The feasibility of an objective measure of the parent-child relationship in health visiting practice: assessment of the Maternal Postnatal Attachment Scale

Background: Positive parent infant relationships are key to achieving long term child outcomes. Identifying parents who may need support is difficult because of a lack of robust assessment tools. Working in partnership with health services we piloted the Maternal Postnatal Attachment Scale (MPAS) in a deprived, multi-ethnic urban community in Bradford, UK. The pilot aimed to assess the clinical utility of MPAS to identify need for support: Was it administered to a representative group of women? Is MPAS valid for this population? Methods: Data were linked to a cohort study in the pilot area (Born in Bradford’s Better Start - BiBBS). Chi Square tests assessed sample representativeness (age, ethnicity, parity, English language, education, deprivation). Exploratory factor analysis explored MPAS’ validity. Results: 563 women in BiBBS were eligible, 210 (37%) completed MPAS. No differences were found between completers and non-completers, suggestive of a representative sample. In total, 336 women (including a number of women living in the service area who had not participated in BiBBS) completed MPAS in the pilot. MPAS had ceiling effects and a satisfactory factor structure could not be identified, indicating poor psychometric properties. Conclusions: Health visitors were successful in administering MPAS to a representative sample, but the lack of psychometric robustness indicates that MPAS is unsuitable for routine use in this setting. A gap for such a measure remains.


Introduction
Background The ability of a mother to interact with her infant sensitively, whilst attuned to their infant's mental state and level of development is a crucial precursor of a child's ability to develop a secure attachment (Ainsworth et al., 1978;Kim et al., 2017). Secure attachment predicts a child's later social-emotional development (Fearon et al., 2010;Le Bas et al., 2020). Insecure attachment prevalence rates are estimated to be high, at 35% in a Danish community study (Skovgaard, 2010), whilst rates of the rarer social functioning disorder reactive attachment disorder in the UK are 1. 4% (Minnis et al., 2013) (reactive attachment disorder is a social functioning disorder related to abuse and neglect and is associated with significant psychiatric morbidity). Attachment is not routinely assessed in, or beyond, infancy, and prevalence is difficult to gauge because of this and the lack of robust measures to assess attachment.
The importance of a healthy parent-infant relationship for children's future development is recognised in the UK in three national clinical guidelines, with assessment, early identification and intervention being key recommendations (NICE guidelines CG192 -antenatal and postnatal mental health (NICE, 2020), NG26 -attachment in children in care (NICE, 2015), and PH40 -social and emotional well-being: the early years (NICE, 2012)).
However, despite the clear importance of early identification and intervention NICE guidance acknowledges that no tools have been identified for use for the 0-12 months postnatal period -a critical time point to allow early identification and prevention of issues. The tools recommended by NICE for identification in pre-school children require clinical expertise and observations making them expensive for use in universal services (NG 26, NICE, 2015). Two recent reviews of self-report tools for measuring maternal dimensions of the parent-infant relationship concluded that no available measures could be recommended for use, in the main due to the lack of evidence about the clinical utility and psychometric properties of the tools (Mathews et al., 2019;Wittkowski et al., 2020). NICE Guidance (NG26) notes this gap in their research recommendations where they state the need to "Develop reliable and valid screening assessment tools for attachment and sensitivity that can be made available and used in routine health, social care and education settings" (NICE, 2015).
In the UK all children and their parents receive a minimum of five mandated visits from a health visitor from pregnancy up to 2.5 years of age to support the child's safety and development (see Box 1 for further details on the role of health visitors). Early parent-infant relationship is recognised as one of the main priorities for health visiting in Early years high impact area 2: Maternal and family mental health (PHE, 2020). However, as far as we are aware, the majority of health visitors across the UK rely on personal observations and professional judgement to identify issues with the parent-child relationship (Appleton et al., 2013;Wilson et al., 2010). Such assessments are subjective and hard to validate. In addition, such observations are recorded in free text rather than coded sections of a healthcare record making extraction of such assessments on a population level challenging. A lack of validated and coded recording may impact on the chances of high quality, joined up clinical care for mother and baby, and limits the ability of researchers and health organisations to characterise prevalence and epidemiology more accurately, identify local levels of need and plan for service provision.

Box 1. The role of health visiting
Health visiting in England is a universal service which visits all children in the home at least five times between 28 weeks of pregnancy and five years of age. Health visitors (Specialist Community Public Health Nurses) are the main health professional in contact with young families and have a key role in health promotion and health needs assessment. There are no eligibility criteria to be met for health visiting and it is provided universally and is free at the point of use. Health visitors are post registration nurses who have taken a one-year masters level specialist training and undertake all health needs assessments of families although some elements of support are offered by less qualified staff. As well as proactive health visiting support focused on child development, maternal health and family transition families also have access to primary care through General Practitioners and emergency care via hospitals.
In Bradford, as a part of the Better Start Bradford programme, (see Box 2), the decision was made to pilot the implementation

Amendments from Version 1
We have amended the text in the Background section to: 'Insecure attachment prevalence rates are estimated to be high at 35% in a Danish community study (Skovgaard, 2010), whilst reactive attachment disorder (RAD) rates in the UK are 1. 4% (Minnis et al., 2013). RAD is a rarer social functioning disorder related to abuse and neglect and is associated with significant psychiatric morbidity. Attachment is not routinely assessed in, or beyond, infancy, and prevalence is difficult to gauge because of this, and the lack of robust measures to assess attachment.' Below we offer more detail around the required sub-samples. The information below has been threaded through the manuscript in the abstract, methods, and results section to facilitate the readers' understanding.
Three sub-samples were needed to fulfil the two objectives -those who had been offered and had completed the MPAS, those who had been offered it and not completed it, and those that had not been offered it.
The research team only had demographic data for parents who were participating in the BiBBS cohort. The BiBBS data enabled us to assess the representativeness of the sample to ensure that our subsequent MPAS psychometric analyses were robust.
There were 833 eligible women across the BSB pilot area, 563 of these were in the BiBBs cohort. In total 435 of the 833 women were offered the MPAS (210/435 were in the BiBBs cohort), and 398 were not offered it (353/398 were in the BiBBs cohort). In total 302 partially/fully completed the MPAS. The psychometric analyses included MPAS data from 198 parents who had fully completed it in English.

REVISED
of an objective and validated assessment tool into universal health visiting practice within an inner-city area of Bradford. Given the lack of recommendations for a tool, the research team worked together with the health visiting service to complete a brief review of potential measures focussing on evidence of validity and reliability as well as potential clinical utility. This review used the same methodology as a larger review by the team (Blower et al., 2019). A number of measures were considered (Brockington et al., 2001;Cuijlits et al., 2016;Høivik et al., 2013;Müller, 1994;Taylor et al., 2005). The Maternal Postnatal Attachment Scale (MPAS) (Condon & Corkindale, 1998) was selected as the best option (from the few existing appropriate measures) for the pilot based on previous research with this measure, its psychometric properties, and because it is freely and easily available.

Box 2. Better Start Bradford
Better Start Bradford is a ten year (2015-2025), £49 million investment by the National Lottery Community Fund in three deprived and ethnically diverse areas of Bradford, UK. It aims to give children a better start in life by providing services that offer preventative early interventions to parents and children aged 0-4 years. Interventions include universal parenting programmes, and more targeted support where needed including an intervention to improve mother-child relationships called Little Minds Matter.
Born in Bradford's Better Start (BiBBS) is a prospective interventional birth cohort study that was initiated to evaluate the impact of the Better start Bradford interventions. BiBBS recruits women during pregnancy who live in the Better Start Bradford areas. An in-depth questionnaire at recruitment collects information on the socio-emotional and economic circumstances of the women and their families; and women also consent to linkage of their and their baby's routine health and education data as well as Better Start Bradford project data (Dickerson et al., 2016). An interim cohort profile of BiBBS (n=2,626) demonstrated that the participants in this cohort are representative of the eligible pregnant population (Dickerson et al., 2022) Better Start Bradford's success is in part reliant on early identification and appropriate referrals into services. BiBBS success is also, in part, reliant on outcomes collected within routine health data using validated tools. At the initiation of this programme, there was no validated objective measure of the mother-child relationship in the health visiting service making referrals into the Little Minds Matter project, and measuring the impact of services on this measure, challenging. It was for these reasons that the pilot described in this paper was undertaken.

The MPAS
The MPAS was developed by John Condon and colleagues in Australia (Condon & Corkindale, 1998). It is a 19-item measure suitable for use with mothers in the first postnatal year. The items are a mixture of forward and reverse scored items with either 2, 3, 4 or 5 answer categories. Each item is equally weighted so some of the item response categories has decimal scoring. The maximum score is 95, and the theoretical minimum is 19. Lower scores indicate more problematic responses. The MPAS does not have validated cut off points for problematic or concerning relationships and is not intended for use as a diagnostic tool on its own, but as a supportive indication within a holistic assessment.
The developer of the MPAS has assessed that validity of the measure and it is described as suitable for use in research and clinical practice (Condon & Corkindale, 1998). A sample of 238 women recruited antenatally completed MPAS at three different timepoints (4 weeks, 4 months, and 8 months). Stability of the measure over time was acceptable (all Pearson correlation coefficients significant at p<0.001) and internal consistency of the measure was acceptably high (alphas>0.7). Factor analysis found that the items loaded onto three factors: Quality of attachment, Absence of hostility and Pleasure in interaction (Condon & Corkindale, 1998).
MPAS has not been widely validated, with only five studies which validate the measure (Condon & Corkindale, 1998;Feldstein et al., 2004;Riera-Martín et al., 2018;Scopesi et al., 2004;van Bussell et al., 2010). These studies were all included in the review by Wittkowski and colleagues (2020) which concluded that the MPAS (and the other included measures) lack evidence of validation, and that if using the measures consideration needs to be given to the robustness of the findings.
The aim of this paper was to assess the clinical utility of the MPAS in universal health visiting services in a disadvantaged and ethnically diverse population.
Specific objectives were to:

·
Explore how feasible and acceptable implementation of this tool was within standard health visiting practice in a disadvantaged and ethnically diverse population · Evaluate the validity and reliability of the tool when used within standard health visiting practice in a disadvantaged and ethnically diverse population

Methods
This was a quantitative study using descriptive statistics to assess the clinical utility of the tool, and using exploratory factor analysis to assess the structural validity and internal consistency of the tool.

Implementation of the MPAS pilot
The MPAS was piloted as a universal assessment at the 3-4-month health visiting contact, over a 1-year period between 8 th May 2017 and 8 th May 2018. The 3-4 month contact is not one of the nationally mandated contacts but is an additional universal contact offered in Bradford. For women who did not speak English, there were options of an Urdu translated MPAS, administered by a bilingual health visitor, or support from a bilingual health visitor or interpreter for other languages.
Several health visitors in the Better Start Bradford (BSB) area speak community languages and work predominately with families in these languages.
Training on MPAS administration and scoring, and what to do if concerns were identified, was provided by local perinatal mental health specialists. A referral pathway into local perinatal mental health services, discrete interventions and children's services was developed.
Health visitors were asked to record: if the MPAS was offered; if declined, the reasons for this; whether it was self-completed, completed with the help of the health visitor or with an interpreter; and the language used to complete the tool.

MPAS sample eligibility
All women with babies, living in the pilot BSB areas, who had a 3-4-month health visitor contact were eligible to complete the MPAS. Of those who were eligible, women who had a reference to the MPAS assessment in their health record were defined as having been offered the assessment. Those who had no record of any questions being completed were defined as not participating, and reasons for non-participation were reviewed.
To assess clinical utility, those who had one or more questions completed in their health record were defined as having participated in the MPAS assessment, and those who completed 15 or more of the 19 questions were defined as having completed the MPAS.
Pilot study eligibility a) Clinical utility All women seen by health visitors for a 3-4 month visit within the time period of the pilot (8 th May 2017 and 8 th May 2018) for whom routine health data was available were included in the analysis of coverage and completion. For the representativeness analysis, BiBBS participants who had an infant aged 3-4 months between the 8 th May 2017 and 8 th May 2018 (the time period of the pilot study) and were living in the Better Start Bradford area were included.

b) Validity & Reliability
The same routine health data used for the coverage and completion analyses were used for the factor analysis. However, participants who did not complete the MPAS in English were excluded, as were all participants who did not complete all 19 questions (see Figure 1).

Data sources
Routine health visitor data for all eligible women was anonymised and shared with the research team. For the representativeness analysis, data on the characteristics of eligible women in the pilot area were obtained using the BiBBS research cohort (see Box 2). Data included sociodemographic characteristics for all women with infants aged 3-4 months. As part of the BiBBS cohort, routine health visiting data (including MPAS data) were linked to cohort data. This enabled a comparison of women in the cohort who did and did not participate in the MPAS assessment.

Objective 1: Feasibility and acceptability of implementation
An acceptable measure would be one which health visitors are willing to ask, and one which women are willing to complete (de Vet et al., 2011). Three key factors were explored, as recommended by de Vet et al. (2011): a) Coverage: What percentage of eligible women took part in the MPAS pilot? b) Completion: What percentage of eligible women completed the tool? Adequate completion was defined as 85% women or higher completing at least 15/19 questions. c) Representativeness: Are there differences in the characteristics of women who took part in the pilot and completed the MPAS, compared to women who were eligible but did not take part?
For a) and b), descriptive statistics were calculated for all eligible women (women with a 3-4-month health visitor check during the pilot period) using routine service data. For c), we compared age, ethnicity, English language ability, education and material deprivation in the BiBBS data using a Chi Square test for differences in proportion. Missing data led to a casewise deletion.

Objective 2: Validity and reliability of the tool
The content validity of the MPAS was established in the original development study (Condon & Corkindale, 1998). but there is limited evidence about other measurement properties. Therefore, the structural validity and internal consistency of the MPAS were assessed in the pilot using exploratory factor analysis (de Vet et al., 2011;Mokkink et al., 2018;Prinsen et al., 2018;Terwee et al., 2018).
This method used a staged approach that: 1) determined the missingness and variation of scores on individual MPAS items; 2) identified the level of correlation between items, and; 3) provided an interpretation of the structural validity and internal consistency.
In stage 1, items which did not show any variation were identified and removed from the analysis. The remaining items were taken forward into stage 2 where a correlation matrix using Pearson correlation coefficients was constructed. For Structural Validity, any items that did not correlate with at least one other item with a coefficient > 0.2 were identified and removed from the analysis, and similarly any items with a coefficient of > 0.9 were identified and removed. No restrictions were made as to the number of factors to be returned by the analysis.
Assessment of the adequacy of the sample for factor analysis was made using the Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy, which should be above 0.5, and Bartlett's test of sphericity, which should be significant. Exploratory factor analysis was selected because using MPAS in routine service and using MPAS in the UK were both new uses of the tool. Items were required to load onto their factor with a loading of at least 0.5, and items which did not meet this threshold were deleted from the measure item by item. Items which have loading of over 0.3 onto multiple factors in the final measure will be considered for deletion but may be retained Eigenvalues were calculated, and a Scree plot was created to determine the number of factors to retain in the analysis. A threshold of a minimum combined proportion of variance of >50% explained by the factors was set.
For internal consistency Cronbach's alpha were calculated for each subscale of the above factor analysis, with an expected minimum of 0.7 and of >0.9 being desirable. When the Cronbach's alpha was not satisfactory, items with poor correlation were deleted and the Cronbach alpha was recalculated.

Ethics approval
The Health Research Authority confirmed that the pilot of the MPAS is considered to be service evaluation, not research, and as such does not require review by an NHS Research Ethics Committee (HRA decision 60/88/81 February 2017). The BiBBS study received ethical approval by Bradford Leeds NHS Research Ethics Committee (15/YH/0455), and research governance approval from Bradford Teaching Hospitals NHS Foundation Trust. All data were anonymised prior to analysis and are stored securely at the Bradford NHS Teaching Hospital.

Results
Objective 1: Feasibility and acceptability of implementation During the study period, 37 health visitors were working in the pilot area and 833 women had a 3-4-month visit. In total, 35 of 37 (95%) health visitors completed at least one MPAS assessment, and the number completed per health visitor ranged from 1 to 66.
Of the 833 eligible women, 435 (52%) had been offered the MPAS and of these, 347 (42% of total eligible women and 80% of those offered the MPAS) women participated in the assessment. Reasons for not participating included refusal, having a conversation instead of using the measure, another person present, and inadequate time. Of the 347 who participated, 302 (87%) completed the assessment. 563 BiBBS participants were eligible for this study. Of these women, 210 had been offered the MPAS assessment. There were no significant differences between women who were and were not offered an MPAS for any of the characteristics examined (See Table 1).
Objective 2: Validity and reliability of the tool 198 MPAS assessments were available for the factor analysis (see Figure 1). Overall, MPAS scores were skewed, with the vast proportion of women scoring very high, indicative of no concern (Figure 2). 21% of women who completed the MPAS in English scored the maximum score of 95 on the tool.
Item response was high. The highest proportion of observed missing data (in 6.7% of cases) was for question 9 ("When I leave the baby…"). This item level missingness is not high enough to suggest that the item should be dropped (de Vet et al., 2011).
Item variation identified that for 9 of the 19 questions, at least one of the response categories was not used by any of the participants. In the case of question 14 "I now think of the baby as…", all 198 women selected the response "very much my own baby". Due to the lack of variation in response in question 14 this item was dropped from the next stage of analysis. Table 2 shows the correlation matrix of the remaining 18 items of the MPAS. Items 7 ("When I am with the baby and other people are present, I feel proud of the baby…") and 12 ("When I am with the baby: … try to prolong time I spend with…") were not correlated with any of the other MPAS items with correlation coefficients <0.2 and were removed from further analysis. The factor analysis moved forward with 16 items (i.e. without question 7, 12 and 14).
Examination of the scree plot ( Figure 3) shows that there is an indication that a three-factor solution similar to that identified by Condon & Corkindale (1998), may be relevant in this population, however, this three-factor solution only explained 41% of the variance. A six factor solution based on all factors with an eigenvalue of >1 explains 67% of the variance. However, on further examination, (Table 3), factor six has only one variable loading onto it. Removing this variable (question 2), another variable (question 9) no longer loads onto the factor solution, leaving factor five with one variable loading onto it (question 4). Removing these three variables leads to a 10-item scale with a four factor solution that explains 60% of the variance. Whilst this solution meets the KMO test for sampling adequacy, and Bartlett's test of sphericity, interpreting the factor structure highlights that, as well as a relatively large number of factors from just ten items, there are three items (8,15,18) which are loading (at >0.3, but less than <0.5) onto multiple factors impairing interpretation of the factor structure of the tool. As no meaningful factors can be extracted from the MPAS data there is no ability to assess the internal consistency of the extracted factors.

Discussion
This study assessed the clinical utility, validity and reliability of the MPAS in universal health visiting services in Bradford. This is the first time that this assessment tool has been used in clinical practice anywhere (to the authors' knowledge at the time of writing), and in a disadvantaged and ethnically diverse population. In the pilot, health visitors' use of the tool was inconsistent and only 52% of eligible women were offered an MPAS assessment by the health visitor at the 3-4-month visit. There was considerable variation between health visitors in how often they used the MPAS with their case load, with some health visitors never recording using it, and one using the tool 66 times during the pilot. There were no socio-demographic differences in who was and was not offered the MPAS, suggesting that health visitors were not being biased in who they offered the assessment to. Of those offered the MPAS, 80% participated and 87% completed suggesting that, when offered, it is acceptable. However, the distribution of the scores was highly skewed with little variance and no indication of any concerns detected in the scores. Furthermore, the analysis on the validity of the MPAS tool in this population failed to find evidence of internal consistency or structural validity. The findings suggest that women in the pilot study population did not interpret or respond to the MPAS questions as intended.
Further in-depth exploration of these findings is required to understand why some health visitors used the tool inconsistently, and what the barriers were to completing the tool with almost half of the eligible population. It is important to understand whether the barriers related to the design of the tool or to contextual factors that could be addressed to improve uptake.
Further exploration is also required to understand the lack of variance in the scores in the pilot study. Whilst this could relate to a lack of validity with women perhaps not understanding or interpreting the questions as intended due to cultural and/or language differences. There may however also be reluctance for women to disclose concerns about their relationship with their baby to health professionals. Previous research completed with a similar population has shown that women from ethnic minorities are less likely to have their perinatal mental health identified by health professionals due to a complex interplay of reluctance to disclose (e.g. due to stigma, fear of having their baby taken away), difficulty in identification by health professionals (e.g. use of interpreters, lack of time etc.,) and problems in capturing issues on IT systems (Prady et al., 2021;Prady et al., 2016a;Prady et al., 2016b).
An additional research study has explored these explanations using qualitative interviews with health visitors during this pilot study. The linked paper by Bird et al. (2022) explores these issues further. Key findings from this paper suggest that although health visitors welcomed the opportunity to discuss the parent infant relationship and there were benefits to using a structured tool, there were also considerable challenges that hindered implementation of the MPAS in a valid and reliable way. Health visitors had concerns around the length of time required to administer the tool, the complexity of the language and the intrusiveness of some questions. These concerns were exacerbated when translation was used. The context that health visitors were working in and lack of time for home visits also posed challenges. Together,    the papers highlight the need for a robust, valid measure to assess parent-child relationships in routine practice, with coproduction to ensure clinical utility and acceptability.
Strengths of this study include the evaluation of the use of the MPAS in routine health visiting practice, meaning that findings relate to 'real world' use of the measure. The evaluation builds on a successful partnership between the service and evaluation teams, from working together to identify a suitable measure through to evaluation and implementation of findings into practice. This meant that the evaluation considered both theoretical and operational perspectives.
There are two key limitations to this study. Firstly, the BSB population has an unusual profile. The population in BSB are very ethnically diverse (only 10% of the women giving birth in the area identify as White British) and economically deprived, live in an urban area, and are not representative of the wider UK population. As such the findings may not be valid for less deprived, less ethnically diverse, or more suburban/rural communities. It is vital that any objective tool is feasible to implement and meaningful to use with all women and health professionals.
Secondly, the routine setting of the study meant we were reliant on health visiting data, not all of which we had full access to. We had no information about the 50% of eligible women who had contact with a health visitor but who had no information recorded as to if they were asked to complete MPAS. Not knowing why MPAS was not asked in these cases limits our ability to understand how acceptable and useful the MPAS was to both women and health visitors. These limitations mean that caution must be exercised when generalising the findings of the BSB pilot.

Conclusions
The MPAS was administered to a representative sample by health visitors, but acceptability was low, and the MPAS had poor psychometric properties. Qualitative research (Bird et al., 2022) confirms that the MPAS was not fully understood by the sample, rendering it unacceptable for the Bradford context. Although health visitors welcomed the opportunity to discuss the parent infant relationship, there were also considerable challenges. This included concerns around the complexity and length of the tool itself and the time-pressured context that health visitors were working in.

Implications for practice and/or further research
Based on the findings from this paper, and Bird et al. (2022), the gap for a robust, valid measure to assess parent-child relationships in routine practice remains, at least in Bradford. Considering this, we coproduced a tool with health visitors, service staff and with input from parents, based on the learning from the current study, and have piloted it in routine care (Bywater et al., 2022).

Data availability
The data are stored securely by Born in Bradford (BiB) at the Bradford Institute for Health Research (BIHR). Data sharing is not applicable to this article, because the participants did not give permission for their data, collected during the service evaluation (not research), to be shared. However, restricted access to an anonymised data set will be considered on a case by case basis, dependant on the relevance of the research question and its' ability to be answered using the existing data.
Before you contact BiB, please make sure you have read our Guidance for Collaborators. The decision for restricted access will be made by our BiB Executive Committee, which reviews proposals on a monthly basis, and we will endeavour to respond to your request as soon as possible. You can find out about all of the different datasets which are available here. If you are unsure if we have the data that you need please contact a member of the BiB team (borninbradford@bthft.nhs.uk).
Once you have formulated your request please complete the 'Expression of Interest' form available here and email the BiB research team (borninbradford@bthft.nhs.uk). Please indicate clearly that you are applying for the restricted dataset used in this article. If your request is approved, we will ask you to sign a collaboration agreement; if your request involves biological samples, we will ask you to complete a material transfer agreement.