Development of the breastfeeding support scale to measure breastfeeding support from lay and professional persons, and its predictive validity in Japan

Item development

Based on the literature on social support and breastfeeding support, the authors conceptualized a framework for the scale. That framework consists of support sources and support type (Table S1). Previous studies in Japan suggested that socio-environmental factors influence breastfeeding. Those factors include appropriate professional support, support from the husband, and the option for mothers to take paid maternity leave for more than 6 months (Inoue et al., 2012; Kaneko et al., 2006). To reflect the fact that breastfeeding is often influenced by socio-environmental factors, we adopted the ecological model (Institute of Medicine, 2000) when considering the source of breastfeeding support. Specifically, we considered that breastfeeding support is provided by social networks that include family, close friends, peers, and health professionals, as well as by mass media and others. In Japan, the marketing of formula milk for babies is not regulated by law, and companies “support” breastfeeding through websites, social networking services, childcare magazines, and face-to-face counseling. Considering the possibility that mothers recognize them as a source of information to support their breastfeeding, we included companies that produce and sell formula and related goods as a possible source of support. Regarding the types of support, the theoretical concept of social support described by House was also considered (House, 1981). Social support is defined as the help provided through social relationships and interactions. House distinguished among four main types of social support: emotional (expressions of empathy, love, trust, and caring), instrumental (tangible aid and services), informational (advice, suggestions, and information), and appraisal (information that is useful for self-evaluation) (House, 1981).

Candidate question-items were developed with reference to the literature on social support, breastfeeding support, and factors associated with breastfeeding, and also on the basis of two of the authors’ (KN and HH) clinical experience in supporting lactating women. To develop items that reflect support from each support source and type, the first author purposively referred to the literature on factors associated with breastfeeding among women who are healthy, live in high-income countries, and have a single healthy full-term baby (Dennis, 2002; Kaneko et al., 2006; Labbok & Taylor, 2008), and also reviewed other literature on breastfeeding support (Brown, Raynor & Lee, 2011; Burns et al., 2010; Hannula, Kaunonen & Tarkka, 2008; Hughes, 1984; Ito, Fujiwara & Barr, 2013; Matich & Sims, 1992; Schmied et al., 2011). Table S2 summarizes the findings of each study reviewed and the items created based on those findings. Reports by Japan’s government on the feeding of infants and young children were also reviewed (Equal Employment Children & Families Bureau, Ministry of Health, Labour & Welfare, 2006; Ministry of Health, Labour and Welfare, 2011). Those reports mentioned that 3-to-5 months postpartum was a common time for mothers to be concerned about infant feeding, but they had no specific information regarding needs for breastfeeding support. Studies on breastfeeding support for working mothers were not included in the literature review because we studied new mothers in the general population, most of whom either were not employed or were on maternity leave.

In addition to the question-items derived from the literature review, the first and the third authors added candidate items on the basis of their experience in clinical pediatrics (KN) and as International Board-Certified Lactation Consultants (KN and HH). Item 6 “I have received free samples or discount coupons for formula milk” and item 20 “In medical facilities I see posters or logos about formula milk” were developed from the idea that free samples and logos at health facilities and elsewhere might make mothers feel that formula is a normative standard supported by health professionals and others while exclusive breastfeeding is considered an extreme choice. Item 18 “I have heard that I must not take any medicine while breastfeeding” and item 22 “I have heard that there are certain things I should not eat while breastfeeding” were created because the authors notice that lactating mothers are often advised to avoid all medications and to avoid specific foods, and that this misinformation is provided both by health professionals and by others. Item 28 “When I leave my baby with other people, I think it will be problematic for them if my baby does not accept a bottle of formula milk” was developed because the authors noticed that mothers often use bottles of formula milk in the belief that doing so reduces the burden on other caregivers, a belief that might reflect inappropriate or insufficient breastfeeding support.

The items were developed in Japanese. A total of 31 items were generated to provide adequate redundancy within each of the four types of support listed above. Some items were positively worded and others were negatively worded. Response choices used a 5-point Likert-type scale from 1 (not agree) through 5 (agree). Those 31 items can be found in Table S3, with English translations.

Content validation testing

As recommended by Fitzner to improve item validity, the items were reviewed by a panel of experts to assess whether they adequately covered each of the concepts to be measured (Fitzner, 2007). The panel comprised five specialists, all of whom were university professors in nursing or midwifery who had clinical experience supporting breastfeeding mothers. Content validity was tested quantitatively by using the Content Validity Index (CVI). CVIs can be computed for each item in a scale (item-level CVI) as well as for the scale as a whole (scale CVI) (Polit, Beck & Owen, 2007). To calculate item-level CVIs, each member of the panel of experts rated each item independently on a 4-point Likert-type scale, with higher ratings indicating greater relevance. For each item, the item-level CVI was computed as the number of experts who gave a rating of either 3 or 4, divided by 5 (the total number of experts). The scale-level CVI was assessed by two methods: the proportion of items that was rated either 3 or 4 by all five experts (universal agreement method), and the average of all item-level CVIs (averaging method).

Item-level CVIs for each item are shown in Table S4. The scale-level CVI for the 31 items was 0.94 by the universal agreement method, and 0.99 by the averaging method, both of which are above the recommended minimum of 0.90 (Polit, Beck & Owen, 2007). Two items for which the CVI was 0.80 were modified after discussion with the expert panel, and were kept for further analysis.

In addition to being asked to rate the items for the CVIs, the experts were also asked to give comments and suggestions regarding clarity and readability of each item. Those comments and suggestions led to minor editorial revisions.

Pilot testing

The items were then pilot-tested about a month prior to the main survey with a convenient sample of 27 healthy mothers who visited a public health center in Tokyo’s Adachi Ward for their infant’s 3-month health check-up. After the participants completed the instrument, which took each mother about 5 min, the first author interviewed them individually to evaluate understandability, face validity, and the instrument’s format. Based on feedback from the first 14 mothers, the wordings of eight items were slightly modified. Then 13 other mothers completed the revised scale, and they found all the items to be clear. None of the mothers noted any problem with the format.

Design and participants

This was a longitudinal study with a survey 3 months after birth and a follow-up survey 2 months later. After the pilot study, we recruited participants for the main study. Mothers who visited any one of four health centers in Adachi Ward from October to December, 2014 for their infant’s 3-month health check-up were invited to participate in the study, if they were at least 18 years old, had a singleton infant, and were fluent in Japanese. Adachi Ward is one of the 23 wards of central Tokyo, and is primarily a residential area. The healthy life expectancy of women in Adachi Ward was the shortest among the 23 wards in Tokyo in 2016 (Toukyouto hukushi hoken kyoku, 2019) and it has no Baby-Friendly certified hospital. Adachi Ward had five health centers but one was not included in this study because it did not have enough space for conducting the survey. Among 414 mothers who were approached, 376 (91%) consented to participate in the study. Mothers were excluded if they or their infants had a medical condition that could significantly interfere with breastfeeding. Those conditions included preterm birth (n = 18), low birth weight (n = 22), macrosomia (n = 3), congenital abnormality of the infant (n = 9), and a history of admission to an NICU (n = 3). In addition, 20 mothers were excluded because they reported that during pregnancy they had intended formula feeding, and 7 mothers were excluded because they visited the health center earlier or later than expected (their infants were younger than 3 months or older than 5 months). Some mothers met more than one exclusion criterion. After the exclusion criteria described above were applied, 243 mothers participated to the study.

All the data were collected by using self-administered questionnaires. The first survey was conducted when the participants were waiting for their appointment at a health center for their infants’ 3-month health checkup. The follow-up survey was conducted by postal mail. Among the 243 mothers who completed the first survey, 177 (73%) returned the follow-up survey.

Psychometric testing

Psychometric testing included factor analysis (described in more detail below) and computation of internal-consistency reliability (coefficient alpha). For validation tests, we hypothesized that higher scores on the breastfeeding social support scale, indicating more support, would be associated with better family functioning and with better infant-feeding status at 3 and 5 months. Family functioning and infant-feeding status were measured as described below.

Family Apgar

As a construct-validation test, we measured family functioning and computed its correlation with scores on the new breastfeeding social support scale. The Japanese version of the Family Apgar scale was used to measure family functioning. It comprises five question-items regarding family adaptation, partnership, growth, affection, and resolve. Responses were on a 3-point (0, 1, and 2) Likert-type scale, yielding minimum and maximum total scores of 0 and 10. Higher total scores indicate better family functioning. Internal-consistency reliability (coefficient alpha) for the Family Apgar scale was 0.77 in this study.

Infant-feeding status

Information on infant-feeding status 3 months and 5 months after birth came from the mothers’ reports. Mothers were asked which of the following six options best described their infant-feeding method over the previous 24 h: (1) full breastfeeding (exclusive and almost exclusive breastfeeding, which means that no formula milk is given), (2) high partial breastfeeding (breastfeeding for more than 80% of all feedings), (3) medium partial breastfeeding (breastfeeding for 20–80% of all feedings), (4) low partial breastfeeding (breastfeeding for less than 20% of all feedings), (5) token breastfeeding (occasional breastfeeding, not for nutritive purposes), and (6) formula feeding (formula feeding only) (Labbok & Krasovec, 1990; Labbok & Coffin, 1997).

Analysis

Factor analysis was done to determine the factor structure and to identify items that contribute to the factors. After the structure and the items to be included in the scale were identified, the reliability and the validity of the scale were tested. To assess reliability, coefficient alpha was computed for the scores on the scale as a whole and for the scores on each subscale. Regarding validity, three hypotheses were tested. First, the total score on the scale was hypothesized to correlate (Pearson’s correlation coefficient) with the score on the Family Apgar scale. Second, the total score on the scale 3 months postpartum was hypothesized to be associated with the infant feeding status at that time, and also with the infant feeding status 5 months postpartum. The Kruskal–Wallis test was used to test that hypothesis. That test was used because it is a non-parametric one-way analysis of variance (ANOVA). It uses the ranks of the data rather than their absolute values. This has the potential advantage of not assuming that residuals are normally distributed, so it may be applicable even when the data do not meet that condition. Because it tests for equal medians among all groups, the results of the Kruskal–Wallis test are interpreted in a way that is analogous to the way in which the results of the more commonly used parametric ANOVA are interpreted. The difference is that with the Kruskal–Wallis test the interpretation refers to medians whereas with parametric ANOVA the interpretation refers to means. For example, a p value less than 0.05 would indicate that the median of the population of one or more of the groups is not the same as the median of the population of one or more of the other groups. The effect size (eta-squared) was computed as chi-squared/(n − 1). Finally, the area under the Receiver Operating Characteristic curve was calculated to assess the degree to which the total score and each subscale score 3 months postpartum could be used to predict which mothers would be in a low category of infant-feeding status 5 months postpartum.

Ethical considerations

Ethical approval was obtained from the Research Ethics Committee of the Graduate School of Medicine at the University of Tokyo (Ethical Application Ref: 10620), and written informed consent was received from all participants.

Characteristics of participants and infant-feeding status

Characteristics of participants are shown in Table 1. Their mean age was 31.7 (SD 5.0) years. Among the 243 participants, 197 (81.1%) had a vaginal delivery, 144 (59.3%) were primiparas, 157 (65.6 %) indicated that during pregnancy they had intended to breastfeed exclusively, and 84 (34.6%) had previous experience of any breastfeeding for more than 5 months. Fourteen mothers (5.9%) were working at 3 months postpartum and an additional 12 (4.9%) had a plan to return to work outside the home before 6 months postpartum. Compared with those who returned the follow-up survey, those who did not return it were younger, less educated, and more likely to be single mothers. However, there was no significant difference between the two groups in mode of delivery, intention to breastfeed, previous experience breastfeeding, parity, country in which the participant was raised, working status, or financial status.

Table 2 shows infant feeding status at the time of the survey. Three months postpartum, 137 (56.4%) reported full breastfeeding within the 24 h before the survey. Five months postpartum, that number was 109 (61.6% of those who responded). Not many were token feeding (4 (1.7%) at 3 months postpartum and 2 (1.1%) at 5 months postpartum) or completely formula feeding (22 (9.1%) at 3 months postpartum and 14 (7.9%) at 5 months postpartum). Among the 159 women who reported their feeding method both 3 months and 5 months postpartum, during the interval between those two reports 7 women increased the exclusivity of breastfeeding, while 22 either reduced the exclusivity of breastfeeding (n = 18) or stopped breastfeeding (n = 4).

Initial psychometric testing, and modifications based on those results

The mean and SD of each of the 31 items’ scores is shown in Table 3. The scores on items 16 and 27 had particularly low SDs.

The scree plot resulting from factor analysis of the 31 items (Fig. 1) clearly had two sections, which were separated by an “elbow” at factor number 4. Looking at the curve in Fig. 1 from left to right, the eigenvalues decrease very “steeply” between factor number 1 and factor number 3, but much more gradually starting at factor number 4. This indicated that the number of factors to be retained was three (Fig. 1). That is, much of the variability in the responses to the 31 questions could be explained by variability in only three underlying factors. This was not unexpected, given that the 31 questions were intended to ask about a smaller number of separate sources and types of breastfeeding support. Figure 1 was interpreted as meaning that the 31 questions were measuring three underlying variables. Next, with the number of factors restricted to three, after varimax rotation the items with factor loadings less than 0.4 were deleted: Those were items 6, 10, 11, 12, 13, 14, 15, 16, 18, 19, 20, 22, 23, 24, 25, 27, 28, and 30. With the remaining 13 items, the same factor analysis procedure was done again and items 21 and 29 were omitted. By that process, the only items retained were those that were best for measuring the three underlying factors.

The resulting scale had 11 items, and factor analysis with those 11 items confirmed that each of them had a loading greater than 0.4 on one, and only one, of the three factors (Table 4). Items 7, 8, 9, and 31 loaded strongly on the first factor. Because of the content that those four items had in common, they were considered to comprise a subscale measuring support from breastfeeding peers and from people in specifically named healthcare professions. Items 2, 5, and 26 loaded strongly on the second factor. Because of the content that those three items had in common, they were considered to comprise a subscale measuring practical help. Similarly, items 1, 3, 4, and 17 loaded strongly on the third factor, and they were considered to comprise a subscale measuring support from people the mother can rely on to help meet emotional needs and address breastfeeding concerns.

Reliability testing

Three months postpartum, coefficient alpha of the scores on the 11-item scale was 0.83, and 5 months postpartum it was 0.85. Three months postpartum, the alphas for the scores on the subscales were 0.78 (breastfeeding peers and named professions), 0.80 (practical help), and 0.81 (support to meet emotional needs). Five months postpartum, the alphas for the scores on the subscales were 0.79 (breastfeeding peers and named professions), 0.86 (practical help), and 0.81 (support to meet emotional needs).

Validation testing

We hypothesized that mothers in well-functioning families would have higher scores on the 11-item breastfeeding support scale. That hypothesis was supported by the positive correlation between the 11-item scale score and the Family Apgar score 3 months postpartum (r = 0.47, p < 0.001).

Table 5 shows the mean scale scores (11 items) of women in each category of infant-feeding status 3 and 5 months postpartum. Among the mothers who completed the scale 3 months postpartum, very few were token feeding (n = 2) or completely formula feeding (n = 3) 5 months postpartum, so those two categories were combined with the low-partial breastfeeding category (n = 6) to make a group of 11 mothers. Higher scores on the 11-item scale 3 months postpartum were associated with more breastfeeding exclusivity both at that time and also 2 months later (i.e. both 3 months and also 5 months postpartum). Specifically, the results of the Kruskal–Wallis test 3 months postpartum were chi-squared = 14.871 (df = 3, n = 211), p = 0.002, eta-squared = 0.071, and the results of the same test 5 months postpartum were chi-squared = 8.556 (df = 3, n = 159), p = 0.036, eta-squared = 0.054.

The area under the Receiver Operating Characteristic curve (AUC) was 0.73 (95% CI [0.57–0.88]), which indicates that scores on the 11-item scale 3 months postpartum may be useful to predict which mothers will be in a low category of infant-feeding status (i.e., low-partial breastfeeding, token breastfeeding, or formula feeding) 5 months postpartum. Regarding the subscales, the AUC was 0.78 (95% CI [0.63–0.94]) for the subscale measuring support from breastfeeding peers and from people in specifically named healthcare professions, 0.44 (95% CI [0.27–0.94]) for the subscale measuring practical help, and 0.74 (95% CI [0.60–0.88]) for the subscale measuring support from people the mother can rely on to help meet emotional needs and address breastfeeding concerns.