New mothers in Korea convalesce and adapt to motherhood for roughly 4.6 weeks at a combination of postpartum care centers (75.1%), their own homes (70.2%), their parents’ homes (19.8%), or their in-laws’ homes (2.4%) after being discharged from the hospital 3~5 days post-childbirth [1]. As new mothers receive professional postpartum services in the hospital for only a short time period, it is important for them to be able to assess their own adaptation to breastfeeding after being discharged, in order to detect any early problems and to find solutions with the help of professionals or other resources. The Breastfeeding Adaptation Scale (BFAS) was developed in 2009 for new mothers to self-evaluate their adaptation to breastfeeding [2]; numerous findings evaluating the BFAS have since been reported in nursing studies conducted in Korea [3-7].

The BFAS was developed by defining the concept of adaptation to breastfeeding based on Roy’s four modes of adaptation through a review of the literature on breastfeeding, existing tools for assessing breastfeeding, and an analysis of interviews with breastfeeding mothers and clinical nurses [2]. During development, the content validity, construct validity, and the internal consistency reliability of the BFAS were verified. As the construct validity of the scale was verified through exploratory factor analysis and known-group validity during development, the validity of the factors should be reexamined through confirmatory factor analysis. The validity of the scale is subject to change according to time and culture, making a continued validation process necessary. Over the last decade, additional verification has not been conducted for scale validation. In particular, the items about nipple problems may not be core items in a tool for assessing breastfeeding adaptation because wounds and pain of the nipple are rarely found at 4 weeks postpartum [8]. The importance of breastfeeding support is also decreasing, as mothers are becoming increasingly able to breastfeed themselves by themselves at 4 weeks postpartum [8]. Therefore, it is necessary to confirm whether the items of the discomfort in breastfeeding sub-dimension and the getting support sub-dimension are in fact necessary to evaluate breastfeeding adaptation. In addition, the reliability of the discomfort in breastfeeding sub-dimension and the getting support sub-dimension of the BFAS scale was low [2]. This study aimed to verify construct validity through confirmatory factor analysis as part of the validation process of the measurement scale, which has not been done previously, and to suggest a shortened scale comprising only core questions.

The original scale is applicable to breastfeeding women at 4 weeks from childbirth, known as the period of adjustment to breastfeeding [2]. Some studies applied this scale to women at less than 4 weeks after childbirth, for whom the validity and reliability had not been evaluated [5,9], and reported that the degree of adaptation to breastfeeding varied according to the postpartum period (1, 4, and 8 weeks) [5]. The validity of the scale has yet to be confirmed within the first 4 weeks postpartum, before the usual time point for assessing adaptation to breastfeeding.

It takes 2~6 weeks after childbirth for the mother to get acquainted with her baby, learn about her newborn, and physically recover, and 2 weeks to 4 months to move toward a new normal [10]. It takes roughly 10 days for transitional breast milk to mature, and for the established lactation to start to be maintained by the auto-secretion system [11]. It takes approximately 4 weeks to adjust to breastfeeding [8]. Despite individual variance, it generally takes 2 weeks from childbirth to adapt to breastfeeding, and in most cases, adjustment to breastfeeding occurs within approximately 4 weeks from childbirth. Thus, the process of adaptation to breastfeeding is completed in the first 2 weeks postpartum in the physiological, self-concept, role function, and interdependence modes, if it is rapid. The process may take more than 4 weeks in case of problems. Therefore, it is important for mothers to assess their level of adaptation to breastfeeding on their own, and to continue to do so after the initial adaptation process. For this reason, it is necessary to evaluate the validity and reliability of the scale used to assess adaptation to breastfeeding to determine its adequacy for use with mothers at 2 weeks from childbirth and to assess whether the longitudinal measurement invariance (MI) of the scale is maintained, making it suitable to compare adaptation to breastfeeding between the first 2 and 4 weeks.

The MI of a scale is essential for comparing the scores of variables across multiple groups defined according to different characteristics (age, gender, time point, etc.). The measurement scores between two groups can be compared more stably if consistency is present in the types of factors of the scale measuring a certain concept and each factor’s indicator variables, factor load, intercept, variance and covariance, and error variance [12]. Consistent results of adaptation to breastfeeding according to the age and education of breastfeeding mothers have been reported [3,6,13]. Conversely, inconsistent results of adaptation to breastfeeding or breastfeeding practices according to employment status [3,4,6,13,14], parity [3,4,6,14], and delivery mode [14,15] have been reported. If any multi-group difference is assumed without checking group invariance, errors may occur in interventions, resulting in unnecessary costs. The MI of the scale should first be verified to compare differences across multiple groups in adaptation to breastfeeding, and a latent mean analysis should be performed to compare the mean level of adaptation in order to obtain accurate results. This study reports the MI of the scale and the results of the latent mean analysis at 4 weeks postpartum. The construct validity, reliability, MI of the scale, and results of the latent mean analysis at 2 weeks postpartum will be reported in a subsequent study.

Therefore, this study presents the development of an abbreviated version of the BFAS (yielding the Breastfeeding Adaptation Scale-Short Form, or BFAS-SF), confirmation of its validity and reliability, and suggestions regarding the applicability of the BFAS-SF. Additionally, by measuring the MI of the BFAS-SF at 4 weeks postpartum, this study derived implications regarding future breastfeeding adaptation assessment. Furthermore, the latent variable means of the BFAS-SF were calculated and presented.

This study was conducted to develop the BFAS-SF, to assess its validity and reliability, and to verify the MI of the scale and differences in adaptation to breastfeeding. First, construct validity and internal consistency reliability of BFAS-SF at 4 weeks postpartum were verified. Second, the MI of the BFAS-SF was verified depending on the mother’s employment status, parity, and delivery mode at 4 weeks postpartum and according to length of postpartum period (2 vs 4 weeks). Third, if any intercept invariance according to employment status, parity, delivery mode, and length of postpartum period was found, between-group differences in the latent variable means of the BFAS-SF were confirmed.

This is a methodological research study conducted to develop a short form of the BFAS (the BFAS-SF) and to test its validity, reliability, and MI.

The target group comprised breastfeeding Korean women at 2 weeks postpartum (14~20 days) who voluntarily participated in the survey. These new mothers and babies had not been hospitalized for any disease after birth. The 4-week postpartum group comprised breastfeeding women who participated in the study by responding to the first survey.

For confirmatory factor analysis of the BFAS-SF, the minimum sample size was 109 individuals (Cronbach’s ⍺ was .05, degrees of freedom=296, power=.90, null root mean square error of approximation [RMSEA]=.00, and alternative RMSEA=.05) [16]. Furthermore, 218 individuals were needed to assess the MI between the two groups. Considering possible primary and secondary dropout rates of 30%, respectively, 500 was set as the sample size. Of the 500 questionnaires distributed, 444 completed questionnaires were returned; 431 questionnaires were used for data analysis, excluding incomplete questionnaires; and 272 individuals responded to the second survey.

Data for this study were collected from December 15, 2017 to May 31, 2018. Data for the first survey at 2 weeks postpartum were collected through convenience sampling at 10 postpartum care centers in Seoul, Daegu, Gwangju, Busan, and Gyeongsang Province. As outsiders are not allowed to enter postpartum care centers to keep infants safe from infection, the researcher hired an employee at each postpartum care center as a research assistant. The assistants provided an oral explanation on the study to the subjects at 2 weeks postpartum, and then distributed the questionnaires to those who agreed to take part in the study and collected their responses upon completion. Small gifts (5,000 Korean won) were given to the participants to express gratitude. In the first survey, the questionnaire gathered information on sociodemographic characteristics, obstetric characteristics, and breastfeeding adaptation.

The second survey was distributed online through social networking services or e-mail to those who agreed to data collection at 4 weeks postpartum and provided their phone number or e-mail address directly. The second survey reexamined the current health status of the baby and breastfeeding adaptation in mothers.

Values for the frequency and percentage of the subjects’ demographic characteristics were calculated, and the mean, standard deviation, skewness, and kurtosis of items and factors were assessed using SPSS version 25.0 (IBM Corp., Armonk, NY, USA). Cronbach’s ⍺ and 95% confidence intervals (CIs) were obtained to assess internal consistency reliability.

To verify construct validity, confirmatory factor analysis was performed using AMOS version 25.0 (IBM Corp., Armonk, NY, USA). The following goodness-of-fit indices were used to evaluate the fit of the models: chi-square (x²), the comparative fit index (CFI; cut-off>.900), RMSEA (cut-off<.060), and the standardized root mean residual (SRMR; cut-off<.0800) [17,18]. Items were deleted by based on a review of their standardized regression coefficient (cut-off<.50), squared multiple correlation (SMC; cut-off<.30), and modified indices (cut-off≥4.0), as well as by theoretically reviewing the content of the items [17]. To confirm the convergent validity of the BFAS-SF, standardized regression coefficients of the observation variables (cut-off≥50), construct reliability (CR; cut-off≥0.70), and average variance extracted (AVE; cut-off≥0.50) were examined. To confirm the discriminant validity of the BFAS-SF, the factor correlation coefficient (cut off≤.80) and the AVE of the latent variables and the squared correlation coefficient between latent variables (AVE of latent variable > squared correlation coefficient between latent variables) were examined [17].

Tests of the MI between the two groups according to employment status, parity, delivery mode, and the postpartum period were performed using the BFAS-SF. The test of the MI involved five steps. The first step was configural invariance (factor structure), the second step was metric invariance (factor loading invariance), the third scalar invariance (intercept invariance), the fourth step was factor variance invariance and factor covariance invariance, and the fifth step was error variance invariance. When MI was rejected at each step, partial MI was confirmed. A change in the model fit in MI testing is generally indicated by an increased x² value relative to the increased degrees of freedom, but the x² value was not used here because it is affected by sample size. MI was tested using change in CFI (△CFI) as the main criterion, and change in RMSEA (△RMSEA) and change in SRMR (△SRMR) as additional criteria [19]. If the total sample size is smaller than 300 or the sample size between the two groups is not even, as in this study, differences in fit between nested models can be evaluated using cutoffs of △CFI≤.005, △RMSEA≤.010, △SRMR≤.0250 (for metric invariance), and △SRMR≤.0050 (for scalar invariance and error variance invariance) as indicative of a nonsignificant decrease in fit across models [19].

When the scalar invariance was verified, the group mean differences in latent variables were tested.

All data were collected after Institutional Review Board (IRB) approval (approval number: CUIRB-2017-0079). The purpose of the study, data collection procedure, confidentiality and anonymity, and possibility of withdrawal were explained to the 2-week postpartum participants. An explanation was also given of the data processing procedure, in which data were encrypted, stored separately on a storage device, and completely deleted after the end of the study. The plan to conduct additional data collection at 4 weeks after birth was explained. Participants voluntarily signed an agreement to participate.

The general background characteristics of participants at 2 and 4 weeks postpartum are described in Table 1. The majority of the participants at 2 and 4 weeks postpartum were employed (51.4% and 52.2%, respectively). The proportion of primiparous participants at 2 and 4 weeks postpartum was 73.3% and 72.1%, respectively. In both groups, the majority of the participants had a vaginal birth (51.7% and 55.1%, respectively).

Confirmatory factor analysis was conducted on the 27 items and eight factors of the BFAS. The fit of the model was unacceptable: x²=511.02 (p<.001), degrees of freedom=296, CFI=.805, RMSEA=.086, SRMR=.0934. Considering the modified indices and model fit, items with a lower standardized regression coefficient or SMC than the cut-off point were deleted (items no. 6, 9, 15~18, 21~23, 26, 27). With the final six factors and 16 items, the standardized regression coefficients of all items were≥.56, with an SMC≥.32, with both significant factor variance and error variance (Table 2). The model fit was also revealed to be acceptable: CFI=.96, RMSEA=.06, and SRMR=.04, except for x²=168.78 (p<.001, degrees of freedom=89).

Upon verifying the convergent validity of the BFAS-SF, the CR of each factor was ≥.73; each AVE was also ≥.57. After examining the discriminative validity of the factors, the correlation coefficient between factors was found to be .25~.68, and the AVE values of all the latent variables were greater than the square of the correlation coefficient between the latent variables.

The names of the sub-dimensions of the BFAS-SF are as follows: factor 1, emotional exchange with one’s baby; factor 2, breastfeeding confidence; factor 3, sufficient breast milk; factor 4, baby’s feeding capability; factor 5, baby’s satisfaction with breastfeeding; factor 6, maintenance of breast milk volume.

The M±SD, skewness, and kurtosis of the items and M±SD of factors of the BFAS-SF are shown in Table 3.

Cronbach’s ⍺ (95% CI) for the overall scale was .88 (.85~.90). The Cronbach’s ⍺ values were as follows for the sub-dimensions: factor 1, .84 (.80~.87); factor 2, .67 (.59~.73); factor 3, .78 (.73~.82); factor 4, .75 (.69~.81); factor 5, .79 (.74~.84); factor 6, .75 (.68~.80) (Table 3).

In MI testing of the BFAS-SF according to mother’s employment status (being employed or not), the fit indexes were acceptable for configural (factor structure) invariance. Although metric invariance was not supported (with a △CFI>.005), partial metric invariance, partial scalar invariance, partial factor variance (covariance), and partial error variance invariance were achieved (with a △CFI≤.005 and a △RMSEA≤.0100), with 15 indicators (except for item no. 25, ‘trying to eat enough food and water for breastfeeding’) having equal factor loadings, item intercepts, factor variance (covariance), and error variance across groups (Table 4).

In MI testing of the BFAS-SF across parity (primipara or multipara), configural and metric invariance were achieved (with a △CFI≤.005, a △RMSEA≤.010, and a △SRMR≤.0250), with all indicators having equal factor structure and factor loadings (Table 4).

In MI testing of the BFAS-SF across delivery mode (vaginal or cesarean delivery), configural, metric, and scalar invariance were achieved (with a △CFI≤.005, a △RMSEA≤.010, a △SRMR≤.0050), with all indicators having equal factor structure, factor loadings, and item intercepts across groups (Table 4).

The longitudinal MI of the BFAS-SF at 2 and 4 weeks postpartum was examined. The fit indexes were acceptable for configural (factor structure) invariance. Although metric invariance was not supported (with a △CFI>.005), partial metric invariance was achieved (with a △CFI≤.005, a △RMSEA ≤.010, a △SRMR≤.0250), with 14 indicators (except for item no. 1, ‘my baby looks so lovely when he or she is drinking breast milk,’ and item no. 12, ‘my breasts feel full when it is time to feed my baby’) having equal factor loadings and item intercepts across groups (Table 4).

At 4 weeks postpartum, mothers who were unemployed had a 0.28-point higher (p=.018) average score for latent variable 2, a 0.38-point higher (p=.007) average score for latent variable 3, and a 0.32-point higher (p=.004) average score for latent variable 5 than those who were employed (Table 5).

Mothers who gave birth vaginally had a 0.24-point higher (p=.028) average score for latent variable 2, a 0.43-point higher (p=.002) average score for latent variable 3, and a 0.23-point higher (p=.043) average score for latent variable 5 than those who had a cesarean section (Table 5).