What tools should be used to identify women in need of additional support in pregnancy?

ABSTRACT

Objective

This study aimed to identify and inform recommendation of self-report and interview-based instruments that are feasible, reliable and valid to evaluate the quality of the maternal-fetal relationship (MFR).

Background

Several constructs predicting parent–infant interaction and later infant adjustment are used to assess mothers’ thoughts and feelings towards their unborn baby, including reflective functioning, mind-mindedness, representation, and fetal attachment. As yet, there is no existing review comparing the quality and accessibility of instruments across each of these constructs.

Methods

A systematic literature review was undertaken to synthesise psychometric information on measures reporting on the MFR. Searches of six databases were conducted. English articles were selected based on inclusion and exclusion criteria. The QATSDD checklist was used to assess study quality.

Results

 Of 669 studies identified, 28 met inclusion criteria. Thirteen different instruments were identified for evaluation. Reported reliability and validity varied significantly across instruments, as well as availability for research and/or clinical use.

Conclusion

Suggestions for research and clinical practice include further evaluation of the psychometric properties of tools, particularly for self-report measures of reflective functioning, use of interviews to scaffold reflexivity, and development of clinical policies and procedures to clarify care pathways for expectant mothers needing further support.

KEYWORDS:

• maternal-fetal attachment
• reflective functioning
• antenatal bonding
• mind-mindedness
• maternal representation

Research into the maternal-fetal relationship (MFR) – expectant mothers’ thoughts and feelings about their unborn baby – began four decades ago with investigations into the quality of the maternal bond with the unborn baby – (‘maternal-fetal attachment’ ([MFA], Cranley, 1981)). Numerous other aspects of the maternal-fetal relationship have since been identified – ‘maternal-fetal representation’, ‘maternal reflective functioning’ and ‘maternal mind-mindedness’ – which focus at a more cognitive level on how the mother’s conscious and unconscious processes influence her thinking about her role as a mother and the relationship with her unborn baby.

The development of instruments measuring these four constructs has established their role as predictors of psychosocial risk factors (Lannert et al., 2013; Malone et al., 2010; Smaling et al., 2016), caregiving behaviours (Foley & Hughes, 2018) and later infant outcomes (Le Bas et al., 2020; Branjerdporn et al., 2017; Cataudella et al., 2016; Zeegers et al., 2017). Several reviews have reported on properties and associated outcomes of instruments assessing maternal-fetal attachment (e.g. Beck, 1999; Perrelli et al., 2014; Van den Bergh & Simons, 2009) and maternal representation, as measured by the Working Model of the Child Interview (e.g. Vreeswijk et al., 2012). Whilst recent literature has reviewed instruments measuring bonding in the perinatal period (Wittkowski et al., 2020), as yet no review has evaluated and compared instruments’ psychometric properties and usability across all four theoretical constructs linked to the MFR. This review aims to address this gap by summarising the psychometric properties of instruments used to measure these constructs and making recommendations regarding the suitability of each tool for use in clinical practice.

How are constructs relating to the maternal-fetal relationship defined?

Four constructs are proposed to describe aspects of the maternal-fetal relationship (Table 1). These include the way expectant mothers think about and bond with their unborn infant (e.g. maternal representation and maternal-fetal attachment), as well as parental mentalisation constructs (e.g. mind-mindedness and reflective functioning).

Table 1. Definitions associated with each of the four constructs.

Prenatal Construct	Definitions
Maternal representation	Maternal representation, drawn from Bowlby’s concept of the internal working model, refers to the way a mother thinks about and behaves towards her unborn baby. Maternal representation is situated in the mother’s own experiences of caregiving, the expectations she has for herself as a mother and her baby, and the meaning attributed to becoming a parent. Responses are categorised into three main typologies that include ‘balanced’ (coherent, richly elaborated and warm narratives), ‘disengaged’ (aloof, minimising and sparse narratives) and ‘distorted’ (disorganised and high-affect narratives, potentially showing role confusion).
Maternal-fetal attachment	Maternal-fetal attachment refers to the mother’s sense of emotional closeness or bond towards her unborn baby, as demonstrated by her thoughts, feelings and behaviours. It differs from the search for comfort that comes with the activation of the attachment system, and instead indicates a desire to protect the unborn baby. It is seen as an early activation of the parental caregiving system.
Maternal reflective functioning	Reflective functioning is the operationalised metric of mentalisation, requiring an understanding of observable behaviour from the intentional stance. It refers to the reflexive abilities of the parent to perceive and interpret their own and their unborn baby’s behaviour in the context of existing or future mental states, such as feelings, wishes, goals and desires.
Maternal mind-mindedness	Mind-mindedness refers to the tendency to view the unborn infant as an individual with a separate mind, rather than merely an entity with needs that must be satisfied. As with reflective functioning, it involves the attribution of intentional states to behaviour, with the addition that these must be spontaneous (aka unprompted) statements. This is seen to be indicative of a natural curiosity about the (future) mind of the unborn baby.

Antenatal maternal representation describes the qualities with which a mother-to-be thinks about her unborn infant, herself as a mother and her experience of pregnancy based on her own internal representations (Rosenblum et al., 2009). These are templates, internalised from her own early relationships, that guide expectations and behaviours in later relationships (Bowlby, 1969). Maternal-fetal attachment (MFA) refers to the emotional bond the mother feels towards her unborn baby (Bowlby, 1982; Condon & Corkindale, 1997; Cranley, 1981). Condon (1993) proposes a model for conceptualising the presence and quality of MFA, highlighting expectant-mothers’ desire to ‘get to know’, protect and take care of their unborn child.

Reflective functioning is the measure of an individual’s capacity to mentalise; that is, ‘to perceive and interpret human behaviours in terms of intentional mental states such as feelings, wishes, goals or desires’ (Fonagy et al., 1995, 1991). Reflective functioning has a particular bearing on parent–infant relationships, driving parents’ curiosity to understand, anticipate, empathise and respond to their child’s needs (Slade et al., 2005). Low levels of parental reflective functioning are demonstrated in an unawareness, inaccuracy or lack of comprehension of the internal state of the infant (Fonagy et al., 2016; Meins et al., 2002; Slade, 2005). Mind-mindedness (Meins, 1997, 1998) refers to the tendency for a caregiver to view their unborn baby as an individual with a capacity for a mind, ‘rather than merely as a creature with needs that must be satisfied’ (Meins, Fernyhough, Fradley et al., 2001, p. 638). In contrast to reflective functioning, which is assessed through invitations to share active reflections, mind-mindedness requires a spontaneous focus on mental states or attributes when describing their infant (McMahon & Bernier, 2017; Meins et al., 2012; Meins, Fernyhough, Russell et al., 2001).

Why assess the maternal-fetal relationship?

Pregnancy is a transitional period, during which the mother’s representations of herself and others, internalised from early experience, activate and organise expectations of her unborn child and herself as mother-to-be (e.g. Ammaniti et al., 1992; Cohen & Slade, 1999; Stern, 1995). Previous research demonstrates that the quality of the maternal-fetal relationship (MFR) predicts multiple infant outcomes, including perceptions about infant temperament, parent–infant interaction quality, infant attachment security, adaptive behaviour and milestone attainment (Le Bas et al., 2020; Branjerdporn et al., 2017; Crawford & Benoit, 2009; Foley & Hughes, 2018; Korja et al., 2010; Siddiqui & Hägglöf, 2000).

A recent meta-analysis (Foley & Hughes, 2018) investigating concordance between antenatal parental representations and later outcomes demonstrates a moderate association between antenatal relationships (including measures of parental representation and fetal attachment) and postnatal parent–infant interaction. Furthermore, research investigating antenatal reflective functioning has found positive associations with maternal sensitivity postnatally (Smaling et al., 2016). Similarly, parents demonstrating greater antenatal mind-mindedness are likely to comment appropriately and spontaneously on their infants’ internal states (Arnott & Meins, 2007) and show better quality parent–infant interactions (Arnott & Meins, 2007). Systematic review evidence from a range of postnatal timepoints demonstrates a positive relationship between mind-mindedness and child developmental outcomes (McMahon & Bernier, 2017).

Studies have also demonstrated that maternal mentalisation can predict infant–parent attachment quality, which in addition to parental sensitivity, has been suggested as a mechanism to explain the intergenerational transmission of attachment security (Oppenheim & Koren‐Karie, 2002; Slade, Grienenberger et al., 2005; Zeegers et al., 2017). A recent study on maternal representation found a correspondence between mothers’ classifications on the WMCI in pregnancy and Adult Attachment Interview (AAI). Follow-up mediation analyses showed that the WMCI accounted for the association between AAI and infant attachment security (Madigan et al., 2015). The fact that this process between mother and baby begins antenatally provides an important window of opportunity to reduce the risks within the parent–infant relationship.

What instruments measure the maternal-fetal relationship?

To date, studies investigating the MFR have used either self-report questionnaires to record expectant mothers’ thoughts and feelings about their unborn baby, or interview-based instruments, coded to draw out mothers’ representations and reflexive capacities. Recent research comparing interview-measures and questionnaire-based tools found no evidence to suggest greater predictive validity of either type of instrument for postnatal parent–infant interaction quality (Foley & Hughes, 2018). To the authors’ knowledge, this has not yet been investigated for mentalisation constructs.

Maternal representation

Standardised semi-structured interviews assess both cognitive and affective features of maternal representation through categorising narrative patterns (Zeanah, 2000). Examples of these include the Working Model of the Child Interview (WMCI; C.H. Zeanah et al., 1986), administered during or after pregnancy; the Interview of Maternal Representations During Pregnancy (IRMAG; Ammaniti et al., 1992; Ammaniti, Tambelli & Peucchini, 2000); the Pregnancy Interview (PI; Slade et al., 2003, 2005) and the Pregnancy Interview Revised (PI-R; Smaling et al., 2016). Self-report questionnaire methods can also be used to measure perinatal representations (e.g. Child Concept Questionnaire; Gloger-Tippelt, 1983).

Maternal-fetal attachment (MFA)

MFA, which has been most widely measured, is assessed using three different tools: (1) the Maternal-fetal Attachment Scale (MFAS; Cranley, 1981) primarily measures affiliative behaviours, (2) the Prenatal Attachment Inventory (PAI; Muller & Mercer, 1993) concentrates more on the emotional relationship to the fetus, whilst (3) the Maternal Antenatal Attachment Scale (MAAS; Condon, 1993) captures both the ‘quality’ (e.g. closeness/distance; tenderness/irritation; joyful/unpleasant) and the ‘intensity’ of the relationship to the unborn baby (e.g. talking to the baby; dreaming about or interacting with him/her).

Reflective functioning and mind-mindedness

Antenatal reflective functioning can be measured using semi-structured interviews, coded with either the Reflective Functioning Scale (Fonagy et al., 1998) or the Reflective Functioning Scoring Manual (Slade, Grienenberger et al., 2005). These interviews elicit representations of attachment relationships and include the Adult Attachment Interview (AAI; George, Kaplan, & Main, 1985), the Pregnancy Interview (PI-R, Slade et al., 1987), and the Parent Development Interview (PDI; Aber et al., 1985). The latter two are primarily used during the third trimester of pregnancy and seek to obtain the reflective quality of expectant parents’ representations of themselves and their unborn baby. Furthermore, Rosenblum et al. (2008) have developed a reflective functioning scale for use with the Working Model of the Child Interview (Zeanah & Benoit, 1995). Antenatal mind-mindedness has not been well investigated, but responses to the ‘Describe your child’ question can be coded for spontaneous mental state references (Arnott & Meins, 2008).

What is already known about instruments assessing the maternal-fetal relationship?

Recent reviews have highlighted a lack of consensus around definitions of constructs related to the maternal-fetal relationship, as well as heterogeneity across instruments (Cataudella et al., 2016; McNamara et al., 2019). This may have implications for the consistency of research findings and applicability of assessment tools to clinical practice. Three systematic reviews have evaluated the psychometric properties of instruments used to measure maternal-fetal attachment. Perrelli et al. (2014) reported on four instruments, finding superior construct and criterion validity for the PAI, in comparison to the MAAS, the MFAS and a Portuguese-language questionnaire (Maternal Adjustment and Maternal Attitudes [MAMA]; Figueiredo et al., 2014). Van den Bergh and Simons (2009) reviewed the psychometric properties of the MFAS, MAAS, and PAI. It reports on the robustness of each instrument and recommends that their psychometric properties be studied in larger and more varied samples, particularly in populations with cumulative risk factors for antenatal bonding. Most recently, Wittkowski et al. (2020) reviewed perinatal instruments assessing the quality of the maternal-fetal bond, identifying five antenatal measures. They conclude that the PAI is the most psychometrically robust antenatal measure, but that the MFAS was the most accessible based on its readability.

In terms of reviews of measures representing other constructs related to the MFR, Vreeswijk et al. (2012) systematically review evidence on maternal representation as measured by the Working Model of the Child Interview (WMCI), including its adaptation and application antenatally (Benoit et al., 1997). However, this does not report specifically on psychometric properties. Similarly, Katznelson (2013) reviewed evidence for the relationship between reflective functioning (including prenatal reflective functioning, assessed with the Pregnancy Interview) and postnatal outcomes relating to the quality of the parent–infant relationship (e.g. Pajulo et al., 2012) that again, does not report on psychometric properties.

Thus, work has yet to be done to synthesise what is known about the psychometric properties of all available instruments measuring prenatal maternal representation, reflective functioning and mind-mindedness, alongside available information on measures assessing maternal–fetal-attachment. Further, no review has reported on the feasibility of all of these tools relative to one another, and provided recommendations for their use in clinical practice, which could contribute to their wider use to identify mothers in need of support in the perinatal period.

Summary and aims

Several related antenatal constructs including mind-mindedness, reflective functioning, maternal representation and maternal-fetal attachment are predictive of both parent–infant interaction outcomes and infant developmental outcomes. The conceptual complexity and variety of methodological issues inherent in the research, raises questions regarding which tools are most reliable and valid, as well as having most clinical utility, in terms of identifying expectant mothers in need of support.

This review therefore seeks to identify both self-report and interview-based instruments that would be feasible, reliable and valid for use in clinical settings. Notably, recent updates to measures (e.g. brief report of reflective functioning) may pose advantages in usability and feasibility but potentially at the expense of predictive validity. With view to informing clinician choice on selecting appropriate and feasible instruments to assess the MFR, the research questions are

1. Which instruments that are used to measure expectant mothers’ thoughts and feelings about their unborn baby have data available regarding their psychometric properties?

2. How valid and reliable are the above tools, and which have most clinical utility in identifying women in need of additional support during pregnancy?

3. Based on the above evidence, which instruments(s) for use in in primary and secondary care can be recommended to measure expectant mothers’ thoughts and feelings about their unborn baby?

Methods

The PRISMA statement informed the method and reporting of this review (Liberati et al., 2009).

Eligibility criteria

Published research articles reporting the psychometric properties of instruments measuring the MFR in expectant mothers were considered for inclusion in the review. Studies selected for review focused on at least one of the following four constructs: (1) maternal representation, (2) maternal-fetal attachment/bonding, (3) maternal mind-mindedness, and (4) maternal reflective functioning.

The publication date range was from 1995 to the present day because literature reporting on maternal–fetal-attachment has captured relevant data from studies preceding this date, whilst measures reporting on reflective functioning, representation and mind-mindedness were developed and studied in populations of expectant mothers only after 1995. In terms of other inclusion criteria, papers had to be published in English. Papers reporting on the psychometric properties of non-English versions of self-report tools were excluded. Conference abstracts, student dissertations and editorials were also excluded.

Information sources

A systematic literature search was conducted using six electronic databases: Medline, PsychINFO, CINAHL, Scopus, EMCARE and PubMed. Database searches were conducted on 31 March 2020 and updated on 31 March 2021, identifying one further paper meeting the inclusion criteria. Search strategies used both article titles and abstracts and comprised all journal articles published up to this date. The database searches were performed by one author (BIS) and accessed from the library of the University of Oxford. The search terms, used across each database, included

antenatal OR pregnan* OR prenatal OR “expectant mother*” AND “reflective function*” OR representation OR “f*etal attachment” OR “f*etal bond*” OR “mind-mindedness” OR “parent development interview” OR “mother-f*etus relationship” AND ”working model of the child interview” OR “pregnancy interview” OR “interview of maternal representation*” OR “child concept questionnaire” OR “maternal antenatal attachment scale” OR “prenatal marschak interaction method*” OR “prenatal attachment inventory” OR “prenatal parental reflective functioning questionnaire” OR “PDI” OR “PAI” OR “WMCI” OR “MFAS” OR “MAAS” OR “PAAS” OR “CCQ” OR “IRMAG*” AND psychometric* OR reliability OR validity OR consistency

The reference lists of identified review studies (N = 5) were hand-searched, yielding a further seven eligible papers.

Study selection

Abstracts were screened by two independent raters to determine if (1) the study involved a measure of representation, maternal-fetal attachment, mind-mindedness or reflective functioning, (2) the measure assessed these constructs in expectant mothers and (3) the study reported on psychometric data of the measure. The agreement for screening of abstracts was 95% (Cohen’s κ = .83). Conflicts in ratings were then resolved between the raters, to ensure that studies met all three eligibility criteria before proceeding to full-text extraction. One author then screened the full texts against the eligibility criteria.

Data extraction and synthesis

A data extraction form was devised to capture the following parameters: (1) psychometric properties: theoretical framework, reliability, validity, sample characteristics and (2) feasibility of use in a routine clinic setting: length of time of administration, number of items, availability and recommendations for research and clinical use. Where information was missing, study authors were contacted directly. Data extraction was undertaken by a second rater on N = 9 papers, yielding a data replication rate of 89% for sample characteristics (Cohen’s κ = .40) and 96% for psychometric properties (Cohen’s κ = .73). Discrepancies were resolved by discussion to reach consensus. This was an integrative review; results are presented in table format, along with a synthesis of and interpretation of findings in relation to existing literature.

Methodological quality appraisal

The QATSDD (Sirriyeh et al., 2012) was used to assess the methodological quality of the studies reviewed. This is a validated 16-item quality assessment instrument shown to have good validity and reliability (κ = 71.5%; Sirriyeh et al., 2012). It assesses a range of study designs, evaluating 14 domains relating to methodological quality. All criteria were assessed by one researcher [BIS]. Twelve papers were double-coded for quality by a second rater. Discrepancies (25%; Cohen’s κ = .43) were discussed and resolved to full agreement.

Results

Figure 1 shows the process by which studies were identified. A total of 28 studies featuring 13 tools met the inclusion criteria. Five tools focused on maternal-fetal attachment/bonding, three on maternal representation, three on maternal reflective functioning and one on mind-mindedness.

Figure 1. PRISMA flow chart indicating study selection process.

The sample characteristics of the studies are presented in Table 2, and information on results, including psychometric information and feasibility of usage is presented in Table 3.

Table 2. Quality of study and characteristics of sample for instruments assessing aspects of the maternal–fetal relationship.

					Sample Characteristics
	Tool	Citation	Study quality	N	Country	Race/ethnicity	Clinical/risk status	Age in years	Marital status	Gestation (weeks)	Employment and education
Maternal representation	Interview of Maternal Representations (IRMAG-R; Ammaniti & Tambelli, 2010).	Flykt et al. (2012)	38/42	101	Finland	Not reported	51 drug-misusing, 50 community controls	Not reported	Not reported	28–40	Not reported
		Ammaniti et al. (2013)	33/42	666	Italy	Not reported	411 without risk factors, 255 with psychosocial and depressive risk factors	M = 31.6; (SD = 4.17); 23–43	Not reported	26–32	Middle social class as assessed by level of education and job status
		Tambelli, Trentini & Dentale (2020)	36/42	50		Not reported	Community sample	M = 33.88 (SD = 4.58)	All recruited part of primiparous couple	28–36	Not reported
	Working Model of the Child Interview (WMCI; antenatal application first reported in Benoit et al., 1997, based on Zeanah et al., 1995)	Dayton et al. (2010)	33/42	164 total (42 prenatal)	United States	62% percent Caucasian, 26% African American, 5% Latina/Hispanic, 5% mixed-race, and 2% other minority groups.	Community sample, 50% had experienced domestic violence	M = 26.7 (SD = 5.0)	43% single women, 45% married, and 12% separated or divorced.	28–40	Monthly incomes ranged from $267 to $9,733 (median = $1,550; SD = $1,665).
		Atkinson et al. (2009)	35/42	102	Canada	83% Caucasian, with small percentages of other ethnicities, including Asian, Hispanic, and African Canadian.	Community sample	M = 31.89 (SD = 3.65)	93.5% of the mothers were married	28–40	Mean of 15.50 (SD = 3.17) years of education; the sample was largely middle class, with 77% earning a family income exceeding 50,000 Canadian dollars
		Benoit et al. (1997)	33/42	96	Canada	100% Caucasian	Community sample	M = 29.17 Range: 20–39	94 married, 2 single	28–40	9−27 years (M = 15.18 years)
Maternal representation	Working Model of the Child Interview-D (Crawford & Benoit, 2009 based on Zeanah et al., 1993)	Crawford and Benoit (2009)	33/42	35	United States	Not reported	Community sample	M = 30.29. Range: 25–29	Not reported	28–40	Most had some college education (M = 16.11, SD = 3.31, range = 10–27)
		Huth‐Bocks et al. (2004)	36/42	206	United States	63% Caucasian; 25% African American; 5% Hispanic; 4% Biracial; 1% Native American; 1%, 1% Asian American, 1% other	Community sample. 44% (n = 91) had experienced domestic violence, 56% (n = 115) no domestic violence	M = 25.4 (SD = 5.0) Range: 18–40	50% never married; 40% married; 5% divorced; 4% separated; 1% widowed	28–40	93 (45%) high school or less; 72 (35%) some college; 14 (7%) associates degree; 16 (8%) Bachelor’s degree; 11 (5%) graduate degree
		Huth‐Bocks et al. (2011)	36/42	173	United States	63% Caucasian; 25% African American; 5% Hispanic; 4% Biracial; 1% Native American; 1%, 1% Asian American, 1% other	Community sample. 91 (44%) had experienced domestic violence, 115 (56%) no domestic violence	M = 25.4 (SD = 5.0) Range: 18–40	50% never married; 40% married; 5% divorced; 4% separated; 1% widowed	28–40 s	93 (45%) high school or less; 72 (35%) some college; 14 (7%) associates degree; 16 (8%) bachelor’s degree; 11 (5%) graduate degree
		Lannert et al. (2013)	35/42	180	United States	63% Caucasian, 24% African American, 5% Hispanic, 4% biracial, 1% Asian Pacific, 1% Native American, 2% other ethnicity	79% (44%) had experienced domestic violence, 101 (56%) no domestic violence	M = 25 Range: 18–40	Not reported	28–40	87% reported that they had graduated high school, and reported monthly income ranged from 0 to $9500 (Mdn = $1500)
		Madigan et al. (2015)	35/42	84	Canada	100% Caucasian	Community sample	M = 29.48 (SD = 4.0) Range: 25–39	97% married/cohabiting	28–40	Average years of education = 15.73 (SD = 3.11). The average income was $50,000–80,000.
		Theran et al. (2005)	31/42	180	United States	63% White/Caucasian, 24% Black/African American, 4% mixed race, 5% Hispanic, 1% Asian-Pacific, 1% Native American, 2% Other	Community sample; 44% had experienced domestic violence	M = 25 Range: 18–40	41% married, 41% single never married, 11% separated/widowed/divorced	32–40	87% high school at least, 43% employed, mean income $1,914
		Vreeswijk et al. (2015)	35/42	308	The Netherlands	82% Dutch, 18% other	Community sample	M = 31.60 (SD = 4.30) Range: 17.34–41.88	Not reported	26	Years of education following primary school varied from 0 to 4 years (11.9%), 5 to 8 years (24.4%), or 9 years or more (63.7%). Employment: 87.1%
Maternal–fetal attachment	Maternal Antenatal Attachment Scale (MAAS; Condon, 1993)	Schwerdtfeger and Goff (2007)	34/42	41	United States	53.7% White (n = 22), 24.4% African American (n = 10)	Community sample (136 multiparous and 166 primiparous mothers)	M = 22.9 (SD = 3.77) Range: 18–33	41 participants, 48.8% (n = 20) were dating or single, 22% (n = 9) were married, 22% (n = 9) were living with a significant other, and 7.3% (n = 3) were divorced	12–41	Total income range of $9,999 to $59,999, with a majority below $19,999 (n = 31, 75.6%). 78.1% (n = 32) high school educated, 36.6% (n = 15) unemployed, 26.8% (n = 11) employed full- time, and 17.1% (n = 7) employed part-time.
		Condon and Corkindale (1997)	26/42	238	Australia	Not reported	Community sample	M = 27.4 (SD = 4.7)	Mean years in partner relationship = 4.2 (SD = 3.5)	33.7 (SD = 7.0).	Professional (24%), skilled (30%), semi-skilled (23%), unskilled (23%)
	Maternal–Fetal Attachment Scale (MFAS; Cranley, 1981)	Bloom, 1997)	29/42	79	United States	29 (60%) Black, 18 (38%) White, 1 (1%) Hispanic, 1 (1%) other	Adolescent community sample	Range: 12–19	42 (86%) single	20–29 and 30–40	Not reported
		Mercer and Ferketich (1995)	29/42	302	United States	Majority anglo-white (68% experienced mothers, 74% inexperienced mothers)	Community sample: experienced and inexperienced mothers	M = 29, (SD = 4.84), Range: 18–45	Majority married: (91% experienced mothers, 84% inexperienced mothers)	24–34	Experienced mothers: 54% professionals, 32% skilled or clerical, and 14% semi- or unskilled. Inexperienced mothers: 58% professionals, 25% skilled or clerical, and 17% semi- or unskilled
	Pictorial Representation of Attachment Measure (PRAM; Van Bakel et al., 2013)	Van Bakel et al. (2013)	36/42	352	the Netherlands	Dutch Caucasian (84%). Other Western (7%) or non-western (9%) ethnic background (e.g. Polish, Moroccan, Turkish, Surinam or Asian).	Community sample	17–44 M = 31.5 (SD = 4.4),	84 single, 268 partnered	26 weeks (SD = 1; Range 23–30).	Average of 16 years of formal education (SD = 4.1).
Maternal–fetal attachment	Prenatal Attachment Inventory (Muller, 1989, Muller & Mercer, 1993)	Gau and Lee (2003)	36/42	349	United States	18.3% African American; 5.5% Asian; 70.6% Caucasian; 2.6% Hispanic; 1.5% Native American; 1.5% African	Inpatient high-risk unit 2.9%; Outpatient high-risk OB unit 8.7%; outpatient	M = 29.21(SD = 5.33) 18–25 = 69 (20.3%); 16–35 = 214 (62.9%); 35+ = 57 (16.8%)	258 (75%) married; 13 (3.8%) divorced; 4 (1.2%) widowed; 59 (17.2%) single with partner; 10 (2.9%) single without partner	28–40	Professional = 101 (29.6%); minor professional = 137 (40.2%); clerical = 51 (15%); semi-skilled = 41 (12%) unskilled/service workers = 11 (3.2%)
		Damato (2004)	39/42	139	United States	136 Caucasian (97.8%)	Community sample	M = 32.1 (SD = 4.9)	137 (98.6%) married	26.5 (SD = 6.43)	Income less than or equal to $60,000 = 75 (53.9%)
	Prenatal Maternal Attachment Scale (LoBiondo-Wood & Vito-ORourke, 1990)	Fowles (1996)	31/42	136	United States	White (92%, n = 125)	Community sample	The average age was 26 years (SD = 3.6, range 18–35).	Married (93%, n = 126). Seven percent (n = 10) were partnered.	24–40	Approximately 60% in clerical, sales and semi-skilled workers, 40% major and minor professionals
	Prenatal Attachment Inventory (Muller, 1989, Muller & Mercer, 1993)	Gau and Lee (2003)	36/42	344	United States	18.3% African American; 5.5% Asian; 70.6% Caucasian; 2.6% Hispanic; 1.5% Native American; 1.5% African	Inpatient high-risk unit 2.9%; Outpatient high-risk OB unit 8.7%; outpatient	18–25 = 69 (20.3%); 16–35 = 214 (62.9%); 35+ = 57 (16.8%)	258 (75%) married; 13 (3.8%) divorced; 4 (1.2%) widowed; 59 (17.2%) partnered; 10 (2.9%) single without partner	28–40	Professional = 101 (29.6%); minor professional = 137 (40.2%); clerical = 51 (15%); semi-skilled = 41 (12%) unskilled/service workers = 11 (3.2%)
Mind-mindedness	‘Describe your child’ question to assess mind-mindedness	Arnott and Meins (2008)	32/42	28 mothers (3 solo), 25 fathers	United Kingdom	26 mothers White British, 2 Asian	Community sample	M = 30.5 Range: 23–41	25 married or cohabiting, 3 single mothers	M = 34.5 weeks gestation, SD = 5.04, range: 27–39	Not reported
Reflective functioning	Prenatal Parental Reflective Functioning Questionnaire (P-PRFQ; Pajulo et al., 2010)	Pajulo et al. (2015)	35/42	124 (n = 29 to review construct validity)	Finland	Not reported	Community sample	M = 30 (SD = 4.3)	121 married/cohabiting	32	Education: low (13, SD = 15) Middle (19, SD = 23) High 68 (SD = 82)
	Pregnancy Interview with Reflective Functioning Scale (Slade et al; 2002; Slade, Grienenberger et al., 2005)	Sadler et al. (2016)	28/42	30	United States	73% Latina. 27% African American	Community sample of expectant mothers in adolescence	M = 17.7 (SD = 1.5) Range: 14–19	65% single with partner contact, 25% no contact with partner, 10% married	30–39	Highest grade in school, 10.4 (SD = 2.4)
		Salo et al. (2019)	33/42	45	Finland	Not reported	Mothers with depression	Not reported	Married: 16 (40%), cohabiting 20 (50%), 4 single (10%)	22–31	Education: elementary school 4 (10%), high school/trade 19 (47.5%), University degree 13 (32.5%), doctorate 5 (10%)
		Smaling et al. (2015)	35/42	162	The Netherlands	84.6% Caucasian, 15.4% other	High (n = 83) and low risk mothers (n = 79)	M = 22.22, (SD = 2.39)	13.6% single parents	27	Monthly family income after tax 2,085.77 (1,265.37); 12.3% unemployed; 93.8% Completed at Least a Secondary Education; 11.1% financial problems
		Pajulo et al. (2012)	34/42	34	Finland	Not reported	Sample at psychosocial risk (drug-misuse)	M = 25.1 (SD = 5.8) Range: 16–38	15 (44%) single parents	21–39 weeks, M = 30.8 (SD = 5.8)	Only basic education (9 years max): 24 (77%)
	Pregnancy Interview Revised with Reflective Functioning Scale (Slade et al; 2002; Slade, Grienenberger et al., 2005)	Smaling et al. (2016)	35/42	142	The Netherlands	89% Caucasian; 11% other	Community sample	M = 22.85; (SD = 2.21) Range: 17–25	93% in a relationship, 7% single parents	27 weeks	32% completed a Bachelor’s or Master’s degree. Monthly family income after tax 2,456.99 (1,215.84);

Table 3. Psychometric properties and feasibility of instruments assessing the maternal-fetal relationship.

Construct	Tool/author	Tool format	Number of items	Subscales or forms	Range and interpretation of output	Reported reliability	Reported validity	Response format and ease of scoring	Clinical or research use	Feasibility	Availability
Maternal- fetal attachment	Maternal Antenatal Attachment Scale (MAAS; Condon, 1993)	Self-report scale	19 items	Two subscales: Attachment (n = 11 items) and Preoccupation (n = 8 items)	19–95 Higher scores indicate stronger attachment bond	Internal consistency α = .69 to .85 (total scale) Individual subscale coefficients ranging from r = .52 to .73	Content: Yes Criterion: Yes Construct: Yes	Self-report; 5-point Likert scale yielding a sum total (simple scoring)	Both	Brief to administer (<5 min)	Available online, free of charge
	Prenatal Maternal Attachment Scale (PMAS; LoBiondo-Wood & Vito-O’Rourke, 1990)	Self-report scale	39 items: 29 items at any stage; 10 items after fetal movement	Not reported	19–95 Subscales: 11–50 (quality); 8–40 (preoccupation) Higher scores indicate stronger attachment bond	Internal consistency: α = .87 (29 item) and .83 to .90 (39 item)	Content: Yes	Self-report; 4-point Likert scale yielding a sum total (simple scoring)	Research use	Brief to administer (<10 min)	Available online, free of charge
	Maternal Fetal Attachment Scale (MFAS; Cranley, 1981)	Self-report scale	24 items (revised to 23 in Mercer & Ferketich, 1995)	8 rating scales assessing qualitative features, 8 rating affective tone	24–120 Higher scores indicate stronger attachment bond	Internal consistency: α = .88 (total scale) Test–retest = Yes	Content: Yes Criterion: Yes	Self-report; 5-point Likert scale yielding a sum total (simple scoring)	Both	Brief to administer (<5 min)	Available online, free of charge
	Prenatal Attachment Inventory (PAI; Muller, 1993)	Self-report inventory	21 items	Uni- dimensional	21–84 Higher scores indicate stronger attachment bond	Internal consistency: α = .81 to .93	Content: Yes Structural: Yes Criterion: Yes Construct: Yes	Self-report; 4-point Likert scale yielding a sum total (simple scoring)	Both	Brief to administer (<5 min)	Available online, free of charge
	Pictorial Representation of Attachment Measure (PRAM; Van Bakel et al., 2013)	Representation of bond as measured by distance between sticker on page (baby) and circle (mother)	Not relevant	Not relevant	Distance: range of 0 to 9.3 cm (lower score = stronger attachment bond)	No reliability analyses performed. Single-item measure, no item analyses or inter-item consistency measures.	Criterion: Yes Construct: Yes	Self-report; measurement of distance (simple scoring)	Both	Brief to administer (<5 min)	Available online, free of charge
Maternal representation	Working Model of the Child Interview (C. H. Zeanah et al., 1994)	Structured clinical interview	15 items	5 rating scales	Derives one of three categorisations; balanced, distorted, disengaged	Internal consistency: α = .94 Inter-rater reliability: κ = .62 to .94 (p < .001)	Content: Yes Structural: Yes Criterion: Yes Construct: Yes	Open ended; clinician rated with a 5-point Likert scale (specialist interpretation required)	Both	Time-consuming to administer (60 min)	Available online, specialist training required to score
	Working Model of the Child Interview-D (Crawford & Benoit, 2009; based on WMCI, C.H. Zeanah et al., 1986)	Structured clinical interview	15 items	8 rating scales assessing qualitative features; 8 rating affective tone	Derives one of four categorisations; balanced, distorted, disengaged, disorganised	Internal consistency: α = .84 to .91 (T1), .78 to .90 (T2), .81 to .92 (T3). Kappa = .94 for overall classification	Content: Yes	Open ended; clinician rated with a 7-point Likert scale (specialist interpretation required)	Both	Time-consuming to administer (60 min)	Available online, specialist training required to score
	Adjective Scale Questionnaire from Interview of Maternal Representations (Ammaniti et al., 1999)	Lists of pairs of adjectives with a Likert-scale to indicate rating	17 pairs of adjectives	5 lists of 17 pairs of mixed adjectives, divided into four representational areas	Range of 3–21. A higher score indicates a more positive, balanced representation.	Internal consistency: α = .84 to .91 (T1), .78 to .90 (T2), .81 to .92 (T3).	Content: Yes Structural: Yes	Self-report; 7-point Likert scale (specialist interpretation required)	Research use	Brief to administer (<10 min)	Unknown, no information available
	Interview of Maternal Representations Revised (IRMAG-R; Ammaniti & Tambelli, 2010).	Structured clinical interview	41 items	7 dimensions including herself as a mother (A1 – A7) and descriptions of infant (B1–B7) are coded for, reflecting features of the mother’s narrative	Derives one of three categorisations: integrated/balanced, restricted/disinvested; not integrated/ambivalent	Inter-rater reliability: 94% (κ = .83, p < .001) Pearson’s r = .89, p = .01 (coherence), Pearson’s r = .96, p = .01 (emergence of fantasies). Mean reliability = .89-.92.	Content: Yes Structural: Yes Criterion: Yes Incremental: Yes	Open ended; clinician rated with 5-point Likert scale (specialist interpretation required)	Both	Time-consuming to administer (45–60 min)	Unknown, no information available
Maternal reflective Functioning	Prenatal Parental Reflective Functioning Questionnaire (P-PRFQ; Pajulo et al., 2010)	Self-report instrument	33 items Revised to 14	Three factors identified	Variable scaling of scores, representing higher quality reflective functioning (high-low, low-high, middle)	Internal consistency: α = .65 (33 item); .77 (14 item)	Content: Yes Structural: Yes Construct: Yes	Self-report; 7-point Likert scale (simple to score, but variable scaling of item responses needs consideration)	Both	Brief to administer (<5 min)	Available online, free of charge
	Pregnancy Interview (Slade et al., 1987, pp. 2002, 2005) with Reflective Functioning Coding Manual (Fonagy et al., 1998)	Semi-structured clinical interview	24-items	Three subscales suggested in paper, 6 items each	Higher scores represent better reflective functioning	Inter-rater reliability: 80% (α = .73 to .95) Internal consistency: α = .78 (anticipation), .73 (differentiation), .78 (interaction)	Construct: Yes	Open ended; clinician rated with 9-point Likert scale (specialist interpretation required)	Both	Time-consuming to administer (60–90 min)	Available from authors to researchers and clinicians trained in its use
	Pregnancy Interview Revised (Slade et al., 1987) with Reflective Functioning Coding Manual (Slade et al; 2003; Slade et al., 2005)	60- to 90-minute semistructured interview	22-items	Not relevant	Higher scores represent better reflective functioning	Mean interrater agreement = .87; overall RF-score = .90	Structural: Yes	Open ended; clinician rated with 11-point scale from −1 (negative RF) to +9 (full or exceptional RF) (specialist interpretation required)	Both	Time-consuming to administer (60–90 min)	Available from authors to researchers and clinicians trained in its use
Maternal mind-mindedness	‘Describe your child’ question, Meins, Fernyhough, Russell et al., 2001	Single question, with open ended response	1 item	Not relevant	Scores for appropriate mind-related comments are expressed as a proportion of the total number of comments produced by the caregiver	Inter-rater reliability: κ = .94	Content: Yes Construct: Yes	Open ended; responses categorised by clinician, frequency of mentalistic comments calculated (specialist interpretation required)	Research use	Brief to administer (<10 min)	Available online, free of charge

Study characteristics and quality

The papers identified were published between 1995 and 2019. Samples were recruited from the United States (N = 13), the Netherlands (N = 4), Canada (N = 3), Finland (N = 4), Italy (N = 2), the United Kingdom (N = 1), and Australia (N = 1). Seven studies reported on sample sizes ≤50, 13 on sample sizes ≤200, and 7 on samples >200, with the smallest sample being 30 (Sadler et al., 2016) and the largest 666 (Ammaniti et al., 2013). Most sample populations comprised expectant mothers in the third trimester of pregnancy (N = 18). Fourteen out of 28 studies included a sample of ‘at-risk’ women. Caucasian participants were heavily represented across studies, with only two studies reporting on a majority of participants from other racial/ethnic groups (Bloom, 1997; Sadler et al., 2016).

Information on study quality as determined by the QATSDD (Sirriyeh et al., 2012) is presented alongside sample characteristics in Table 2. Quality ratings demonstrated a trend whereby more recent papers scored higher across all domains. Notably, no studies reported on co-production in relation to either the development of the measure or the study design. Few measures (N = 3) described the statistical consideration of sample size. Psychometric properties were discussed in all studies. Papers reporting on self-report tools presented a greater range and depth of information on psychometric properties due to these either being the main or secondary aim of the study; and papers describing interview measures reported such information in the description of the measure, along with a short comment on inter-rater reliability and/or agreement. No studies reported on predictive validity.

In terms of measures of maternal–fetal attachment, papers reporting on the self-report PAI were of the highest quality (M = 37.5), whilst papers reporting on the MFAS were of the lowest quality (M = 29), reflecting the level of detail and clarity of reporting. Papers reporting on the Pregnancy Interview showed the most variation in quality, with scores ranging from 29 to 36. For example, Sadler et al.’s (2016) small sample size resulted in limitations in terms of representativeness and lacked information on recruitment. Smaling et al. (2015) provided a comprehensive overview of theoretical issues, methodology and results, though (like many) did not report on either sample size calculations or the use of co-production in study design. Pajulo et al. (2015), reporting on the P-PRFQ, and van Bakel et al. (2013), reporting on the PRAM, showed a similar profile.

Psychometric properties

Measures of maternal-fetal attachment

The Maternal-Fetal Attachment Scale ([MFAS], Cranley, 1981), Maternal Antenatal Attachment Scale ([MAAS], Condon, 1993), Pictorial Representation of Attachment Measure ([PRAM], Van Bakel et al., 2013), Prenatal Maternal Attachment Scale ([PMAS], LoBiondo-Wood & Vito-ORourke, 1990) and Prenatal Attachment Inventory ([PAI] Muller, 1989, Muller & Mercer, 1993) were identified in the search.

The MFAS is a 37-item self-report scale, revised to 19-items. It is widely used in clinical research, reporting good reliability across the included studies. Bloom, 1997) reports a Cronbach’s coefficient alpha of .88 for the total scale, with subscales – identified through expert consensus – at .73 for role taking, .53 for interaction with fetus, .56 for differentiation from fetus, .55 for giving of self and .77 for attributing characteristics.

The MAAS is a 19-item self-report scale comprising two subscales identified by factor analysis: ‘quality of attachment’ (11-items) and ‘intensity of preoccupation’ (8-items). There is some evidence for construct validity in its correlation with other measures of social wellbeing (Condon & Corkindale, 1997), as well as adequate reliability as assessed by internal consistency (Schwerdtfeger & Goff, 2007). The lack of norms, categories, or cut-off scores for the MAAS limits its utility in clinical settings.

The PMAS is a 29-item self-report scale, with an additional 10 items available to women who have experienced fetal movement. Reported psychometric properties include good internal consistency (α = .90) and content validity. Currently, the PMAS is used in research settings only.

The PAI is a 29-item with good reliability and validity, including concurrent validity with the Maternal Fetal Attachment Scale, construct validity with Maternal Adjustment and Maternal Attitudes Scale, content validity achieved through review by panel of expert clinicians, theorists and new mothers, and concurrent validity with the Maternal Attachment Interview (Damato, 2004). Confirmatory factor analysis (CFA) has found that it fits a unidimensional scale (Gau & Lee, 2003); the advantage of CFA, as opposed to an exploratory approach, is that it allows for testing of an a priori hypothesis, in keeping with the development of the PAI which is based on theory and literature review.

Finally, the PRAM is a single pictorial representation of the maternal-fetal bond; it showed convergent validity with verbal and non-verbal measures of maternal-fetal attachment assessed with the MAAS. However, as it is a single-item measure, no item analysis or inter-item consistency measures can be performed, nor were other investigations of validity or reliability reported (Van Bakel et al., 2013).

Measures of maternal representation

Studies reporting on three interview-based measures of maternal representation were identified: the Working Model of the Child Interview ([WMCI], C. H. Zeanah et al., 1994), the Working Model of the Child Interview-D ([WMCI-D] Crawford & Benoit, 2009; based on the WMCI; C.H. Zeanah et al., 1986), and the Interview of Maternal Representations-Revised ([IRMAG], (Ammaniti & Tambelli, 2010). These require between 45 to 60 minutes for delivery, with versions of the WMCI consisting of just 15 items compared with the IRMAG-R which comprises 47 items. One self-report form was also identified (ASQ-IRMAG; Ammaniti et al., 1992).

The IRMAG-R is a revised version of the IRMAG interview (Ammaniti et al., 1999), with five further questions exploring the influence of the mother’s trauma, fears and preoccupations. The coding system allows the mother’s transcript to be classified as integrated/balanced, restricted/disinvested, or not integrated/ambivalent. Interrater reliability for IRMAG-R scales showed a mean reliability of .92. Interrater reliability for the main category was 94% (Cohen’s κ = .83, p < .001). Factor analyses were reported with principal component analysis showing two factors, indicative of good structural validity. Evidence demonstrates the criterion and incremental validity of IRMAG-R as predictive of postnatal emotional availability and attachment security (Tambelli, Trentini & Dentale, 2020). Flykt et al. (2012) report on an adjective scale questionnaire form from the IRMAG: whilst it showed good reliability (κ = .83), no evidence for validity was reported. The authors note the questionnaires are available for research use only.

The WMCI is a 60-minute structured interview examining perceptions and subjective experiences of the current and future relationship between mother and baby. Questions can be changed into the future tense to assess antenatal representations in pregnant women. Percentage agreement for overall classification was between 85% (κ = .57, Benoit et al., 1997) and 96% (κ = .62 to .94; p < .001; Dayton et al., 2010).

Studies evaluating the properties of the WMCI-D (Huth‐Bocks et al., 2004, 2011; Lannert et al., 2013; Madigan et al., 2015; Theran et al., 2005; Vreeswijk et al., 2015), adapted from the WMCI to capture disorganised representations, have reported on reliability and validity. The WMCI-D shows good inter-rater agreement (90%, κ = .71; Crawford & Benoit, 2009) and good construct validity, with an 83% correspondence with the Adult Attachment Interview (Madigan et al., 2015). The subscales that were developed for the WMCI-D scale have variable interrater reliability.

Measures of maternal reflective functioning

Three instruments assessing maternal reflective functioning were identified: the P-PRFQ (Pajulo et al., 2015), PI (Slade et al; 2003; Slade, Grienenberger et al., 2005) and PI-R (Slade et al., 1987, pp. 2002, 2011).

The PI and PI-R are semi-structured interviews, lasting between 60 and 90 minutes and capturing parents’ reflective functioning capacity. The coding for the PI shows good inter-rater agreement (80%; Sadler et al., 2016) and reliability (α = .73 to .95; Salo et al., 2019). Construct and structural validity are shown to be robust (Smaling et al., 2015, 2016). Extensive training on scoring the transcribed interviews is required to access the coding framework. The P-PRFQ (Prenatal Parental Reflective Functioning Questionnaire; Pajulo et al., 2010) is a 14-item self-report instrument, designed to address this issue. It shows good reliability (α = .77) and reports both good content and structural validity, with factor analysis demonstrating a three-factor model: ‘Opacity of mental states’, ‘Reflecting on the fetus–child’, and ‘The dynamic nature of mental states’ (Pajulo et al., 2015).

Measures of maternal mind-mindedness

Mind-mindedness has not been well studied in populations of expectant mothers. Only one study was identified that used a measure of mind-mindedness in the antenatal period (Arnott & Meins, 2008). This asked mothers an adaptation of the ‘Describe your Child’ question typically used to assess mind-mindedness, assessing the degree to which expectant mothers use mentalising, behavioural, physical or general descriptors to predict what their child will be like. Coding with the mind-mindedness coding manual (Meins & Fernyhough, 2006) showed good inter-rater reliability (κ = .94). Although the construct validity of antenatal mind-mindedness was not demonstrated, it showed good discriminant validity with respect to the MAAS (Condon, 1993). The measure is currently only recommended for research use.

Feasibility for use within clinical settings

The usability of the tools within clinical or research settings varied. An overview of factors relating to usability is provided in Table 3. The majority of instruments (N = 10) were designated suitable for use in both clinical and research settings. The number of items for self-report tools ranged from 14 (P-PRFQ; Pajulo et al., 2010) to 39 (PMAS; LoBiondo-Wood & Vito-O’Rourke, 1990). For interview-based tools, it ranged from 15 (WMCI; C. H. Zeanah et al., 1994) to 47 (IRMAG-R; Ammaniti & Tambelli, 2010); the time for administration was longest for the Pregnancy Interview (Slade et al., 1987, pp. 2002, 2005) at up to 90 minutes, and shortest for the IRMAG-R (Ammaniti & Tambelli, 2010) at 45–60 minutes. The majority of tools, whether self-report or clinician-rated, used a response or coding format of Likert scales (N = 11). Seven tools were self-report instruments, while the remaining six were clinician-rated, using coding manuals that require specialist training to administer. Whilst most tools could be located online and were free to use, the Pregnancy Interview coding scheme is only provided to practitioners trained in its application.

Discussion

The purpose of this systematic review was to identify and evaluate the quality and availability of contemporary measures of representation, fetal attachment/bonding, mind-mindedness and reflective functioning as applied in populations of expectant mothers, including their psychometric properties. The measurement of these constructs is important for policymakers as they serve as strong predictors of later difficulties in the parent–infant relationship and later infant outcomes (e.g. Branjerdporn et al., 2017), thereby providing an important window of opportunity in terms of prevention.

What tools are used to measure the maternal-fetal relationship?

Thirteen tools were identified, each measuring one of the four central theoretical constructs relating to the MFR. The majority of instruments measured attachment/bonding (N = 5), followed by representation (N = 4), reflective functioning (N = 3) and finally mind-mindedness (N = 1).

How valid and reliable are tools used to measure the maternal-fetal relationship?

The papers identified for review varied in the extent to which they reported on instrument reliability and validity, highlighting limitations in the availability of psychometric information. For example, the MFAS (Cranley, 1981), whilst widely used in clinical research studies, shows significant gaps in the evaluation and reporting of validity. Previous reviews (e.g. Beck, 1999; Doan, Cox & Zimmerman, 2003; Van den Bergh & Simons, 2009) note longstanding issues with both content validity and the method through which the dimensions were constructed, which was primarily through expert perspective as opposed to statistical methods.

Of the tools assessing MFA, studies using the PAI (Muller, 1989, Muller & Mercer, 1993) demonstrated the strongest evidence for reliability and validity across a range of domains. One issue identified across MFA tools was the relative lack of known-groups validity – that is, differentiation between community and at-risk samples – compared to tools measuring reflective functioning and representation. Further research is needed to identify cut-off scores and criteria for raising concerns about the developing maternal-fetal bond. Of instruments assessing maternal representation, studies reporting on the WMCI (C. H. Zeanah et al., 1994) showed the most substantial evidence for reliability as well as demonstrating content, structural, criterion and construct validity. The Adjective Scale Questionnaire from the IRMAG also showed promise in this regard, although would benefit from further research. Finally, regarding antenatal mentalisation, studies using the Pregnancy Interview (Slade et al., 2002) reported most robust evidence for reliability and validity. Meanwhile, initial investigations of the P-PRFQ (Pajulo et al., 2010) indicated potential with respect to its psychometric properties. Measurement of antenatal mind-mindedness has only been reported in one study, which was limited in its reporting of psychometric data. Further investigation is required before it can be recommended for use in clinical settings.

What is the feasibility of tools in identifying mothers in need of further support?

The feasibility of using each tool was defined in relation to a number of different factors (e.g. number of items, time to administer, self-report or clinician-rated, suitability for use in clinical settings and availability). Mothers considered at highest risk include those who have experienced childhood maltreatment (Malone et al., 2010) and domestic violence (e.g. Huth‐Bocks et al., 2004; Lannert et al., 2013). Other risk factors include detrimental health behaviour during pregnancy (Fares & Adler, 1998), the presence of two or more children under 7 years in the household (Pajulo et al., 2006), poor marital adaptation (Sitrin, 2001; Fares & Adler, 1998) and maternal education, social support, and substance use (Smaling et al., 2016).

Brief self-report tools are most suitable for use in primary care settings. For example, the 29-item PAI (Muller, 1989, Muller & Mercer, 1993) performs well in psychometric terms, presents minimal participant burden, and is easily scored by clinicians. In contrast, interview-based measures may be less well suited to screening, requiring application of a coding scheme that is both costly and resource-intensive. Unfortunately, this has implications for the assessment of maternal representation, reflective functioning and mind-mindedness and raises questions about whether and why these constructs are under-represented in the literature compared to maternal-fetal attachment. The 14-item self-report tool, the P-PRFQ (Pajulo et al., 2010), measuring reflective functioning, may present a viable alternative. However, having been developed more recently, it is still in the early stages of validation and reliability testing. Further investigation is needed to confirm its utility and robustness as a screening tool.

Implications and recommendations for clinical practice

Whilst there is no literature directly comparing the predictive validity of different MFR tools, previous research demonstrates a relationship between psychosocial risk factors, constructs related to the MFR, and postnatal outcomes. This includes evidence for the detrimental impact of maternal risk factors on quality of antenatal representation (Vreeswijk et al., 2012, 2015), reflective functioning (Huth‐Bocks et al., 2004; Smaling et al., 2015; Theran et al., 2005), and maternal-fetal attachment, with further implications for their association with postnatal bonding (McNamara et al., 2019). The accurate identification of at-risk mothers provides an important window of opportunity to deliver targeted support to reduce the likelihood of adverse child outcomes. Further research should identify which construct, and indeed which measure, is most strongly associated with parent–infant interaction quality and infant outcomes.

Findings from this review suggest that self-report instruments, such as the PAI and MAAS, can be recommended for use in assessing maternal-fetal attachment in primary care settings (see Table 4). Both show good psychometric properties and usability, with higher scores indicating a stronger attachment bond. Worth noting is that self-report measures may be less well-suited for at-risk populations where reflexivity, self-awareness and insight may be compromised. The high face validity of these measures may result in expectant mothers over-reporting their degree of bonding with their unborn baby, particularly in contexts where they may feel stigmatised. Following further investigation into its psychometric profile, the self-report P-PRFQ may be a more suitable, less transparent measure for screening levels of reflective functioning. Moreover, as a relatively brief measure that does not require specialist training to administer, it can be used by practitioners such as midwives, health visitors and public health nurses in the third trimester of pregnancy. Importantly, it should be noted that in the absence of cut-off scores and norms for self-report instruments, where lower scores indicate concerns within the MFR, clinicians should also utilise structured professional judgement to identify mothers in need of further support. Accordingly, scores on self-report measures should be interpreted alongside any pre-existing risk factors. Where lower scores on self-report measures are present, alongside other risk factors, onward referral may be necessary to tertiary services that have the resources to utilise interview measures and can offer further intervention.

Table 4. Recommendations for instruments to be used in primary and secondary care settings.

	Instrument	Strengths	Weaknesses
Primary care	Prenatal Attachment Inventory ([PAI] Muller, 1989, Muller & Mercer, 1993)	Self-report (21-items), freely available for clinical use, quick to administer and easy to interpret irrespective of professional background Substantial evidence for robust psychometric properties Unidimensional, providing an overall score indicating strength of attachment bond Lower scores indicate compromised attachment bond, indicating potential relevance of onward referral for further assessment	High face validity, indicating validity may be compromised in populations where insight is limited, or expectant mothers are anxious about service involvement No cut-off scores or clinical norms available
	Maternal Antenatal Attachment Scale ([MAAS], Condon, 1993)	Self-report (19-items), freely available for clinical use, quick to administer and easy to interpret irrespective of professional background Substantial evidence for robust psychometric properties Provides scores across two subscales: attachment (level of bond) and preoccupation (level of focus on baby) Lower scores indicate compromised attachment bond, indicating potential relevance of onward referral for further assessment	High face validity, indicating validity may be compromised in populations where insight is limited, or expectant mothers are anxious about service involvement No cut-off scores or clinical norms available
	Prenatal parental reflective functioning questionnaire (P-PRFQ; Pajulo et al., 2010)	Self-report (14-items), freely available for clinical use, quick to administer Low face validity, appropriate for use in populations where insight may be compromised or anxiety about service involvement contributing to over/under-reporting Emerging evidence for psychometric robustness, requiring further investigation Lower scores indicate compromised reflective functioning, indicating potential relevance of onward referral for mentalisation-based intervention	Further psychometric investigation needed Interpretation requires awareness of variable scaling of scores across items; may be more complex to interpret for clinicians who lack prior knowledge of mentalisation theory and associated profile of reflective functioning deficits No cut-off scores or clinical norms available
Secondary care	Interview of Maternal Representations Revised (IRMAG-R; Ammaniti & Tambelli, 2010)	Derives one of three representation categorisations to inform conceptualisation of difficulties in the maternal-fetal relationship (MFR): integrated/balanced, restricted/disinvested; not integrated/ambivalent Can be used in more complex populations to inform formulation and treatment planning in relationally focused interventions	Clinician administered interview, requiring specialist training to deliver, with associated costs Lengthy to administer (45–60 min) Requires specialist training in use of coding manual as well as significant time-commitment to interpret
	Working Model of the Child Interview-D ([WMCI-D] Crawford & Benoit, 2009)	Derives one of four representation categorisations to inform conceptualisation of difficulties in the MFR: balanced, distorted, disengaged and disorganised; the latter category renders it useful for application in settings where parents may have experienced complex trauma Can be used in more complex populations to inform formulation and treatment planning in relationally focused interventions	Clinician-administered interview, requiring specialist training to deliver, with associated costs Lengthy to administer (60 min) Requires specialist training in use of coding manual as well as significant time-commitment to interpret
	Pregnancy Interview Revised (PI-R; Slade et al., 1987, pp. 2002, 2011)	Provides rich account of the expectant parent’s reflective capacities, indicating the potential relevance of a mentalisation-based intervention to support the parent–infant relationship following birth Can be used in more complex populations to inform formulation and treatment planning, particularly with respect to services offering mentalisation-based interventions	Clinician-administered interview, requiring specialist training to deliver, with associated costs Lengthy to administer (60–90 min) Requires specialist training in use of coding manual as well as significant time-commitment to interpret

In such specialist settings, the Pregnancy Interview Revised should be used to ascertain reflective functioning capacities. This interview would not only help to clarify the level of support needed but also has the potential to facilitate reflexivity by helping the mother to consider her thoughts and feelings towards her unborn baby, thus acting as an intervention in itself. These women could then be directed towards appropriate resources (e.g. the Getting to Know Your Baby programme, www.your-baby.org.uk), advice around techniques to improve the developing relationship, in addition to early intervention services. These might include mentalisation-based interventions, such as Minding the Baby (Sadler et al., 2013), which may be beneficial for women with more challenging psychosocial histories and contexts (e.g. domestic violence, drug misuse, experience of trauma).

Strengths and limitations

To our knowledge, this is the first review to synthesise what is known about the psychometric and clinimetric properties of instruments measuring the quality of the maternal-fetal relationship across all four theoretical constructs, providing recommendations for use in clinical practice. One limitation of this systematic review is the possibility that not all instruments for evaluating the MFR were identified or included. There is also potential for gaps in information on psychometric properties owing to the exclusion of papers reporting on translations of self-report tools. For example, studies reporting on Polish and Swedish translations of the PAI yield a five-factor structure based on principal component analysis (Bielawska‐Batorowicz & Siddiqui, 2008; Siddiqui et al., 2000) as opposed to the single-factor model, reported in the present paper, derived from confirmatory factor analysis (Gau & Lee, 2003).

Additionally, a comprehensive evaluation of other areas of feasibility has not been reported, including information regarding required reading level and length/cost of training for interview-based measures. Unfortunately, this information was not available in any paper selected for review, although it is an important aspect of feasibility. Similarly, despite efforts to contact authors of instruments included in this review, the authors of the IRMAG-R could not be reached to ascertain its availability and cost. Finally, whilst the QATSSD was chosen as a quality-rating tool because it has the scope to assess studies of a range of methodologies included in the review, there is a need for the development and evaluation of quality rating tools that specifically assess studies reporting on both psychometric and usability data.

Recommendations for further research and clinical practice

Further research should continue to evaluate the psychometric properties, availability and feasibility of using these tools such that clearer comparisons can be made of their psychometric properties where information is currently lacking. The development of the P-PRFQ is promising with a view to broadening the study of antenatal reflective functioning where the resource-intensive use of interview-based instruments might otherwise limit the scope of its application. Work should also be undertaken to compare the predictive validity and strength of association between instruments measuring each of the four constructs and postnatal parent–infant outcomes, with a view to establishing which are the most helpful in identifying mothers in need of support. Additional efforts should be made not only in early identification but also to ensure that such support is both available and accessible. Services may want to consider a stepped care model to demarcate appropriate care pathways, with psychoeducational materials, guided self-help and support groups followed by more intensive provision such as mentalisation-based treatment programmes (Sadler et al., 2013). Accordingly, future directions for research include establishing clear clinical cut-offs and criteria within these measures to identify mothers at-risk of experiencing difficulties in their relationship with their unborn baby.

Conclusion

The quality of the maternal-fetal relationship has important implications for parent–infant interaction and infant outcomes, which in turn have significant longer-term effects on child development. This suggests a need for further research into instruments measuring the MFR that can be used to enrich current understanding of risk and transmission pathways. Advances in this understanding represent an important opportunity to anticipate risk of difficulties in the parent–infant relationship and thereby offer preventative intervention to disrupt the intergenerational transmission of attachment difficulties and scaffold positive outcomes for child socio-emotional development. Yet, this possibility depends on successful translation of theoretical understanding into clinical practice. The present review contributes recommendations for robust, clinically feasible instruments that can be used by clinicians to identify expectant mothers in need of further support.

Disclosure statement

No potential conflict of interest was reported by the author(s).