Introduction

Improvements in perinatal care, including technical practices in neonatal intensive care units (NICUs) and management of high-risk pregnancies, have led to an increase in the survival rate of very preterm infants [1]. However, preterm birth is associated with a high risk of long-term sequelae, such as neuromotor disabilities, cognitive impairment, emotional disturbance and behavioural problems [2–3]. Several risk factors have been identified in the genesis of these morbidities, including somatic factors like neonatal cerebral lesions, bronchopulmonary dysplasia, sepsis, and protein and energy undernutrition [4–7], and environmental factors like the family’s sociocultural level and the parental affective state [8–11].

Parental emotions and attitudes following a preterm birth may affect the quality of the early parent-infant relationship and possibly the child’s later competencies and development [12–15]. Indeed, prematurity is a risk situation that may well disrupt parent-infant interactions. Admission to the NICU means intensive technical care and a long stay in an incubator, which limits parental proximity to their newborn. These risks have been taken into account in the developmental care programs now implemented in the NICUs of many countries, giving a central place to parental competencies and well-being through a family-centred approach [16]. The physiological characteristics of preterm infants also pose a risk for interactions. During the neonatal period, preterm infants may display signs of weaker or shorter attentional behaviour, more fragile wakefulness, and less responsivity to caregivers than full-term infants [17]. In turn, the stress and traumatism of the preterm birth may alter the mother's sensitivity and availability, although these elements are crucial for the development of mother-infant bonding [18]. Alterations in the quality of the mother-infant interaction and maternal attachment representations have been reported, indicating a more mother-controlling dyadic pattern of interaction than after a term birth [13, 19, 20]. This pattern has been associated with infant’s behavioural symptoms, eating difficulties, and lower interest or abilities for social communication in the first two years of life [13]. Another potential consequence of a disorganized maternal caregiving system is social withdrawal behaviour, which is more frequent in children born preterm [21]. The early identification of maternal or parental interaction patterns that signal risk for infant development could prompt supportive intervention at a very early age: while the infant is still in the NICU and/or shortly after discharge to home.

Few observational-assessment tools are available to evaluate early mother-infant interactions, particularly in the context of premature birth [22]. Massie and Campbell developed the Attachment During Stress (ADS) scale to observe mother and child behaviours during paediatric examinations or other stressful situations [23]. The scale was particularly designed to detect insecure attachment behaviour. Studies have reported on its use in clinical paediatric settings as part of a Chilean attachment and intervention program [24] and in the investigation of mother-infant interactions with adoptive mothers or HIV-positive women [25]. This scale is suitable from birth to 18 months, but no study has yet established its feasibility in the context of a discharge examination of a very preterm infant.

EPIPAGE 2 is a cohort study of all preterm infants born between 22 and 31 completed weeks of gestation in 2011 in France, covering a period of 8 months [26]. The OLIMPE project (Observation LIen Mère-enfant PrématurE, observation of mother-premature infant bonding) provides a window for more detailed exploration of mother-infant interactions in a subsample of the French EPIPAGE 2 cohort. To this end, a multidisciplinary team of perinatal professionals was simultaneously trained to the same standards for observing interactions and assessing social withdrawal behaviour in two different periods: when the child was discharged from the NICU and at 6 months corrected age (chronological age plus prematurity).

The primary objective of this study was to determine the feasibility of using the ADS scale for very preterm neonates, both at the time of discharge to home and at 6 months corrected age. The secondary objectives were to (i) describe the mother-infant interactions and any disorders at discharge and 6 months, (ii) explore the association between maternal emotional state—including depression, anxiety, and stress—and these interactions, and (iii) assess the relationship between the disordered interactions and infant social withdrawal behaviour. An exploratory objective included the relationship between disordered interactions and family-centred care, evaluated by the centres’ policies.

Materials and methods

Results

Maternal and neonatal population

From the 12 participating NICUs, 332 neonates were eligible. Among them, 158 (48%) were not included because of (i) the unavailability of a trained professional to observe the dyad just before discharge in 149 cases, (ii) a lack of parental consent in 4 cases, and (iii) unknown reason in 5. One hundred and seventy-four parents agreed to participate in the study. The ADS scores were available for 163 mother-infant dyads at discharge, 148 at 6 months, and 136 at both times. Tables 1 and 2 present the maternal and neonatal characteristics of the eligible sample, the included sample, and the included sample still participating in the OLIMPE study at 6 months corrected age. The proportion of missing data on maternal and neonatal characteristics in the included sample was less than 10% in more than 90% of the variables. Compared with the non-included population, OLIMPE mothers were older, had been hospitalized less often during pregnancy, and had less in-utero transfer. Depressive symptoms and trait anxiety were also less frequent at discharge. Compared with the non-included population, OLIMPE neonates required a shorter duration of invasive mechanical ventilation and had lower rates of severe bronchopulmonary dysplasia and abnormal auditory screening results. They were more frequently supported with KC and an individual developmental care program and were also more often breastfed with direct nipple sucking.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Maternal characteristics in the eligible OLIMPE study population, the included population and the included population still participating at 6 months corrected age.

https://doi.org/10.1371/journal.pone.0188942.t001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Neonatal characteristics in the eligible OLIMPE study population, the included population and the included population still participating at 6 months corrected age.

https://doi.org/10.1371/journal.pone.0188942.t002

OLIMPE mothers who still participated in the study at 6 months corrected age had a lower risk of being probably depressed (as measured by the CES-D scale) and were less anxious (as measured by the STAI scale) than OLIMPE mothers who discontinued study participation.

ADS scale components

At discharge, 43.3% of the i-ADS scales and 15.3% of the m-ADS scales had at least one unobserved component. Two unobserved items were particularly frequent in infants: touching (a) and touching (b), respectively 78.3% and 25.0%. For mothers, the most frequently unobserved items were touching (b) and holding, respectively 41.0% and 14.7%. An association was found between the observation (or not) of the items touching (a) in i-ADS and touching (b) in m-ADS (p<0.001).

At 6 months corrected age, the rates of unobserved components on the infant and mother subscales had decreased to 13.9% and 12.7%, respectively. All items on both scales were observed in more than 90% of the cases (Table 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Frequency of unobserved items on ADS infant and mother subscales at discharge and 6 months.

https://doi.org/10.1371/journal.pone.0188942.t003

ADS scale interrater reliability

The ADS scale was independently assessed by two professionals in 59 dyads at discharge: 36.2% (59/163), and in 53 dyads at 6 months corrected age: 35.8% (53/148) (Table 4).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Interrater reliability of ADS infant and mother subscales at discharge and 6 months.

https://doi.org/10.1371/journal.pone.0188942.t004

At discharge, the percentages of agreement varied between 60.5% (touching (b)) and 77.6% (holding) for infants and between 58.6% (affect) and 78.0% (vocalizing) for mothers (unweighted percentages), while weighted Cohen’s kappa coefficients extended from 0.39 (affect) to 0.62 (proximity) for infant items, and from 0.43 (affect) to 0.76 (vocalizing) for maternal items.

At 6 months, the percentage of agreement and the weighted kappa coefficients for infant items were slightly higher compared with those at discharge and slightly lower for maternal items.

Interrater reliability for the presence or absence of an interaction disorder was determined for each subscale. Reliability was satisfactory for m-ADS at discharge [kappa: 0.66 (0.47–0.86)] and i-ADS at 6 months [kappa: 0.63 (0.38–0.88)], and moderate for i-ADS at discharge [kappa: 0.56 (0.34–0.78)] and m-ADS at 6 months [kappa: 0.53 (0.28–0.79)].

Mother-infant interaction disorders

Due to the high rate of unobserved components on "touching (a)" on the infant subscale and "touching (b)" on the mother subscale at discharge, these items were not taken into account in classifying the mother-infant interaction disorders at discharge.

At discharge.

At discharge (n = 163), interaction based on m-ADS was classified as disordered in 48.6 [40.9–56.4]%, including 49.8 [38.7–61.0]% resistant behaviour, 46.2 [35.3–57.5]% avoidant behaviour, and 3.9 [1.2–11.8]% combining resistant and avoidant behaviours. Based on i-ADS, interaction was classified as disordered in 36.5 [29.3–44.2]%, including 17.4 [9.5–29.8]% resistant behaviour, 72.2 [58.9–82.4]% avoidant behaviour, and 10.4 [4.6–21.8]% combined insecure behaviours.

Thus, 35.9 [28.8–43.7]% of the dyads had no disorder of interactions on either of the 2 sub-scales, 43.5 [36.0–51.4]% had an interaction disorder observed on one of the two subscales, and 20.6 [15.0–27.6]% an interaction disorder observed on the 2 sub-scales.

At six months.

At 6 months (n = 148), m-ADS suggested altered interaction in 32.6 [25.4–40.7]%, including 40.1 [26.8–55.0]% resistant behaviour, 49.2 [34.9–63.5]% avoidant behaviour and 10.7 [4.4–23.9]% combining resistant and avoidant behaviours. Altered interaction was observed in 26.0 [19.4–33.8]% of i-ADS, with 18.8 [8.9–35.4]% resistant, 73.1 [56.1–85.3]% avoidant and 8.0 [2.5–23.2]% combined insecure behaviours.

Thus, 57.8% [49.6–65.6]% of the dyads had no disorder of interactions on either of the 2 sub-scales, 25.8 [19.3–33.6]% had an interaction disorder observed on one of the two subscales, and 16.4 [11.2–23.4]% an interaction disorder observed on the 2 sub-scales.

Disordered interactions on i-ADS significantly decreased between discharge and 6 months (p<0.001, McNemar’s test for paired data).

At 6 months, infant and mother disordered interactions based on i-ADS and m-ADS were significantly associated (p<0.001). Interaction disorders based on i-ADS were also associated with the previous observation of interaction disorders at discharge on both subscales (Table 5).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Relationships between disordered interaction at discharge and 6 months based on ADS infant (i-ADS) and mother (m-ADS) subscales.

https://doi.org/10.1371/journal.pone.0188942.t005

Association between maternal emotional state and mother-infant interaction disorders

Table 6 shows the proportions of interaction disorders in mothers and children at discharge and 6 months corrected age, according to the maternal mental health evaluations (CES-D, STAI, and mPPQ scales). No association was found between maternal depression, anxiety, or stress and interaction disorders based on the infant and mother subscales of the ADS.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 6. Disordered interaction at discharge and 6 months: Relationships with maternal emotional state and infant social withdrawal behaviour.

https://doi.org/10.1371/journal.pone.0188942.t006

Discussion

This study showed that perinatal care professionals were able to assess the early interactions between mothers and their very preterm infants with acceptable reliability after a period of training. Insecure attachment behaviours at 6 months infant corrected age were associated with previous identification at discharge. The behaviours in the mothers were anticipated by the observation of disordered infant interactions, and in the infants by the non-optimal interactional patterns of both the mothers and their infants. Also, insecure infant attachment at 6 months corrected age was not related to maternal mental health but was strongly associated with infant social withdrawal behaviour.

To our knowledge, OLIMPE is the first large cohort of very premature neonates in which early mother-infant interactions were studied just before discharge from the NICU. We selected a scale that could be scored during or just after live observation and rapidly coded [22]. Although the ADS scale is compatible with self-guided training, two one-day meetings were organized to ensure standardized procedures in all centres, especially for item scoring, and a written protocol summarizing all instructions was distributed before the start of the study. Despite the containment provided during the examination to promote interactions, the infants’ immature motor skills were a limiting factor in the observation of "touching" items on the infant subscale at discharge, and thus the corresponding maternal reaction, "touching (b)". Although previous studies have not reported this restriction in scale use, it should be noted that they included only term infants of at least 3 months old, rather than very premature neonates examined near term [25, 36]. Theoretically, the CARE-Index also assesses the quality of interaction and dyadic characteristics from birth to 15 months [37]. However, its performance in the context of premature infants having less than 6 months corrected age has never been explored [13, 20]. In addition, this tool requires extensive training, systematic videotaping of the interactions, and experienced coders [22]. Censullo’s dyadic mutuality code (DMC) is another instrument that can be used in the first 6 months of life to specifically explore synchronicity in the dyad [38], but its sensitivity has been questioned in a context quite comparable to ours [39].

The literature indicates that different procedures have been followed to interpret ADS scoring. Hale et al. [25] averaged the mean scores of the seven items, using the subscale scores. We nevertheless found this scoring method to be inappropriate for detecting insecure behaviours, which are positioned at the extremes of the 5-point scale. We therefore based our rating on the occurrence of insecure behaviours determined independently from the infant (i-ADS) and mother (m-ADS) subscales, as did Carcamo et al. [36]. However, we did not use their classification which prevented the coding of a substantial number of scales. Thus, we considered that the observation of at least one typical behaviour of insecure attachment or two rather avoidant behaviours was enough to confirm a disordered interaction, which allowed us to settle all cases. The rate of nearly 30% for insecure behaviours on both the maternal and infant subscales at 6 months was fairly consistent with the controlling pattern of interaction observed in 28% of mother-infant dyads using the CARE-Index at the same age [25, 36]. Moreover, in a cohort of 117 children born before 32 weeks of gestation or below 1500 g, a recent study observed 12.8% avoidant and 23.1% resistant/ambivalent attachment at a corrected age of 2 years using the gold standard: the strange situation procedure [40].

Our observation that a disordered maternal interaction at discharge may anticipate a disordered infant interaction at 6 months was consistent with the transactional theory of development, which proposes that non-optimal interactional patterns exacerbate already disorganized infant behaviour [41]. Intervention programs based on this theory and begun during NICU hospitalization therefore may improve infant temperament, mother-infant interactions and parenting stress in the early postnatal years [42].

Most of the research on parental mental health following very preterm birth has focused on maternal depression, with rates of clinically significant depression ranging between 15% and 40% in the first postnatal year, compared with approximately 10% after term delivery [43]. Despite the documented adverse effects of postpartum depression on mother-infant interactions, we found no association between the CES-D and ADS scales. This result was not explained by maternal antidepressant treatment, which might have potentially decreased the impact of depression on these interactions, as mothers suffering from a psychiatric disorder, including severe depression requiring treatment, were excluded from the study. CES-D screens for women at risk for depression, but it cannot be used to diagnose major depression. Furthermore, the impact of early maternal depression on attachment security is statistically low, with the magnitude of effect extremely variable from one study to another [44]. Maternal anxiety has been associated with less touching, speaking and responsiveness to the premature infant during NICU hospitalization, which may decrease the infant's engagement in interaction [45]. However, the studies that have assessed these early interactions found that anxiety had inconsistent effects depending on its intensity, duration, co-occurrence with depression, and the tools used to measure it [12, 18, 39]. Attention is increasingly being focused on the symptoms of maternal posttraumatic stress, which is experienced by 35–45% of the mothers of premature infants at 6 months infant age, as noted in the OLIMPE study as in others [19, 20, 46]. Yet to date, very few studies have examined the influence of maternal traumatic reaction on the quality of mother-infant interaction at 6 months. Consistent with our results, Muller-Nix et al. found no significant difference in maternal and infant interactional behaviours [19]. However, maternal stress influenced the dyadic interactive pattern, with more "controlling mother-compliant infant" dyads [20].

A robust association was found at 6 months between disordered interaction on the ADS infant subscale and infant social withdrawal behaviour. In our opinion, this finding adds external validity to the results of the ADS scale, as the ADBB scale includes elements of observation in relationship to the observer rated outside the mother-infant interaction sequences. Indeed, withdrawal has been interpreted as a way for the infant to deal with a lack of synchrony in the parent-infant relationship, and the association between early mother-infant interaction and later social withdrawal has been suggested [47]. Our rate of withdrawal was nevertheless difficult to interpret, in part because the OLIMPE study is the first to provide such data in very preterm neonates. A study performed in our country found a prevalence of 14% at 12 months and a clear association with low birth weight and preterm birth [21].

The ADS scale results suggested that, for family-centred care, the centres’ policies regarding KC or BF did not seem to influence the occurrence of disordered interactions at discharge and at 6 months. Although policies supporting neurodevelopmental care have progressed in our country [27], few children of the EPIPAGE 2 and OLIMPE cohorts were enrolled in an individualized developmental care program with a family-centred approach, yet these programs may well increase parental self-confidence and competence and promote a positive interactional style.

Conclusions

This prospective multicentre observational cohort study confirmed the high prevalence of disordered mother-infant interactions and maternal distress in the few months following very preterm birth. We do not as yet have long-term outcomes for the OLIMPE infant study, but neurodevelopmental outcome will be assessed on the basis of the difficulties observed in the early interactions. Parents and the environment can modulate the effects of biological factors on child development. Our work suggests that the ADS scale can help to target those dyads that would most benefit from early hospital- and home-intervention programs designed to improve parenting behaviours, the quality of the home environment, and the outcomes for children and their families [50].

Acknowledgments

OLIMPE study writing group includes, in addition to the authors of the manuscript, the following members: Isabelle Souksi-Medioni, MD (Department of Neonatal Medicine, Carémeau University Hospital, Nîmes, France); Richard Barbier, MD (Neonatal Intensive Care Unit, Dax-Côte d'Argent General Hospital, Dax, France); Gael Mazeiras, MD (Neonatal Intensive Care Unit, Côte Basque Hospital, Bayonne, France); Isabelle Glorieux, MD (Department of Neonatal Medicine, Children's University Hospital, Toulouse, France); Olivier Brissaud, MD (Department of Neonatal Medicine, Pellegrin Children's University Hospital, Bordeaux, France); Anne-Laure Sutter, MD, PhD (Department of Adult Psychiatry, Charles Perrens University Hospital, Bordeaux, France); Bernard Guillois, MD (Department of Neonatal Medicine, Caen University Hospital, Caen, France); Sebastien Mur, MD (Department of Neonatal Medicine, Jeanne de Flandre University Hospital, Lille, France); Laurent Storme, MD, PhD (Department of Neonatal Medicine, Jeanne de Flandre University Hospital, Lille, France); Anne-Catherine Rolland, MD, PhD (Child and Adolescent Psychotherapy Unit, Robert Debré University Hospital, Reims, France), Claire Langlet, MD (Department of Neonatal Medicine, Hautepierre University Hospital, Strasbourg, France), Sylvie Nezelof, MD, PhD (Infant and Child Psychiatry Unit, Besançon University Hospital, Besançon, France), and Hubert Daudé, MD (Medico-Social Early Intervention Centre (CAMSP), St Eloi University Hospital, Montpellier, France).