeding regardless of feeding method [16].
Sample
Mothers of preterm infants (gestational age < 37 weeks) who were providing mother’s milk to any extent at discharge were invited to participate in the study. The exclusion criteria included serious maternal medical or psychiatric problems at discharge, language problems that could not be resolved, transfer of the infant to another hospital/unit after discharge, and infants that were terminally ill. In total, 493 mothers of 547 infants contributed data to this study, and a flowchart of eligible, included, excluded and lost to follow-up mothers is shown in Fig. 1. In the study population few mothers fed their infants expressed mother’s milk by bottle. Two mothers exclusively pumped at discharge and had ceased breastfeeding at 8 weeks after discharge. Some mothers stated that they gave their infants expressed mother’s milk by bottle occasionally, but otherwise breastfed direct at the breast. Hence, this study measured breastfeeding satisfaction in mothers who predominantly breastfed their preterm infants directly at breast.
Data collection and measurements
At each neonatal unit there was a breastfeeding support team consisting of staff from the neonatal unit who informed mothers about the study, asked the mothers for consent, and collected the demographic data close to the discharge. For detailed information on the original RCT study and recruitment see Ericson et al. [17]. All participating mothers signed a written informed consent form after receiving oral and written information and having had the opportunity to ask questions.
Demographic data included the following data for the mothers: age, educational level, parity, country of birth other than Sweden, multiple births and mode of delivery. The following demographic data were collected for the infants: gestational age at birth, birth and discharge weights, small for gestational age status, gender, neonatal illnesses and breathing support during hospitalization, length of stay and breastfeeding at discharge, i.e., exclusive or partial. Follow-up questionnaires were sent by post to the mothers at 8 weeks after discharge, and at 6 and 12 months after birth. The mothers were asked to fill out the questionnaires in relation to how they were feeling at the time regardless of breastfeeding status. The questionnaires included a question on breastfeeding status i.e., exclusive, partial or no breastfeeding and the following validated scales: the Maternal Breastfeeding Evaluation Scale (MBFES) [18], the Maternal Postnatal Attachment Scale (MPAS) [19] and the Swedish Parental Stress Questionnaire (SPSQ) [20]. It took about 20–30 min for the mothers to complete the entire survey.
The MBFES consists of 30 items in the following three dimensions: maternal enjoyment and role attainment, infant satisfaction and growth, lifestyle and maternal body image. Each item is scored on a five-point scale from strongly disagree to strongly agree. A summary score is calculated with a minimum of 30 and a maximum of 150, with higher scores indicating a more positive breastfeeding experience. In our study, the Cronbach’s alpha for the MBFES was 0.93 at all follow-ups. The MBFES is a valid and reliable instrument and was developed to measure aspects of breastfeeding that mothers identify as important in defining successful breastfeeding. The MBFES evaluates the breastfeeding experience, i.e., the mother’s satisfaction with breastfeeding. The three dimensions in MBFES could be used to assess important aspects of breastfeeding satisfaction compared to a single item [18, 21].
The MPAS consists of 19 items, which are scored on a two-, four- or five-point scale and have a total global attachment score with a minimum of 19 and a maximum of 95; higher scores indicate greater maternal-to-infant attachment. In our study, the Cronbach’s alpha was 0.78, 0.76 and 0.74 at 8 weeks after discharge, and at 6 and 12 months after birth for the MPAS, respectively.
Because there is no Swedish version of these two scales, translations were made according to the study by Wild et al. [22], which describes guidelines and standards for the translation and cultural adaptation of patient-reported outcome measures.
The SPSQ, which is a modified Swedish version of the Parenting Stress Index (PSI) [23], consists of 34 items. The items are scored on a five-point scale from strongly disagree to strongly agree. The total score is calculated as the mean of all responses. Higher scores indicate greater perceived parental stress. In our study, the Cronbach’s alpha for the SPSQ was 0.90, 0.89 and 0.90 at 8 weeks after discharge, and at 6 and 12 months after birth, respectively.
The participants’ personal data were obtained and stored in accordance with the European General Data Protection Regulation and the Swedish Ethical Review Act. All collected data were protected against unauthorized access in a secure database with daily back up. The results are presented at the group level, and no individual can be identified.
Data analysis
As the intervention in the original RCT study had no effect on breastfeeding satisfaction and exclusive breastfeeding [13], all mothers were included in the analyses in the present study, regardless of randomization group. Descriptive statistics of maternal and infant characteristics and the MBFES were calculated and are presented as numbers, percentages, means and standard deviations (SD) for normally distributed variables and as medians and interquartile ranges (IQRs) for non-normally distributed variables.
To assess associated factors and changes in breastfeeding satisfaction, linear mixed-effects modelling was applied to describe the association between each of the factors and the MBFES summary scores and the scores on each dimension separately over time. The linear mixed-effects model is a useful statistical analysis for longitudinal data as it accounts for the dependence among the repeated measurements on the same individual. It also has an advantage in handling missing outcome values in that an individual can be included in the analysis even if some outcome values are missing. Factors included in the model were maternal age (years), upper secondary school or less/higher education, mothers born in Sweden/other birth country, primipara/multipara, singletons/twins, infant gestational age < 32 weeks/32–36 weeks, breastfeeding at discharge (partial/exclusive) and breastfeeding 8 weeks after discharge and 6 and 12 months after birth (no/partial/exclusive). In the linear mixed-effects models, we used the AR (1) covariance structure with a fixed effect of time. The linear mixed-effects models included repeated measurements 8 weeks after discharge and 6 and 12 months after birth. The results from the linear mixed-effects model analyses are presented with estimates, which should be interpreted as the mean difference in MBFES scores between groups, and 95% confidence intervals (CIs) between the compared groups after adjusting for confounders. If the estimated value was positive, the breastfeeding satisfaction level was higher in the reference group; if the value was negative, the breastfeeding satisfaction level was lower in the reference group. The baseline confounders included in the regression models were based on previous research and clinical experience. To test the model, a likelihood ratio test was performed comparing the full model to a null model with a time variable as the only variable. The model fit was assessed visually by comparing the residual distribution to a normal distribution. Calculations were performed with IBM SPSS Statistics for Windows (version 26.0, IBM Corp., Armonk, NY).