A single Spanish version of maternal and paternal postnatal attachment scales: validation and conceptual analysis

Introduction

During the first years of life, newborns depend for their survival and development on their caregivers (Bowlby, 1969), who will provide the child’s environmental context (De Cock et al., 2016). During this period, the parent–child relationship is crucial for the child’s biological, cognitive, emotional, and social development (Belsky & Fearon, 2002; Feldman, 2007). One of the most important processes in the postpartum period is the postnatal bonding between the parents and the baby (Brockington, 2004, 2011).

In the current state of scientific literature, the conceptualization of postnatal bonding presents inconsistencies, both linguistic and epistemological, across authors and disciplines, generating confusion and disagreement in practice (Bicking Kinsey & Hupcey, 2013).

Linguistically, inconsistencies in the use and meaning of postnatal bonding are due to confusing the term with attachment, bonding and attachment being used interchangeably. Nonetheless, different authors have specifically differentiated the terms (Altaweli & Roberts, 2010; Kennell & McGrath, 2005; Taylor et al., 2005). For example, Taylor et al. (2005) comment that bonding is used to describe how the mother feels toward her baby; while attachment includes the behavior of the baby toward the mother. Epistemological inconsistencies are related to the directionality of the process, the domains that comprise it (affective, behavioral, and biological) and its temporality (Bicking Kinsey & Hupcey, 2013).

Despite these inconsistencies, there is some consensus about identifying the emotional component of postnatal bonding in its different definitions (Brockington, Fraser & Wilson, 2006; Condon & Corkindale, 1998; Condon, Corkindale & Boyce, 2008; Figueiredo et al., 2009; Taylor et al., 2005; Van Bussel, Spitz & Demyttenaere, 2010), in addition to being consistently operationalized by measuring instruments (Bicking Kinsey & Hupcey, 2013). Therefore, according to the above, bonding is the term adopted in this study, meaning the unique emotional tie established between parents and their baby in early childhood.

Insights from studies on the negative consequences of postnatal bonding disorders both for the infant and for the parents highlight the importance of this concept. First, concerning the consequences for the infant, some studies relate low postnatal bonding quality with complications in the child’s socioemotional development, increased parental perception of the infant’s difficult temperament (De Cock et al., 2016; Mason, Briggs & Silver, 2011), problems in the development of executive functions (De Cock et al., 2017), and externalizing behavior difficulties at 18 months of age (Hairston et al., 2011). Secondly, as to the consequences for parents, it was found that low-quality bonding predicts greater parental stress (De Cock et al., 2017; Mason, Briggs & Silver, 2011), higher anxiety levels, less perception of the partner’s support (De Cock et al., 2016), a negative effect on parenting skills and feelings of parental adaptation (Müller, 1994; Siddiqui & Hägglöf, 2000), as well as poorer quality of parent–child interactions, for example, maternal response to the baby’s signals (Hornstein et al., 2006).

Due to the importance of bonding and the consequences it can have on the infant and the parents, in recent years, specific instruments have been developed for professionals of nursing, psychology, pediatrics, and primary care to detect and prevent problems in early bonding. Easy-to-use instruments that do not need much time or require sophisticated infrastructure to evaluate qualitative information are required (Van Bussel, Spitz & Demyttenaere, 2010). In this sense, self-reports are the most useful and applied evaluation strategies, and the following three are the most extensively employed to evaluate early emotional bonding between the mother and the newborn: the maternal postnatal attachment scale (MPAS: Condon & Corkindale, 1998), the Postpartum Bonding Questionnaire (Brockington et al., 2001) and the mother–infant bonding scale (Taylor et al., 2005). These three instruments provide a reliable and valid indication of the early emotional bond between a mother and her newborn infant (Van Bussel, Spitz & Demyttenaere, 2010).

Although mother–infant bonding has received much attention over the last decade, the study of the father’s role in the context of early relations is still little addressed in research, despite studies that have determined its great impact on the infant’s development (De Cock et al., 2016, 2017; Ramchandani et al., 2013; Sarkadi et al., 2008). For example, Scism & Cobb (2017) have shown that father–infant bonding in the postpartum period reduces cognitive delay, promotes weight gain in premature infants, and improves the rate of breastfeeding. In addition, incorporating fathers into the study of postnatal bonding, allows analyzing this concept from a family system perspective.

Only one of the above instruments, the MPAS, is based on a theoretical framework that encompasses the father figure and provides an instrument adapted to fathers—the paternal postnatal attachment scale (PPAS: Condon, Corkindale & Boyce, 2008)—thereby incorporating a gender perspective in the study of postnatal bonding. Additionally, it has adequate psychometric characteristics, is easy to apply, and has been widely used in different studies (De Cock et al., 2016, 2017).

These two scales, the MPAS and the PPAS, draw on the conceptualization of bonding as an emotional state (“love”) of the parents toward their baby. This emotional state leads to a series of needs and dispositions directed toward the baby, which will determine paternal and maternal behaviors (Condon, 1993; Condon & Corkindale, 1998; Condon, Corkindale & Boyce, 2008). These dispositions or needs, according to Condon, Corkindale & Boyce (2008), differ in men and women. In women, they focus on the pleasure of proximity, tolerance, the need for gratification and protection, and the acquisition of knowledge (Condon & Corkindale, 1998). In men, the bonding indicators are patience and tolerance, pleasure in interaction, affection and pride (Condon, Corkindale & Boyce, 2008).

A review of the literature on the MPAS and the PPAS has identified several works related to the adaptation of these scales to other countries. The MPAS has been adapted for use in Italy (Scopesi et al., 2004), Belgium (Van Bussel, Spitz & Demyttenaere, 2010), Portugal (Nunes et al., 2014), the United States (Feldstein et al., 2004), and Iran (Ghadery-Sefat et al., 2016). With regard to the PPAS, it has been adapted for use in Portugal (Pires et al., 2014) and Turkey (Güleç & Kavlak, 2013). Various studies report appropriate values of reliability and validity but not all the studies have analyzed their structural validity (Van Bussel, Spitz & Demyttenaere, 2010; Feldstein et al., 2004; Ghadery-Sefat et al., 2016). Among those who have done so, inconsistencies have been found, as the three factor structure of the original scales could not be replicated (Scopesi et al., 2004; Nunes et al., 2014; Pires et al., 2014). In this sense, in the Italian version of the MPAS, six factors were obtained (Scopesi et al., 2004) and in the Portuguese version of the two scales, two factors were obtained in each scale (Nunes et al., 2014; Pires et al., 2014). The version of the PPAS adapted in Turkey is the only one to have managed to replicate the original three-factor structure (Güleç & Kavlak, 2013). All studies mentioned above were performed on the general population. No studies have been found on clinical populations.

In short, contributing to progress in perinatal mental health requires validated instruments that allow the early detection of problems in mother–infant and father–infant bonding. Currently, there is no instrument validated in Spanish to assess postnatal bonding in both parents, making it difficult to study bonding within the family system. As previously mentioned, only the MPAS and the PPAS have both versions but, equally in this case, there is some discrepancy around the instruments’ structure.

Therefore, the present study had a twofold objective: to study the psychometric properties of the Spanish version of the MPAS and the PPAS and, at the same time, to contribute to clarifying the dimensions of the postnatal bonding construct in mothers and fathers.

Materials and Methods

Participants

Initially, the sample consisted of 1,980 women and 1,721 men recruited at postpartum discharge from the hospital. As inclusion criteria, all participants were parents of a full-term newborn baby, with low or medium risk pregnancies and births (according to the classification of obstetric risk of the Mother Care Program of local health council), who could speak and read Spanish without difficulty. Participants whose contact details were erroneous (353 men and 470 women) were excluded from the sample. Some participants decided to leave the study at 5 months postpartum (55 men and 27 women). Hence, 1,313 men and 1,483 women were included in the study, but only 571 women and 376 men completed the required questionnaires. Therefore, these data indicate a response rate of 38.5% for mothers, and 28.6% for fathers (see Fig. 1).

Mothers and fathers who participated in the study differed significantly from those who did not participate regarding the following characteristics: they were older, primiparous, Spanish nationality, higher educational level, and higher economic income. There were no significant differences in their marital status, nor according to whether they were living together with their partner.

Regarding the participants’ characteristics, the women’s mean age was 34.13 years, and the men’s was 35.98 years. Most of the participants were of Spanish nationality, married and living with their partner, had university studies, and their income was between 12,000 and 30,000 euros a year (see Table 1).

Table 1:

Characteristics of the study sample (n = 947).

	Mothers (n = 571)	Fathers (n = 376)
	Mean ≠ SD(min–max)	Mean ≠ SD(min–max)
Age	34.13 ≠ 4.26(20–47)	35.98 ≠ 4.34(23–51)

	n(%)	n(%)
Marital status
Single	89 (15.6)	39 (10.4)
Married	400 (70.1)	253 (67.3)
Separated or divorced	12 (2.1)	7 (1.9)
Widow	2 (0.5)	0 (0)
Missing Values	67 (11.7)	77 (20.5)
Coexistence with the partner
Yes	485 (84.9)	292 (77.7)
Some days of the month	9 (1.6)	7 (1.9)
No	9 (1.6)	0 (0)
Missing Values	68 (11.9)	77 (20.5)
Education
Elementary school or lower	16 (3.8)	22 (5.9)
High school	96 (16.8)	85 (22.6)
High school senior or Vocational training	109 (19.1)	60 (16.0)
University	279 (48.8)	131 (34.9)
Missing values	65 (11.4)	78 (20.7)
Income level
<6,000	30 (5.3)	14 (3.7)
6,000–8,999	26 (4.6)	8 (2.1)
9,000–11,999	51 (8.9)	15 (4.0)
12,000–17,999	108 (18.9)	58 (15.4)
18,000–29,999	142 (24.9)	100 (26.6)
30,000–44,999	81 (14.2)	61 (16.2)
45,000–60,000	33 (5.8)	27 (7.2)
More than 60,000	18 (3.2)	12 (3.2)
Missing values	82 (14.4)	81 (21.5)
Nationality
Spanish	524 (91.8)	292 (77.7)
Other	25 (4.4)	14 (3.7)
Missing values	22 (3.9)	70 (18.6)

DOI: 10.7717/peerj.5980/table-1

Note:

SD, standard deviation.

Results

Reliability and validity of a single version for mothers and fathers

Due to the inadequacy of the prior confirmatory analyses, a panel of experts met to determine, on the one hand, the degree of relevance of the items of the two scales; on the other hand, the existence of differences between maternal and paternal bonding. Six experts participated in the study, all of them researchers in the concept of postnatal bonding, four of whom also had a clinical profile, with a long history of working with infants and families. The results showed a 100% inter-judge agreement, highlighting the absence of differences in the adequacy of the items as a function of the parents’ gender. Therefore, we continued the analysis, using a common version for mothers and fathers, called postnatal bonding scale (PBS), which was made up of 15 items, after deleting the unshared items (items 5, 6, 8, and 17 of the mother’s version, and items 6, 8, 9, and 15 of the father’s version).

Firstly, a confirmatory analysis was performed of the new version, using as reference the original three-factor version of the MPAS, the first instrument developed by the authors (see Fig. 2, for mothers; Fig. 3, for fathers). The results showed an adequate fit (Table 4), as their indexes reached the recommended values (in mothers: GFI = 0.96, CFI = 0.91, RMSEA = 0.049; in fathers: GFI = 0.93, CFI = 0.90, RMSEA = 0.060).

The data obtained showed how the correlation coefficients between the three subscales took high and significant values, showing a high interrelation between the components of the instrument: quality of bonding and absence of hostility (Mothers: r = 0.81, p < 0.001; Fathers: r = 0.84, p < 0.001) quality of bonding and pleasure in interaction (Mothers: r = 0.83, p < 0.001; Fathers: r = 0.73, p < 0.001), and absence of hostility and pleasure in interaction (Mothers: r = 0.66, p < 0.001; Fathers: r = 0.53, p < 0.001). All the standardized coefficients were statistically significant and obtained a value near or greater than 0.40, with the exception of Items 4 (λ = 0.30), 9 (λ = 0.26), and 19 (λ = 0.32) in the case of mothers; and of Item 4 (λ = 0.20) in the case of fathers. Lastly, the internal consistency index, analyzed through Cronbach’s alpha coefficient, was 0.70 for mothers and 0.78 for fathers. Regarding the subscales, the strategies used were the Cronbach’s alpha coefficient and the CR. The results obtained were as follows: in the case of the mothers α = 0.50 and CR = 0.74 for quality of bonding; α = 0.55 and CR = 0.93 for absence of hostility, and α = 0.60 and CR = 0.83 for pleasure in interaction. In the case of fathers, results were respectively: α = 0.54 and CR = 0.80, α = 0.64 and CR = 0.94 finally α = 0.72 and CR = 0.90. With regard to the analysis of group invariance (see Table 5), the results revealed empirical evidence of configural (Model 1) and metric invariance (Model 2) when comparing the responses of mothers and fathers. The overall results indicated the viability of constraining the factor loading to be the same across the groups. However, the CFI difference between Model 3 and Model 2 was higher than 0.001, which indicated that invariance of the intercepts was not complete across the two groups. The lack of invariance at this level does not permit the analysis of strict invariance, so we could not conduct analysis of differences in the latent or observed variables.

Table 5:

Fit indices for invariance tests.

	χ2	df	p	CFI	ΔCFI	RMSEA	RMSEA 90% CI
Mothers–Fathers
Configural	376.634	166	0.000	0.905	–	0.037	(0.032–0.041)
Metric	395.052	176	0.000	0.901	0.004	0.036	(0.031–0.041)
Scalar	542.416	185	0.000	0.838	0.063	0.045	(0.041–0.049)

DOI: 10.7717/peerj.5980/table-5

Note:

χ², chi-square; df, degrees of freedom, p, probability; CFI, comparative fit index; ΔCFI, change of the comparative fit index; RMSEA, root mean square error of approximation; RMSEA (IC-90%), RMSEA confidence interval.

In the analysis of concurrent validity, moderate negative correlations were found for the EPDS (in mothers, r = −0.41, p < 0.001; and fathers, r = −0.38, p < 0.001), and positive correlations for the DAS (in mothers, r = 0.39, p < 0.001; and in fathers, r = 0.44, p < 0.001). Thus, higher scores in postnatal bonding were associated with lower scores in depressive symptomatology and higher dyadic adjustment.

Discussion

The objective of this study was to examine the psychometric properties of the Spanish version of the MPAS and the PPAS through the analysis of internal consistency, construct and concurrent validity, and thereby to contribute to clarifying the dimensions of the construct of postnatal bonding in mothers and fathers. The results of the study were obtained from a 15-item brief version, common for mothers and fathers, called PBS, which showed a good fit to the original three-factor structure proposed by Condon & Corkindale (1998): seven items belonged to the quality of bonding dimension, four items to the absence of hostility dimension, and four items to the pleasure in interaction dimension. Worthy of note, to maintain the coherence of the study’s theoretical framework and avoid further terminological ambiguity, the term “attachment” (in the original scale) was changed for “bonding.” This latter term corresponds to the construct evaluated by the instrument.

The results obtained in the sample used in this study did not confirm the original three-factor solution of the MPAS (Condon & Corkindale, 1998) or the PPAS (Condon, Corkindale & Boyce, 2008), or the alternative models of the Italian version of the MPAS (Scopesi et al., 2004), or the Portuguese version of the two scales (Nunes et al., 2014; Pires et al., 2014). Therefore, none of the evidence found in previous research on the structure of the scale was observed in this study. These results could indicate problems in the initial structure of the scales, due to the construction process: unstructured interviews with small samples (10 women in the case of the MPAS, and 15 men in the case of the PPAS), the reduction from 31 to 19 items in both scales due to statistical decisions (Condon & Corkindale, 1998; Condon, Corkindale & Boyce, 2008), and the heterogeneity in the response option alternatives.

Due to the lack of adequacy of previous confirmatory analyses, we considered the option of using a single version for men and women, using the 15 items common to both scales. This decision was initially proposed for theoretical reasons, and we subsequently confirmed its psychometric properties.

The theoretical reasons that justified this proposal are as follow. First, most psychological constructs are evaluated using instruments common to men and women, despite recognized gender differences (e.g., emotional intelligence; Gartzia et al., 2012). In this sense, it is understood that the essence of the concept is the same, although there may be differences in its expression. Therefore, in this way, the postnatal bonding construct could be assessed using a common instrument for men and women. Secondly, to separate an instrument by sex variable—men and women—is made difficult due to the gender identity variable, since this variable may be more important than sex to explain the differences between men and women (Gartzia et al., 2012). Thus, having a postnatal bonding’s instrument which is suitable both for men and women, avoids the possibility of mistaking due to disregarding the distinction between sex and gender identity. Thirdly, disposing of common tools made up of the same items permits direct comparison between the bonding of men and women. In this sense, De Cock et al. (2016) defended the need for an instrument that evaluates prenatal and postnatal bonding conjointly in order to advance in parental bonding research. Likewise, this same argument can be used for postnatal bonding in men and women. Fourthly, different studies have used the same instrument of postnatal bonding in men and women (Edhborg et al., 2005; Hall et al., 2015; Salian & Shah, 2017), despite the fact that no studies have been found that specify the adaptation of this instrument to fathers.

In addition, it was considered necessary to use a panel of experts to obtain more evidence on the degree of relevance of the items both in men and women. The experts fully agreed that there were no differences in the degree of relevance of the items for men and women.

Therefore, evaluating these arguments, we considered that, at a theoretical level, a common instrument for men and women was adequate, and this was supported empirically.

We confirmed the three-factor structure of the original MPAS, the first instrument created by the authors (Condon & Corkindale, 1998), showing adequate fit index values. The majority of the standardized coefficients were statistically significant, with the exception of four items (items 4, 9, and 19 in the case of mothers, and Item 4 in the case of fathers). It will be necessary to analyze the functioning of these items in future studies.

The results of the global Cronbach alpha coefficients in the total sample were adequate (0.70 for mothers and 0.78 for fathers) and similar to the values found in the original versions (Condon & Corkindale, 1998; Condon, Corkindale & Boyce, 2008) and in the Italian version (Scopesi et al., 2004). However, the Cronbach alpha for the subscales showed lower values, above all in the case of mothers. In this case, we could not compare it with the original version or the Italian version, because these findings were not indicated. On the other hand, the Portuguese version (Nunes et al., 2014) and the Belgian version (Van Bussel, Spitz & Demyttenaere, 2010) of the MPAS also reported values similar to those obtained in the internal consistency of the subscales.

Due to the limitations of this method for reliability analysis (Domínguez-Lara & Merino-Soto, 2015; Lozano, García-Cueto & Muñiz, 2008), the CR was used. It is based on a structural equation modeling approach and represents a more accurate alternative for calculating reliability (Peterson & Kim, 2013). In this sense, as Raykov (2001) comments, the modeling of structural equations has the ability to empirically assess and overcome some of the limitations of the alpha coefficient. The CR results were optimal in the subscales quality of bonding (mothers, 0.74; fathers, 0.80), absence of hostility (mothers, 0.93; fathers, 0.94) and pleasure in interaction (mothers, 0.83; fathers, 0.90). Therefore, the CR results indicate that the reliability of the different subscales is adequate. We considered it necessary to maintain both procedures, coefficient alpha and CR, to allow comparing the results with previous studies and provide all necessary data for future research regarding the scale’s reliability.

With regard to the scale invariance analyses carried out, we confirmed configural variance, which indicates that the construct was conceptualized in the same way in both samples (Cheung & Rensvold, 2002); that is, we concluded that mothers and fathers shared the same definition of bonding. We also observed metric invariance, which indicates that the participating mothers probably interpret each item that makes up the scale in the same way as the fathers do. However, it was not possible to confirm complete or partial scalar invariance between these two groups. The lack of invariance in this model may be due to the difference in the sample size of each group, as the size of the group of mothers was twice as large as that of the fathers. As a result, the strict invariance hypothesis is not supported.

Concurrent validity, measured through postnatal depression and dyadic adjustment, constructs used in the original PPAS (Condon, Corkindale & Boyce, 2008), showed high associations in the same line of the scientific literature and were consistent with the hypotheses proposed in the study. These results support the validity of the scale.

Conclusions

The new version of the scale of postnatal bonding, showing appropriate structural and concurrent validity and adequate reliability, is suitable for use in mothers and fathers. It evaluates the postnatal bond and allows comparing that of mothers and fathers: this contributes to progressing in perinatal mental health and in understanding gender differences in postnatal bonding. In this regard, we emphasize the need to include fathers in assessment and intervention programs within the context of transition to parenthood, thus favoring a systemic perspective on the family.

Supplemental Information