Mother-infant interaction in mother and baby unit patients: Before and after treatment

Maeve Kenny a, Susan Conroy a, Carmine M. Pariante a, Gertrude Seneviratne b,Susan Pawlby a,*

1. Introduction
Research has shown that untreated maternal mental illness has profound effects on the woman, family and the developing child, in relation to the child’s own attachment relationships (Hipwell et al., 2000; Wan and Green, 2009) and subsequent psychiatric and psychological morbidity (Pawlby et al., 2009; MacCabe et al., 2007).
Moreover, different psychotherapeutic interventions addressing the mother’s mental health needs per se have not been found to be effective in improving the outcome for the baby (Murray et al., 2003; Forman et al., 2007). When mothers suffer from mental health problems, early interventions with their babies, such as those offered during an admission to an MBU, may be crucial in improving outcomes (in the short and longer term) for both the
mother and for the developing child. From the perspective of the baby, recent UK government documents, such as the Allen (Allen, 2011) and Munro reports (Munro, 2011), highlight the importance of early years’ interventions to promote optimal outcomes for  children.

Since the work of Main, describing the treatment of mothers with their children at the Cassell Hospital in 1958, the UK has led the way in developing Units to admit psychiatrically ill mothers with their babies (Cassell and Coleman, 1990). Mothers admitted to such units may suffer from a range of disorders including puerperal
psychosis, psychotic disorders, severe affective illness, as well as disruption of the mother infant relationship. The National Institute for Health and Clinical Excellence Guidelines for Antenatal and Postnatal Mental Health (NICE, 2007) recognised the importance of specialist perinatal inpatient services, specifically for mothers and infants, typically with 6e12 beds, to be staffed by specialist perinatal mental health staff and having available the full range of therapeutic services.

However, there are virtually no studies demonstrating the effectiveness of admissions to Mother and Baby Units (O’Keane, 2006). Changes and, in particular, improvements in the mothere infant relationship are important markers of the effectiveness of an admission to a psychiatric Mother and Baby Unit for both the mother and the baby (Joy and Saylan, 2007). In economically challenging times, providers of publicly-funded national health services are also keen to see evidence relating to the effectiveness and cost effectiveness of such specialist services. The economic argument in the longer term is compelling, and Heckman and Masterov (2004) have described the significant financial saving by providing interventions and services early in a child’s life to prevent subsequent difficulties, compared with intervening later in the child or adult’s life. Thus, a psychiatric Mother and Baby Unit provides an ideal environment in which to be offering early interventions. During the first year of life, infantecaregiver interaction plays an essential role in organising the infant’s experience. Good infante caregiver interaction may moderate the risk of adverse biological or psychosocial factors on a child’s development (Tronick and Weinberg, 1997). There is evidence that maternal mental illness has a detrimental effect on mothereinfant interaction (Steadman et al., 2007) and therefore is an important aspect of the treatment of mothers with SMI on an MBU. Moreover there is also evidence that treating the mother’s symptoms per se improves her mental health, but that this does not extend to the interaction between herself and her baby (Murray et al., 2010). This highlights the importance of addressing this specific therapeutic need with interventions that are beyond the mere treatment of the mental disorder. One way of focussing on infantecaregiver interaction is through video feedback intervention which aims to enhance the developing relationship between a parent and the child. Some preliminary findings suggest that it could be an effective way of improving interaction in mothers with SMI and their babies (Stein et al., 2006). However, research also points out that video feedback intervention may be beneficial in improving interaction but this cannot be separated from other factors such as improving parental confidence (Bilszta et al., 2012).

This paper explores the extent to which admission to the Channi Kumar Mother and Baby Unit, and taking part in video-feedback intervention, is efficacious in improving mothereinfant interaction in women with SMI. The Channi Kumar Mother and Baby Unit  is a 13-bedded in-patient unit at the Bethlem Royal Hospital, South London & Maudsley NHS Foundation Trust & King’s Health Partners in the UK. The multidisciplinary team of psychiatrists, psychologists, nurses, occupational therapists, social workers and nursery nurses combine treatment of the mother’s mental illness with work to promote her relationship with her baby and develop skills to care for the baby. This includes the use of psychiatric interventions such as medication, psychological therapies, and video interaction work where the psychologists support the mother to interpret the baby’s cues and enhance sensitive caregiving. We thus test the hypothesis that, following admission to the MBU and receiving a video feedback intervention, mothers with a severe mental illness were more sensitive and less unresponsive interacting with their infants on discharge than on admission, and their infants were more cooperative and less passive. A second approach for evaluating the intervention is based on comparing the mothereinfant interaction of MBU inpatients to mothers with psychiatric diagnoses living in the community who had not received in-patient care or a video feedback intervention, and to a group of community well mothers; here we test the hypothesis that, after admission, and receiving the intervention, MBU maternal and infant interaction scores improved so that there were no differences when compared with a healthy reference group, and their scores were better than those of the
community ill reference group.

2. Method

2.1. Participants

Participants were 138 mothereinfant pairs. Forty nine mothers were in-patients on a Mother and Baby Unit (MBU group); 67 mothers had a psychiatric diagnosis but were living in the community (community ill group); 22 mothers had no psychiatric diagnosis (healthy group).

2.2. MBU group

The mothers were a representative sample of in-patients on an MBU, a 13-bed ward for mothers who have an acute postpartum psychiatric illness where they are admitted with their babies (aged 0e52 weeks) for treatment (for further details see Pawlby et al., 2010). Patients were given a range of psychotropic medication which was tailored to their symptoms and changed accordingly through the admission. This included antidepressants, antipsychotics, mood stabilisers, anxiolytics and sedative medication. As part of their treatment programme, mothers are supported in developing their relationships with their babies with individuallytailored feedback following video-recordings of mothereinfant interaction. The first video is made within one week of admission, when mothers are unwell, and the final one following an assessment by the multidisciplinary team confirms that the mother no longer needs in-patient care. Her mental health is stable and care of her baby is good enough for her to be discharged home with her baby into the care of the community psychiatric team. Discharge would take place at the earliest opportunity following this assessment made in ward round. The mean time between admission and discharge was 13.04 weeks (SD 8.04, range 1-33). The number of video feedback sessions in the study reported here ranged from 2 to 9 (mean 2.88, SD 1.32) depending on the length of the mother’s admission to the MBU. There was a high correlation (r =.64, p < .001) between the duration of the mother’s admission and the number of video feedback sessions. The number of sessions per week (dosage) ranged from 0.07 to 2 (mean 0.30, SD 0.29). DSM-IV diagnoses (APA, 1994) were given retrospectively to each mother by the consultant perinatal psychiatrist (G.S.), based on ICD-10 diagnoses made during admission together with details from the discharge summaries. Mothers were then assigned to one of three clinical groupings based on symptom profile: (1) schizophrenia (DSM-IV code: 295.30, n =8); (2) depressive mood disorders with or without psychosis, ‘depression group’ (296.20, n = 8; 296.23, n = 1; 296.24, n = 3; 296.30, n = 3; 296.33, n = 6; 296.34, n = 1; 296.54, n = 1); (3) mood disorders where manic symptoms were present, with or without psychosis, ‘mania group’ (296.41, n = 1; 296.60, n = 4; 298.90, n = 13). One mother had a comorbid personality disorder and two had mild learning disabilities. At the admission video the mean infant age was 12.4 weeks (range = 1-52 weeks) and at the discharge video was 20.6 weeks (range = 5-61 weeks); 63% of the babies were male.

2.3. Community ill group

A comparison group of 67 mothers with comorbid diagnoses of depression and personality disorder were used as a sample of convenience. They had been recruited into a different study from the postnatal ward of a large local hospital in the same geographical area as the Mother and Baby Unit and were all living at home with their babies. Comparison data were drawn from a single videotaped play session with their infants (Conroy et al., 2010).  Participants were diagnosed with DSM-IV personality disorder, using the Structured Clinical Interview for DSM IV Personality Disorders M. Kenny et al. / Journal of Psychiatric Research 47 (2013) 1192-1198 1193 (SCID-II; First et al., 1997) and with DSM-IV depression using the mood disorder module of the Structured Clinical Interview for DSM-IV Axis I Disorders Non-patient Edition (SCID-I NP; First et al., 2002). There was considerable diagnostic overlap between personality disorder clusters and categories. Just over a third of women in the co-morbid group met criteria for a cluster A PD (n = 25; 37%); two-fifths met criteria for a cluster B PD (n =27; 40%) and more than two-thirds met criteria for a cluster C PD (n = 47; 70%). The most prevalent categories of PD were avoidant (n =27; 40%) borderline (n = 25; 37%), paranoid (n = 24; 36%) and obsessive
compulsive (n = 23; 34%). None of the mothers required in-patient admission for their mental health nor did they receive any video feedback intervention. Further details of treatment in the community were not available. The mean infant age was 8.9 weeks (range ¼ 6-15 weeks, SD ¼ .2) and 45% were male.

2.4. Healthy group

A second comparison group of 22 mother-infant dyads were randomly selected from the control group who participated in a study of mother-infant bonding difficulties. The mothers had no current or past psychiatric history and no bonding difficulties with their babies. They provided data from a single video-taped play session. The mean infant age was 22.32 weeks (SD = 4.47 weeks, range = 14-29 weeks) and 50% were male.

Sociodemographic characteristics of the three groups are shown in Table 1. There were no significant differences between the groups in maternal age at the time the interaction video was made, or in the mothers’ ethnicity or cohabiting status, or in the baby’s gender. Compared to women in the healthy group, mothers in both the MBU group and in the community ill group were more likely to be multiparous. Infant age was significantly different between the three groups: the MBU babies were significantly older than the babies of the community ill group, but there was no significant difference in the ages of the MBU and healthy groups.

2.5. Procedure

All mothers were invited to have a 3-min unstructured play session with their baby video-recorded. They were asked to play and talk with their baby as they would normally, preferably without the use of toys. Video recordings for MBU mothers were made in a specially-designed laboratory setting. Infants were seated in a baby-seat with mothers sitting adjacent to them. A mirror, placed beside the baby chair, provided the camera with a full view of the mother’s face. During their admission all mothers took part in an individually-tailored video-feedback treatment package designed to promote maternal sensitivity. The aim of the feedback session is to make the experience as positive as possible for the mother and to build on the skills and strengths that she already has. The therapeutic work, with a professional trained in infant observation, is primarily behavioural in approach and centres on watching and learning about the baby’s cues so that the mother becomes more sensitive to seeing her baby as a person and learning what works and what does not work in creating a smooth dialogue between them. Feedback addresses the different domains of communication used by both mother and infant: facial expression, body language and verbal communication. Play sessions of the comparison groups were recorded on one occasion in the participants’ homes, with no feedback given.

2.6. CARE-index

The video-recorded play sessions were analysed using the CARE-Index, a screening tool developed for research purposes to assess the quality of adulteinfant interaction. It is based on 3- 5 min of unstructured playful interaction that occurs under nonthreatening conditions and is videotaped. Because this procedure is robust in terms of physical context, the videotaping can be done in a clinical setting, research laboratory or home (Crittenden, 2004). The CARE-Index assesses adult sensitivity in a dyadic context. Specifically, “adult sensitivity in play is any pattern of behaviour that pleases the infant and increases the infant’s comfort and attentiveness and reduces its stress and disengagement” (Crittenden, 2004; page 3). Therefore, it assesses the dyadic relationship as opposed to assessing an individual. The coding focuses on seven aspects of both adult and infant behaviour. Four aspects focus on affect (facial expression, verbal expression, affection and body contact) and three focus on temporal contingencies (turn-taking, control and developmental appropriateness of chosen activity). Each aspect of adult and infant behaviour is evaluated individually and summed to make seven scale scores. For adults, these are sensitivity, unresponsiveness and controlling. For the infant (birth to 15 months of age) they are cooperativeness, difficultness, compulsivity and passivity. Interactions receive a score on each aspect of adult and infant behaviour and scores are then summed to create the seven scale scores, each on a range from 0 to 14. The scales of maternal unresponsiveness and controlling are highly and negatively related to one another (r=-.75, p < .001) and so we thus only report the maternal scales of sensitivity and unresponsiveness. Likewise, the infant scales of difficultiness and passivity are significantly and negatively related (r =-.51, p < .001), as are compulsiveness and passivity (r=-.33, p < .001). Thus we only report the infant scales of cooperativeness and passiveness.


2.7. Inter-rater reliability

Coding training for the CARE-Index is based on the dyadic nature of the interaction. The rater is trained to look at the mother and infant interaction in relation to the other and assesses the fit between the two. Reliability to score the interaction is gained for both mother and infant. The same rater codes both mother and infant. All the raters were trained to the same high level of reliability, reaching 80% agreement on each scale.

The videotaped recordings of the MBU and healthy groups were all coded by two trained raters who were completely unaware of the study’s hypotheses and of the fact that some mothers had a diagnosed mental illness. Differences in the codings were discussed and consensus scoring reached. A third rater coded the play sessions of the comparison ill group, and 15% of these video recordings were rated by a fourth trained assessor. The inter rater intraclass correlation coefficient for maternal sensitivity was 0.90, indicating acceptable inter-rater reliability.

2.8. Statistical analysis

Distribution and skewness of scores were assessed for normality and linearity and as data were normally distributed parametric analyses were used. Tests for homogeneity of variance were conducted and found not to be significantly different between the groups. General linear repeated measures models were used to compare MBU interaction scores at admission and discharge with time as the within-subjects factor. Differences in the interaction scores between the diagnostic groups on the MBU were further examined adding diagnostic group as the between-subjects factor. A general linear model was used to compare the MBU group scores at discharge with the scores of the community ill group and the healthy group. Full ethical approval was granted for the three studies that provided data for this paper. Informed written consent was obtained for the recordings of the interactions to be used for research and all procedures were conducted in accordance with the British Psychological Society ethical guidelines. Mothers with psychiatric illness were informed that they could withdraw at any time without implications for their treatment.

3. Results

3.1. Mothereinfant interaction on admission and at discharge from the mother and baby unit

Changes in both maternal and infant behaviours between admission and discharge were explored in a series of general linear repeated measures models with the infant’s age on admission added as a covariate. Mothers were significantly more sensitive and less unresponsive, and infants more cooperative and less passive, at discharge than on admission with infant age controlled for (Table 2). Moreover, the changes in maternal sensitivity and infant cooperativeness were positively correlated (N = 46, Pearson’s r =.50, p < .001). The analyses were repeated with the dosage of video feedback sessions per week as an additional covariate. There were no significant independent effects of the dosage of video feedback sessions on the changes in maternal or infant behaviour. Compared with admission, at discharge mothers’ scores continued to be more sensitive (F1,45 = 6.30, p < .02; effect size, partial h 2 ¼ .12), less unresponsive (F1,45 = 12.99, p < .001; effect size, partial h 2 = .22) and the infants continued to be more co-operative (F1,43 = 12.60, p < .001; effect size partial h 2 = .23) and less passive (F1,43 = 14.17, p < .001; effect size, partial h 2 = .25).

3.2. Changes in mother-infant interaction scores on admission and at discharge by diagnostic group

Admission and discharge scores for maternal and infant interaction are reported separately for each of the three MBU clinical groups (schizophrenia, depression, mania) in Table 3. Women in each of the diagnostic groups had improved scores for sensitivity (F1,44 = 19.50, p < .001; effect size, partial h 2 = .31) and unresponsiveness (F1,44 = 23.17, p < .001; effect size, partial h 2 = .34) on discharge compared with on admission, while controlling for infant age. There was no significant effect of diagnosis on the change in either maternal sensitivity or unresponsiveness scores from admission to discharge, as shown by a lack of group by time interaction, (respectively, F2,44 = .91, p = .41; effect size, partial h 2 = .04 and F2,44 = .17, p = .85; effect size, partial h 2 ¼ .01). Compared with their scores on admission, infants of mothers in the three clinical groups were more cooperative (F1,42 = 23.48, p < .001; effect size, partial h 2 = .36) and less passive (F1,42 = 26.59, p < .001; effect size, partial h 2 = .39) on discharge, even when controlling for infant age, and again there was no significant diagnosis by time interaction (respectively, F2,42 = .83, p = .44; effect size, partial h 2 = .04 and F2,42 = 3.14, p = .053; effect size, partial h 2 = .13).

3.3. MBU group scores following a video feedback intervention, compared to those of a community ill group and a healthy group.

Videos of the two reference groups were made on only one occasion. The mother and infant scores of the community ill group, who received no video-feedback intervention, provided a benchmark against which to measure the scores of the MBU dyads, following the video feedback intervention (Figs. 1-4). The scores of the healthy dyads were used to test the hypothesis that the scores of the MBU dyads, following the video feedback intervention, would not differ from those of healthy mothers (Figs. 1-4). Maternal parity was unrelated to any of the mother and infant interaction measures and was thus not considered further.

3.3.1. Maternal sensitivity and unresponsiveness

Infant age was unrelated to the maternal measures of sensitivity and unresponsiveness and was thus not included in the analysis of variance. Following video feedback intervention there was a significant difference in sensitivity (F2,129 = 3.13, p < .05; effect size, partial h 2 = .05) and in unresponsiveness (F2,129 = 4.31, p < .02; partial h 2 = .06) between the 3 groups (Figs. 1 and 2). Post hoc analyses showed that MBU mothers were more sensitive (p < .05) and less unresponsive (p < .02) to their infants than community ill mothers, and did not differ from the healthy group.

3.3.2. Infant cooperativeness and passiveness

Infant age was correlated with the infant measures of cooperativeness (r = .16, p = .06) and passiveness (r = .21, p < .05) following the video feedback intervention, and was thus included as a covariate in the analyses of covariance. There was a significant difference in the cooperativeness (F3,128 = 3.91, p < .05; effect size, partial h 2 = .06) and passiveness scores (F2,128 = 5.45 p < .01; effect size, partial h 2 = .08) of the 3 groups, independently of the infants’ age (Figs. 3 and 4). Planned contrasts showed that MBU infants were significantly more cooperative (p < .01) and less passive (p < .001) than the infants of the community ill group, and did not differ significantly from the infants of the healthy group.

4. Discussion

Our main aim was to examine the mothereinfant interaction of dyads in a mother and baby psychiatric in-patient unit and to see whether the quality of interaction changed from admission to discharge, following a video feedback intervention. A secondary aim was to compare the mothereinfant interaction of these dyads with that of (1) a community ill group of mothers with a diagnosis of depression and personality disorder who had not received the video feedback intervention and (2) a healthy group of mothers with no diagnosed mental disorder.

Over the course of treatment, MBU mothers and infants improved so that when they were discharged from the MBU the mothers were more sensitive and less unresponsive than on admission, and their infants were more cooperative and less passive. On average the mothers received one video-feedback session every 3 weeks. We did not find any relationship between the weekly dosage of the video-feedback intervention and the improvement in the mother or infant interaction scores over the period of admission. This is primarily explained by the high association between the number of video feedback sessions and the duration of the mother’s stay on the MBU which in turn reflects the severity of her illness and the time needed for her to recover. However the efficacy of our treatment package on mother-infant interaction is consistent with a meta-analysis of intervention studies, which concludes that video feedback is the most successful technique in enhancing maternal sensitivity, and that a small number of sessions (less than 5) is optimal (BakermansKranenburg et al., 2003). We were also able to show that the improvement in maternal and infant scores was related within each individual dyad, so that the more sensitive a mother became, the more cooperative her infant became. This supports previous findings that, although the mother and infant ratings are independent, there are strong correlations between the sensitive and cooperative, and unresponsive and passive dimensions (Crittenden, 1998).

Our findings also showed that mothers with diagnoses of schizophrenia, mania or depression all made significant improvements in their interactions with their infants, as did the infants themselves. Mothers with schizophrenia have often been shown to have greater interaction deficits than those with affective disorders (Riordan et al., 1999). However, our finding suggests that, given appropriate support through video feedback intervention, mothers with schizophrenia improve in their interactive skills with their infants as much as mothers with other mental health diagnoses. Comparisons with a community ill group of mothers and their infants showed that, following the video feedback intervention, MBU mothereinfant dyads were significantly more attuned to one another than the ill mothers and their babies. Furthermore, comparisons with a healthy reference group showed improvement to an average level of interaction. This supports the importance of early relationship intervention for mothers with mental health problems and their infants. Such early interventions are likely to be significant in preventing psychological and psychiatric problems in the developing child and therefore need further investment in terms of research and resources.

5. Strengths and limitations

To our knowledge this paper is the first to look at mother-infant interaction in MBU dyads, and to show improvements following a video feedback intervention during their admission to an in-patient psychiatric mother and baby unit. A further novel feature is that comparisons were made with the interactions of mother-infant dyads in a community ill group and in a healthy group, drawn from the same geographical area as the MBU group. The CARE-Index is a dyadic assessment of mother-infant interaction, meaning that it assesses the fit between adult and infant. This index is particularly appropriate in evaluating an intervention as the assessment varies according to the mother-infant interaction on the day, without any practice effect over time, and, therefore, can be used multiple times (Cramer et al., 2006). Furthermore, coding of the CARE-Index was conducted by trained assessors who were unaware of the mental health status of the mothers. Two limitations of the study need to be discussed. First, this is not a randomised controlled trial, and therefore the improvements in the mother-infant interactions may be attributable to the multidisciplinary therapeutic environment of the MBU, including medication and psychological therapies. However, the review by Murray et al. (2010) suggests that, while treatments of maternal depression following childbirth have been shown to be successful in ameliorating the mother’s mental health, they have been less successful in improving mother infant interactions. The review concludes that the most successful interventions are those which focus on the mother-infant interaction, and this intervention aimed to do this. Therefore, we believe that it is unlikely that such dramatic changes in mother-infant interaction could be driven only by general therapeutic intervention and overall improvement in psychopathology. Nevertheless, it is important to stress the limitation of the study in disentangling the direct effect of the video
feedback intervention from that of other concurrent treatments. In order to do this, further research would need to compare mothers and babies receiving the same MBU in-patient treatment, some participating in the video feedback intervention and others not. However, there are ethical dilemmas involved in randomising women in one MBU to different treatment options in this way, because of depriving one group of a potentially effective intervention within the limited time-window where it can exert its longlasting consequences. Alternatively, one could compare different mother-infant interventions, although obviously this implies that the comparison treatment has also shown some evidence of efficacy before. Finally, comparing the outcomes for mothers and babies in different MBUs, where some offer the video feedback intervention and others do not, also has drawbacks, since the MBUs in the UK do not all offer the same multidisciplinary treatment options. Therefore, although the design remains a methodological limitation of this study, we also believe that it is challenging to address this question using alternative strategies. The second limitation is that the community ill and healthy comparison groups were not matched on demographics. In particular, the infants in the community ill groups were significantly younger than those in both the MBU group and the healthy group. This limitation was addressed by taking infant age into account in the analyses, and indeed the study findings remain robust. Notwithstanding these limitations, this study does demonstrate the effectiveness of an admission to an MBU with a unique mothereinfant relationshipfocussed intervention and shows significant improvements in the dyadic interaction by the end of the admission. This study thus goes some way to addressing the concerns of others (Forman et al., 2007; Murray et al., 2010) that one focus of treatment for mothers suffering from a mental illness should include the relationship with their baby in order to maximise the benefit for both.

Finally, in the current economic climate enormous emphasis is placed on measuring the outcomes of different treatments. Outcome measures following admission to in-patient psychiatric mother and baby units, such as the Health of the Nation Outcome Scales (HoNOS: Wing et al., 1996) and the Mental Health Cluster Tool (MHCT) have largely focussed on aspects of the mothers’ mental health and wellbeing. These may not be enough. In specific patient groups, such as mothers admitted with their infants to a psychiatric mother and baby unit, where the relationship of the dyad is of paramount importance, other measures such as the CARE-Index described here, may be helpful in demonstrating more specialist outcomes.

Scroll to Top