Archives of Women's Mental Health (2021) 24:619–625
https://doi.org/10.1007/s00737-021-01105-8
ORIGINAL ARTICLE
Prenatal attachment: using measurement invariance to test
the validity of comparisons across eight culturally diverse countries
Sarah Foley1 & Claire Hughes1 & Aja Louise Murray2 & Adriana Baban3 & Asvini D. Fernando4 & Bernadette Madrid5 &
Joseph Osafo6 & Siham Sikander7 & Fahad Abbasi7 Susan Walker8& & Bao-Yen Luong-Thanh9 & Thang Van Vo9 &
Mark Tomlinson10,11 Pasco Fearon12 Catherine L. Ward13 Sara Valdebenito14 Manuel Eisner14& & & &
Received: 28 September 2020 /Accepted: 19 January 2021 / Published online: 9 February 2021
# Crown 2021, corrected publication March 2021
Abstract
Studies in high-income countries (HICs) have shown that variability in maternal-fetal attachment (MFA) predict important
maternal health and child outcomes. However, the validity of MFA ratings in low- and middle-income countries (LMICs)
remains unknown. Addressing this gap, we assessed measurement invariance to test the conceptual equivalence of the
Prenatal Attachment Inventory (PAI: Muller, 1993) across eight LMICs. Our aim was to determine whether the PAI yields
similar information from pregnant women across different cultural contexts.We administered the 18-item PAI to 1181mothers in
the third trimester (Mean age = 28.27 years old, SD = 5.81 years, range = 18–48 years) expecting their first infant (n = 359) or a
later-born infant (n = 820) as part of a prospective birth cohort study involving eight middle-income countries: Ghana, Jamaica,
Pakistan, Philippines, Romania, South Africa, Sri Lanka and Vietnam.We usedMultiple Group Confirmatory Factor Analyses to
assess across-site measurement invariance. A single latent factor with partial measurement invariance was found across all sites
except Pakistan. Group comparisons showed that mean levels ofMFAwere lowest for expectant mothers in Vietnam and highest
for expectant mothers in Sri Lanka. MFA was higher in first-time mothers than in mothers expecting a later-born child. The PAI
yields similar information about MFA across culturally distinct middle-income countries. These findings strengthen confidence
in the use of the tool across different settings; future studies should explore the use of the PAI as a screen for maternal behaviour
that place children at risk.
Keywords Maternal-fetal attachment . Pregnancy .Measurement invariance . Lower-middle income . Cross-cultural . Parity
* Sarah Foley 8 Caribbean Institute for Health Research, The University of the West
sf412@cam.ac.uk Indies, Kingston, Jamaica
9 Institute for Community Health Research, Faculty of Public Health,
1 Centre for Family Research, University of Cambridge, Hue University of Medicine and Pharmacy, Hue University,
Cambridge, UK Hue, Vietnam
2 Department of Psychology, University of Edinburgh, Edinburgh, UK 10 Department of Psychology, Stellenbosch University, Stellenbosch,
3 Department of Psychology, Babes-Bolyai University, South African Institute for Life Course Health Research, Department
Cluj-Napoca, Romania of Global Health, Stellenbosch University, Cape Town, South Africa
4 Department of Paediatrics, Faculty of Medicine, University of 11 School of Nursing and Midwifery, Queens University, Belfast, UK
Kelaniya, Colombo, Sri Lanka
12
5 Research Department of Clinical, Educational and HealthChild Protection Unit, University of the Philippines, Psychology, University College London, London, UK
Manila, Philippines
13
6 Department of Psychology, University of Ghana, Legon, Department of Psychology, University of Cape Town,
Accra, Ghana Cape Town, South Africa
7 Health Services Academy, Islamabad, Pakistan 14 Institute of Criminology, University of Cambridge, Cambridge, UK
620 S. Foley et al.
Postpartum depression (PPD) is a key area of research in the field measuring MFA in India with a sample of 230 pregnant women
ofmaternal and child health (Field 2010), and research is increas- using the Maternal-Fetal Assessment Scale (MFAS; Cranley
ingly recognising the crucial importance of pregnancy as a period 1981). Their findings showed average MFA scores (M= 87.43,
of potential risk and opportunity for intervention. However, re- SD = 10) were at the lower end of the range (minimum 70–
search on maternal psychological health has scarcely been con- maximum 114), which the authors interpreted as reflecting social
ducted outside of high-income countries (HICs) (known as the and economic restrictions on women’s ability to develop a rela-
10/90 gap; COHRED 1990; Henrich et al. 2010). That said, tionship with the fetus, coupled with a cultural emphasis upon
reviews that have focused on research in low- and middle- birth rather than pregnancy. However, another study in India
income countries (LMICs) have demonstrate that prenatal stress (not included in Tichelman et al. 2019) found no association
is as strongly associated with adverse infant outcomes as in HICs between PAI scores and depression in gestational surrogates or
(Buffa et al. 2018; Parsons et al. 2012) and that rates of PPD matched controls (Lamba et al. 2018). Existing findings therefore
often exceed those reported for HICs (e.g. Fisher et al. 2012; paint amixed picture regarding the construct validity of the PAI in
Woody et al. 2017). Given this, it is critically important to test LMICs. Furthermore, in both theBangladeshi and Indian samples,
the universality of processes underlyingmaternal and infantmen- the Cronbach’s alpha was lower than for English versions and
tal health in culturally and economically diverse contexts, as included no further psychometric analysis of the questionnaires.
outcomes differ crucially across contexts (Vidyasagar 2006). Punamäki et al. (2017) asked 511 expectant Palestinian
One area of focus, as illustrated in a recent systematic review mothers living in Gaza to complete the PAI. They reported that
(Tichelman et al. 2019), has been the role of maternal-fetal lack of social support mediated the association between war
attachment (MFA)—defined as the strength of mothers’ emo- trauma and reduced MFA, but found MFA was unrelated to
tional ties with the fetus (and also known as prenatal bonding; PPD. Given the scarcity of research on MFA in LMICs, it is
Walsh 2010).MFA assessments capture variability in expectant difficult to judge whether this lack of association reflects the
mothers’ behaviours, cognitions and emotions towards the fe- particular challenges of living in a war zone, a more general
tus, which appear important for positive prenatal health prac- contrast in the salience of MFA across resource settings or cul-
tices such as giving up smoking (e.g. Lindgren 2001), attending tural differences in the meaning or validity of the instruments
clinics and exercising regularly (Cinar et al. 2017). In a meta- used (note Punamäki et al. (2017) were unable to replicate the
analysis of 14 studies that both spanned prenatal and postnatal established PAI factor structure).
periods and included direct observational ratings of parent- A key step in evaluating the conceptual equivalence of MFA
infant interactions, Foley and Hughes (2018) found that MFA ratings across different contexts is to construct latent factor
predicted caregiver sensitivity in interactions with their infant, models and test for measurement invariance across groups.
which is a potent influence on children’s cognitive and socio- That is, assessing whether a tool shows an equivalent structure
emotional development (Mills-Koonce et al. 2015). Thus, valid and meaning across distinct groups. Theoretically, assessing
measurement of MFA has potentially far-reaching clinical im- whether an instrument yields similar information in different
plications for the health of both mothers and infants (Laxton- LMICs is necessary to test the cultural universality of inferences
Kane and Slade 2002). about specific causal pathways. Practically, demonstrating
Predictors of MFA include gestational age (Yarcheski et al. across-site equivalence is useful in establishing whether similar
2009), good maternal mental health (Rollè et al. 2020), social or distinct methods of identifying at-risk individuals should be
support and first pregnancy (Tichelman et al. 2019). In contrast, used across settings.
MFA appears unrelated to demographics (e.g., education, age, The current study first aimed to assess the conceptual and
marital status), fertility treatment, fetal defects or planned preg- measurement equivalence of the PAI across eight LMICs in
nancies (Tichelman et al. 2019). In a systematic review of 41 widely different global regions. By testing for measurement in-
studies, Rollè et al. (2020) showed that treatment for depression variance across eight LMICs, we aim to establish whether PAI
can improve MFA. However, this research field is limited by its items capture variation in how MFA is manifest across different
narrow focus on parents (mainly mothers) living in HICs groups. Finally, if measurement invariance is established, our
(Brandon et al. 2009). For example, all but two of the 77 studies next two aims were to compare mean levels of MFA across site
included in Tichelman et al.’s (2019) review were from HICs. and birth parity.
Of the two MFA studies with LMIC samples included in
Tichelman et al.’s (2019) review, the first involved 672 mothers
in rural Bangladesh. Edhborg et al. (2011) found prenatal MFA Methods
(assessed via a translation of the widely used Prenatal Attachment
Inventory; PAI: Muller 1993) was positively associated with Participants
bonding when the infant was 2–3 months old and negatively
associated with PPD symptoms. In the second study, Trained female fieldworkers invited women attending antena-
Lingeswaran and Bindu (2012) assessed the feasibility of tal appointments to participate in Evidence for Better Lives
Prenatal attachment: using measurement invariance to test the validity of comparisons across eight... 621
Study – Foundational Research (EBLS-FR), a prospective demographic factors/reproductive history as part of a struc-
birth cohort study in eight different sites (N = 1208) that rep- tured interview. The Ethics Boards of each university ap-
resent distinct social and cultural conditions across major proved the EBLS-FR protocol (Valdebenito et al. 2020).
world regions: Koforidua (Ghana, n = 150), Kingston
(Jamaica, n = 152), Tarlai Kalan (Pakistan, n = 150),
Valenzuela (Philippines, n = 154), Cluj-Napoca (Romania, Measures
n = 150), Ragama (Sri Lanka, n = 152), Worcester (South
Africa, n = 150) and Hue (Vietnam, n = 150). Inclusion Maternal-fetal attachment Expectant mothers completed the
criteria were as follows: (i) third trimester of pregnancy (i.e. 18-item PAI (Muller 1993). Participants rated how often they
weeks 29–40), (ii) aged over 18 and (iii) living primarily with- engaged in specific thoughts or behaviours towards the fetus
in the study’s defined geographical area. The average recruit- (e.g., ‘I feel love for the baby’ or ‘I stroke the baby through my
ment rate across sites was 82%. Table 1 presents sample char- tummy’) on a 4-point scale: 1 = almost never, 2 = sometimes,
acteristics by site for the 1181 women expecting singleton 3 = often, 4 = almost always. A review ofMFA questionnaires
pregnancies in the current study (excluding 27 multiple preg- concluded that the PAI is psychometrically sound (e.g.,
nancies). On average, mothers were 28.27 years old, SD = Cronbach’s alpha ranges between .81 and .93) and, unlike
5.81 years, range: 18–48 years (note expectant mothers in other measures, includes behaviours and thoughts and feel-
Jamaica, Pakistan and South Africa were significantly youn- ings towards the fetus (van den Bergh and Simons 2009).
ger than mothers in Romania, Ghana, Sri Lanka and
Vietnam). For 359 /1181 (30%), this was their first pregnancy Background measures Expectant mothers also reported on
(note half of the Romanian mothers were experiencing their their age, socioeconomic status, education and previous
first pregnancy whilst Pakistan had the smallest proportion of pregnancies/ births.
women experiencing their first pregnancy across sites). The
sample was diverse in terms of level of education: 4% none, Data analysis
11% primary, 48% secondary, 16% vocational and 21% uni-
versity (lowest to highest level of maternal education, respec- We analysed the data using a latent variable framework inMplus
tively: Pakistan, Ghana, Jamaica, South Africa, Vietnam, Sri (Version 8.4; Muthén andMuthén 2018) and used Confirmatory
Lanka, Philippines and Romania). Factor Analysis (CFA) to test measurement models and compare
mean levels of MFA. Data screening indicated that there was no
variability in responses to item 3 ‘I enjoy feeling the baby move’
Procedure in the Philippines and so it was removed from further analyses.
Furthermore, the limited spread in the 4 response options across
Information sheets, consent forms and questionnaires were sites for the remaining the items indicated scores should be col-
translated (using guidelines from the World Health lapsed into dichotomous indicators (i.e., disagree/agree). This is
Organization; http://www.who.int/substance_abuse/ in line with Rutkowski et al.’ (2019) recommended approach for
research_tools/transla- tion/en/) into the most frequently dealingwith floor/ceiling effects in response categories. Analyses
spoken languages by the participants, who provided either applied the weighted least squares mean and variance adjusted
written or audio-recorded informed consent. Expectant (WLSMV) estimator. Following Muller’ (1993) original conten-
mothers in their third trimester of pregnancy (M = 33.23weeks tion that MFA is unidimensional, we tested a 1-factor model
gestation, SD = 3.36) completed the PAI and reported on alongside a 3-factor solution reflecting the dimensions of
Table 1 Site-specific sample demographics
Ghana Jamaica Pakistan Philippines Romania South Africa Sri Lanka Vietnam Between-site
n=146 n=151 n=143 n=152 n=148 n=143 n=149 n=149 differences
M (SD) M (SD) M (SD) M (SD) M (SD) M (SD) M (SD) M (SD) F η2
Age of mother 29.05 (6.38) 25.57 (5.69) 27.28 (5.24) 27.63 (6.05) 30.07 (4.62) 26.89 (5.99) 29.75 (5.56) 29.93 (5.16) 13.07*** .07
(years)
Gestation (weeks) 34.35 (3.78) 33.93 (3.06) 31.82 (3.72) 32.75 (2.98) 33.16 (3.12) 34.25 (3.31) 32.38 (2.86) 33.21 (3.00) 11.18 .06
Grades passed in 8.33 (4.44) 10.46 (1.06) 7.80 (4.79) 11.44 (2.66) 12.80 (1.93) 10.51 (1.88) 11.83 (1.47) 10.53 (2.48) 49.67 .24
education system
% % % % % % % % χ2 Cramer’s V
First pregnancy 19.9 26.5 12.6 25 50 36.4 43 29.5 71.93 .25
***p < .001
622 S. Foley et al.
anticipation, differentiation and interaction suggested by some Table 2 Model fit indices for a 1-factor and 3-factor model of the
previous studies (e.g., Pallant et al. 2014). We evaluated model Prenatal Attachment Inventory by site
fit using three primary criteria: Comparative Fit Index (CFI) > Site 1-factor model fit 3-factor model fit
0.90, Tucker Lewis Index (CFI) > 0.90, RootMean Square Error
of Approximation (RMSEA) < 0.08 (Brown 2015). To test mea- RMSEA CFI TLI RMSEA CFI TLI
surement invariance across site and parity, we used multigroup
Ghana 0.070 0.939 0.930 .066 .948 .939
CFA, which involves systematically adding equality constraints
Jamaica 0.044 0.935 0.926 .036 .957 .949
to the model and testing the change in model fit of these nested
Pakistan 0.038 0.962 0.957 .036 .967 .962
models (Byrne 2012). Nestedmodel comparisonswere judged to
Philippines 0.037 0.954 0.947 .036 .959 .952
be invariant if the CFI decreased by ≤0.020 and RMSEA in-
Romania 0.086 0.829 0.804 .086 .827 .798
creased by ≤ .003 (Svetina et al. 2020). First, we used a
South Africa 0.028 0.973 0.969 .015 .992 .991
configural model to test for equivalence in the pattern of item
Sri Lanka 0.047 0.928 0.917 .047 .929 .917
loadings across site and then parity groups (i.e., the organisation
Vietnam 0.065 0.892 0.876 .067 .888 .869
of the construct is similar). If this model showed acceptable fit,
we then proceeded to test metric (weak factorial) invariance,
specifying that factor loadings were equal across sites (i.e., each
item contributes in a similar way to the construct). If metric invariant). Thus, a partially scalar invariant model was con-
invariance was not achieved, we sought to improve model fit sidered to have an acceptable model fit, CFI = 0.901, TLI =
by inspecting modification indices to evaluate whether releasing 0.897, RMSEA = 0.060.
item constraints would yield a partially invariantmodel. Next, for
items that showed metric invariance, we tested for scalar (strong Cross-cultural differences in prenatal attachment
factorial) invariance, which involves equivalence of item thresh-
olds (i.e., the cut-off underling the distribution of scores is con- Using the partial scalar-invariant model, we tested for mean
sistent). Finally, if we established scalar invariance, we differences in the latent factor across sites. We adopted a
proceeded to test the mean differences in MFA by constraining Bonferroni correction and adjusted the alpha level (i.e.,
the mean of the reference group to zero and freely estimating the p < .007) to account for multiple comparisons. With Ghana
other group means (Byrne 2012). as the reference site, we found mean between-site differences
in MFA, with lower levels in Vietnam, b = −.50, z = −3.78,
p < .001, and higher levels in Sri Lanka, b = .73, z = 3.00,
Results p < .001, however no significant contrasts with expectant
mothers in Jamaica, Romania, Philippines and South Africa.
MFA factor structure
Both the 1-factor and 3-factor solution showed an acceptable Parity and prenatal attachment
fit to the data for all sites (Table 2). However, given the high
overlap between the 3 factors (r = > .8) we proceeded to test The model testing the equivalence of the configural model
the across-site equivalence of the 1-factor solution. across parity showed acceptable fit, CFI = 0.911, TLI =
0.894, RMSEA = 0.053, SRMR= 0.052. Adding metric con-
Tests of measurement invariance across site straints led to a slight improvement in model fit, CFI = 0.912,
TLI = 0.902, RMSEA = 0.051, SRMR = 0.06 and imposing
The configural model showed poor model fit, CFI = 0.627, constraints to examine scalar invariance did not significantly
TLI = 0.633, RMSEA = 0.11, SRMR= 0.061. Inspection of reduce model fit, CFI = 0.903, TLI = 0.902, RMSEA= 0.052,
the modification indices suggested the removal of Pakistan SRMR = 0.066. We subsequently compared the mean of the
would improve the model fit. Indeed, the configural model MFA latent factor across primiparous and multiparous wom-
for the 7 remaining sites showed good model fit, CFI = en, which demonstrated higher MFA in first-time mothers:
0.925, TLI = 0.912, RMSEA = 0.05, SRMR= 0.061. Adding b = −.21, z = −2.19, p = .029.
metric constraints did not significantly reduce model fit:
CFI = 0.921, TLI = 0.914, RMSEA = 0.05, SRMR = 0.075,
but imposing constraints to examine scalar invariance did re- Discussion
duce model fit. Inspection of modification indices led to re-
leasing equality constraints of factor intercepts for 15 items Evidence for links between MFA and clinically relevant pre-
(items 10 ‘I know when the baby is asleep’, 12 ‘I feel love for and post-natal behaviours (Rollè et al. 2020; Tichelman 2019)
the baby’ and 16 ‘I stroke the baby through my tummy’ were has relied heavily on studies from a fewHICs. Addressing this
Prenatal attachment: using measurement invariance to test the validity of comparisons across eight... 623
gap, 1181 expectant mothers completed the PAI to examine externalise their developing bond. Highlighting the impor-
the conceptual equivalence of MFA across eight LMICs: tance of cultural appropriateness, Arafah et al. (2020) included
Ghana, Jamaica, Pakistan, Philippines, Romania, South four new items in the PAI to capture the MFA amongst
Africa, Sri Lanka and Vietnam. The PAI showed conceptual Arabic-speaking women in Qatar (e.g., ‘I am careful with
and measurement equivalence for pregnant women in seven my activities so nothing will hurt my baby’). Future qualitative
of the eight sites, with mean levels of MFA being lowest research might therefore help highlight relevant questions for
amongst expectant mothers in Vietnam and highest amongst pregnant women in Pakistan or whether the MFA construct
expectant mothers in Sri Lanka. Whilst conceptually similar lacks validity in this context.
across levels of parity, MFA was greater in first-time mothers.
Site-specific differences in MFA
A Modified Measure of MFA shows Cross-Cultural
Equivalence Whilst we often lack good cross-cultural research, establishing
the conceptual and measurement equivalence of measures across
Consistent with findings from studies that have adapted the different cultures is a key pre-requisite for drawing valid group
PAI for use in non-English speaking HICs (e.g., Pavše et al. comparisons (Boer et al. 2018). Having established that the PAI
2019), our analyses support a one-factor MFA model. Similar shows measurement invariance across seven sites, we found
to previous research with the PAI in India (Lamba et al. 2018), lower levels of MFA in Vietnam and higher levels in Sri
the items contributing to MFA were dichotomised into agree/ Lanka, than in the other five countries. One factor that might
disagree options. TheMFA latent factor exhibited partial mea- contribute to these site contrasts is variation in antenatal care.
surement invariance across seven out of the eight sites—with For example, in Sri Lanka, the overwhelming majority of expec-
some of item thresholds freely estimated at the scalar level of tant mothers use the free, comprehensive maternal healthcare
invariance. Taken together, this suggests that a simplified two- package, which includes antenatal classes that encourage expec-
option (agree/disagree) PAI is appropriate for use across cul- tant mothers to promote specific behaviours, to seek partner sup-
turally diverse groups, which should increase confidence in port and to think about their developing infants as individuals
the validity of cross-cultural comparisons, as well as in the use (Hemachandra 2011). Future research is needed to unpack the,
of the PAI in efforts to improve the health and wellbeing of potentially diverse, origins of differences in MFA, for example
women across the world (UN General Assembly 2015). cultural norms, sampling and economic factors. Our study rein-
That said, the PAI did not pass the first level of equivalence forces the importance of testing the equivalence of measures in
testing in Pakistan. Closer inspection of item responses in multi-site studies in order to make meaningful comparisons.
Pakistan revealed a near-floor effect for seven items. For ex-
ample, few expectant Pakistani mothers endorsed items such First-time pregnancy: a time to reflect?
as: I stroke the baby through my tummy or I imagine calling
the baby by name. This may reflect other contrasts between First-time mothers reported higher MFA than mothers in later
Pakistan and the other sites, including elevated rates of infant pregnancies. Whilst similar results studies have been reported
mortality (i.e., < 5 years of age = 74 per 1000 live births), in high- (e.g., Tichleman et al. 2019) and upper-middle-
poverty (i.e., GDP pc = $1550) and education (as highlighted income countries (e.g. Turkey; Özcan et al. 2018), the mea-
in Table 1: expectant mothers in Pakistan received significant- surement invariance tests employed in the current study in-
ly lower levels of education than all sites aside from Ghana). crease the confidence that there is a true and meaningful effect
The lack of item endorsement may reflect a coping strategy for of parity. One simple explanation is that first-time mothers
the mothers. For example, in seminal ethnographic research in have greater psychological space for thinking about their
the Alto do Cruzeiro, a shantytown in northeast Brazil, new infant than mothers who are already taking care of youn-
Sheper-Hughes (1992) described how, in the context of high ger infants. Our results suggest this parity effect is culturally
infant mortality, impoverished mothers inhibited attachment universal—indeed, the magnitude of this effect appears stron-
towards their infant after birth (e.g., not attributing human ger in this study than in previous meta-analyses (Yarcheski
qualities towards their infants). Alternatively, prevailing belief et al. 2009; Tichelman et al. 2019).
systems may render specific items or explicit endorsement of
items inappropriate. Ethnographic studies have emphasised Strengths and limitations
expectant mothers in Pakistan are very discreet about their
pregnancies, with open conversations both within and outside The inclusion of data from eight LMICs is a key strength of
of the family seen as immodest (i.e., inferring disclosure of the study. Notably, similar studies of measurement equiva-
sexual activity) and inviting nazar (the evil eye of jealousy), lence have excluded data from LMICs due to a low response
which may harm the unborn infant (Qureshi and Pacquiao rate (e.g., Zhang 2020). That said, our samples were recruited
2013; Qamar 2016). Thus, women may internalise rather than using a non-probabilistic sampling strategy, which potentially
624 S. Foley et al.
limits the generalisability of our results to the wider popula- References
tion. Furthermore, although the average rate of recruitment
was high (82%), there was between-site variability (i.e., rang- ArafahD, Thomas B, Fenton TR, Sabr Y,Metcalfe A (2020) Validity and
ing from 50% in Sri Lanka to 98% in Pakistan) which suggests reliability of the Arabic version of Muller’s prenatal attachment
inventory. J Psychosom Obstet Gynecol 1–9. https://doi.org/10.
further research is required to rule out the possibility that ele- 1080/0167482X.2020.1713083
vated rates of MFA reported by expectant mothers in Sri Boer D, Hanke K, He J (2018) On detecting systematic measurement
Lanka are not the result of selection bias. In addition, sample error in cross-cultural research: a review and critical reflection on
characteristics differed across sites, for example in terms of equivalence and invariance tests. J Cross-Cult Psychol 49:713–734.
https://doi.org/10.1177/0022022117749042
education (see Table 1). However, having established the Brandon AR, Pitts S, Denton WH, Stringer CA, Evans HM (2009) A
measurement properties of the PAI as an index ofMFA, future history of the theory of prenatal attachment. J Prenat Perinat Psychol
research can examine the cultural universality/specificity of Health 23(4):201–222
these associations as well explore the use of the PAI as a Brown T (2015) Confirmatory factor analysis for applied research, 2nd
screener for maternal behaviours that may place children at edn. Guilford Press, London
Buffa G, Dahan S, Sinclair I, St-Pierre M, Roofigari N, Mutran D,
risk. Rondeau J-J, Dancause KN (2018) Prenatal stress and child devel-
opment: a scoping review of research in low- and middle-income
countries. PLoS One 13:e0207235. https://doi.org/10.1371/journal.
pone.0207235
Conclusions ByrneB (2012) Structural equationmodelingwithMplus: basic concepts,
applications, and programming (multivariate applications series).
Routledge, New York
By adapting the PAI to yield a simple but valid tool that shows Cinar N, Caka SY, Topal S, Yuvaci HU, Erkorkmaz U (2017) The rela-
cross-cultural equivalence in seven LMICs and across birth tion of health-related practices of pregnant women, fatigue and pre-
order, we hope to stimulate further work examining prenatal natal attachment. J Coll Physicians Surg Pak 27:693–698
psychological processes underpinning maternal and infant ad- Commission on Health Research for Development (1990) Health re-
search: essential link to equity in development. Oxford University
justment across the first 1000 days. Press, New York
CranleyM (1981)Development of a tool for themeasurement ofmaternal
Funding The work of the Evidence for Better Lives Study was supported attachment during pregnancy.MSNurs Res 30:281–284. https://doi.
by the Jacobs Foundation, UBS Optimus Foundation, Fondation Botnar, org/10.1097/00006199-198109000-00008
the Consuelo Zobel Alger Foundation, the British Academy, the Edhborg M, Nasreen H-E, Kabir ZN (2011) Impact of postpartum de-
Cambridge Humanities Research Grants Scheme, the ESRC Impact pressive and anxiety symptoms on mothers’ emotional tie to their
Acceleration Account Programme, a Queensland University of infants 2–3 months postpartum: a population-based study from rural
Technology Postgraduate Research Award, Higher Degree Research Bangladesh. Arch Womens Ment Health 14:307–316. https://doi.
Student Supplementary Research Funding from Queensland University org/10.1007/s00737-011-0221-7
of Technology, the University of Edinburgh College Office for the Field T (2010) Postpartum depression effects on early interactions, par-
College of Arts, the Humanities and Social Sciences SFC ODA Global enting, and safety practices: a review. Infant Behav Dev 33:1–6.
Challenges Internal Fund, the University of Cambridge GCRF Quality https://doi.org/10.1016/j.infbeh.2009.10.005
Research Fund and the Wolfson Professor of Criminology Discretionary
Fisher J, Mello MCD, Patel V, Rahman A, Tran T, Holton S, Holmes W
Fund. The first author was funded by an Economic and Social Research
(2012) Prevalence and determinants of common perinatal mental
Council Post-Doctoral Fellowship.
disorders in women in low-and lower-middle-income countries: a
systematic review. Bull World Health Organ 90:139–149. https://
Declarations doi.org/10.2471/BLT.11.091850
Foley S, Hughes C (2018) Great expectations? Do mothers’ and fathers’
Conflict of interest The authors declare no competing interests. prenatal thoughts and feelings about the infant predict parent-infant
interaction quality? A meta-analytic review. Dev Rev 48:40–54.
Disclaimer The content of this paper is the responsibility of the authors https://doi.org/10.1016/j.dr.2018.03.007
and does not reflect official views of these funding bodies. Hemachandra N (2011)Maternal care package: a guide to field healthcare
workers. Family Health Bureau, Ministry of Health. Sri Lanka
Open Access This article is licensed under a Creative Commons Henrich J, Heine SJ, Norenzayen A (2010) The weirdest people in the
Attribution 4.0 International License, which permits use, sharing, adap- world? Behav Brain Sci 33:1–75. https://doi.org/10.1017/
tation, distribution and reproduction in any medium or format, as long as S0140525X0999152X
you give appropriate credit to the original author(s) and the source, pro- Lamba N, Jadva V, Kadam K, Golombok S (2018) The psychological
vide a link to the Creative Commons licence, and indicate if changes were well-being and prenatal bonding of gestational surrogates. Hum
made. The images or other third party material in this article are included Reprod 33:646–653. https://doi.org/10.1093/humrep/dey048
in the article's Creative Commons licence, unless indicated otherwise in a Laxton-KaneM, Slade P (2002) The role of maternal prenatal attachment
credit line to the material. If material is not included in the article's in a woman’s experience of pregnancy and implications for the
Creative Commons licence and your intended use is not permitted by process of care. J Reprod Infant Psychol 20:253–266. https://doi.
statutory regulation or exceeds the permitted use, you will need to obtain org/10.1080/0264683021000033174
permission directly from the copyright holder. To view a copy of this Lindgren K (2001) Relationships among maternal-fetal attachment, pre-
licence, visit http://creativecommons.org/licenses/by/4.0/. natal depression, and health practices in pregnancy. Res Nurs Health
24:203–217. https://doi.org/10.1002/nur.1023
Prenatal attachment: using measurement invariance to test the validity of comparisons across eight... 625
LingeswaranA, Bindu H (2012) Validation of Tamil version of Cranley’s Sheper-Hughes N (1992) Death without weeping: the violence of every-
24-item maternal–fetal attachment scale in Indian pregnant women. day life in Brazil. University of California Press, Berkeley, Los
J Obstet Gynaecol India 62:630–634. https://doi.org/10.1007/ Angeles, Oxford
s13224-012-0175-3 Svetina D, Rutkowski L, Rutkowski D (2020) Multiple-group invariance
Mills-Koonce W, Willoughby M, Zvara B, Barnett M, Gustafsson H, with categorical outcomes using updated guidelines: an illustration
CoxMJ (2015) Mothers’ and fathers’ sensitivity and children’s cog- using Mplus and the lavaan/semTools packages. Struct Equ Model
nitive development in low- income, rural families. J Appl Dev Multidiscip J 27:111–130. https://doi.org/10.1080/10705511.2019.
Psychol 38:1–10. https://doi.org/10.1016/j.appdev.2015.01.001 1602776
Muller ME (1993) Development of the prenatal attachment inventory. Tichelman E, Westerneng M, Witteveen AB, Van Baar AL, Van Der
West J Nurs Res 15:199–215. https://doi.org/10.1177/ Horst HE, De Jonge A, Berger MY, Schellevis FG, Burger H,
019394599301500205 Peters LL (2019) Correlates of prenatal and postnatal mother to-
Muthén LK, Muthén B (2018) Mplus. The comprehensive modelling infant bonding quality: a systematic review. PLoS One 14:
program for applied researchers: user’s guide 5. Muthén, B & e0222998. https://doi.org/10.1371/journal.pone.0222998
Muthén, LK, Los Angeles UN General Assembly, Transforming our world : the 2030 agenda for
Özcan NK, Boyacıoğlu NE, Dikeç G, Dinç H, Enginkaya S, Tomruk N sustainable development, 21 October 2015, A/RES/70/1, available
(2018) Prenatal and postnatal attachment among Turkish mothers at: https://www.refworld.org/docid/57b6e3e44.html. Accessed 24
diagnosed with a mental health disorder. Issues Ment Health Nurs Sept 2020
39:795–801. https://doi.org/10.1080/01612840.2018.1455773 Valdebenito S, Murray AL, Hughes C, Baban A, Asvini D et al (2020)
Pallant JF, Haines HM, Hildingsson I, Cross M, Rubertsson C (2014) Evidence for Better Lives Study: A comparative birth-cohort study
Psychometric evaluation and refinement of the Prenatal Attachment on child exposure to violence and other adversities in eight low- and
Inventory. J Reprod Infant Psychol 32:112–125. https://doi.org/10. middle-income countries – Foundational Research. BMJ Open 10:
1080/02646838.2013.871627 e034986.
Parsons CE, Young KS, Rochat TJ, Kringelbach ML, Stein A (2012)
van den Bergh B, Simons A (2009) A review of scales to measure the
Postnatal depression and its effects on child development: a review
mother-foetus relationship. J Reprod Infant Psychol 27:114–126.
of evidence from low- and middle-income countries. Br Med Bull
https://doi.org/10.1080/02646830802007480
101:57–79. https://doi.org/10.1093/bmb/ldr047
Pavše L, Tul N, Velikonja V (2019) Analysis of the internal structure of Vidyasagar D (2006) Global notes: the 10/90 gap disparities in global
the Slovenian version of the Prenatal Attachment Inventory (PAI) health research. J Perinatol 26:55–56. https://doi.org/10.1038/sj.jp.
[Analiza notranje strukture slovenskega prevoda Lestvice vezi med 7211402
nosečnico in plodom (PAI). Psiholoska Obzorja 28:11–18. https:// Walsh J (2010) Definitions matter: if maternal-fetal relationships are not
doi.org/10.20419/2019.28.496 attachment, what are they? ArchWomensMent Health 13:449–451.
Punamäki RL, Isosävi S, Qouta SR, Kuittinen S, Diab SY (2017) War https://doi.org/10.1007/s00737-010-0152-8
trauma and maternal–fetal attachment predicting maternal mental Woody CA, Ferrari AJ, Siskind DJ, Whiteford HA, Harris MG (2017) A
health, infant development, and dyadic interaction in Palestinian systematic review and meta-regression of the prevalence and inci-
families. Attach Hum Dev 19:463–486. https://doi.org/10.1080/ dence of perinatal depression. J Affect Disord 219:86–92. https://
14616734.2017.1330833 doi.org/10.1016/j.jad.2017.05.003
Qamar AH (2016) Belief in the evil eye and early childcare in rural Yarcheski A, Mahon N, Yarcheski T, Hanks M, Cannella B (2009) A
Punjab, Pakistan. Asian Ethnology 75:397–418. https://doi.org/10. meta-analytic study of predictors of maternal-fetal attachment. Int J
2307/asianeth.75.2.397 Nurs Stud 46:708–715. https://doi.org/10.1016/j.ijnurstu.2008.10.
Qureshi R, Pacquiao D (2013) Ethnographic study of experiences of 013
Pakistani women immigrants with pregnancy, birthing, and postpar- Zhang RJ (2020) Social trust and satisfaction with life: a cross-lagged
tum care in the United States and Pakistan. J Transcult Nurs 24:355– panel analysis based on representative samples from 18 societies.
362. https://doi.org/10.1177/1043659613493438 Soc Sci Med 251:112901. https://doi.org/10.1016/j.socscimed.
Rollè L, Giordano M, Santoniccolo F, Trombetta T (2020) Prenatal at- 2020.112901
tachment and perinatal depression: a systematic review. Int J
Environ Res Public Health 17:2644. https://doi.org/10.3390/ Publisher’s note Springer Nature remains neutral with regard to jurisdic-
ijerph17082644 tional claims in published maps and institutional affiliations.
Rutkowski L, Svetina D, Liaw Y-L (2019) Collapsing categorical vari-
ables and measurement invariance. Struct Equ Model Multidiscip J
0:1–13. https://doi.org/10.1080/10705511.2018.1547640