Study design
The present study is part of a longitudinal study that had a non-randomised cluster-controlled design. Participants completed 6 measurement points (T) as follows: during gestational weeks 16-22 (T1), 24-30 (T2) and 31(T3), and at 6 weeks (T4), 4 months (T5) and 6 months (T6) after birth. However, after study commencement the interval for T1 was extended to increase recruitment, and the intervals for the other measurement points were extended due to delayed responses from participants. This article is based on data from T1, T4 and T5 (see Table 1 for study design). Participants completed T1 between gestational weeks 13 and 39 (median 23.0, mean 23.0, SD 3.62), T4 between 5 and 15 weeks after birth (median 7.6, mean 8.1, SD 1.94), and T5 between 3 and 9 months after birth (median 4.0, mean 4.4, SD 0.83). The time between completing T4 and T5 ranged between 5 and 28 weeks (median 13.0, mean 13.1, SD 3.62). The study design and procedure have been described in more detail earlier (45).
Participants were allocated to the NBO intervention group or to care as usual based on their home address, which determined at which of six well-baby clinics they would receive their postpartum follow-up. Cluster randomisation of the well-baby clinics to learning the NBO was not feasible in this routine practice setting. Families belonging to one specific well-baby clinic received follow-up with the NBO plus care as usual. The NBO intervention was extended to three clinics during the study to increase the size of the NBO-group. Families at the remaining well-baby clinics received care as usual.
Participants and procedure
All Norwegian-speaking pregnant women and their partners from Tromsø municipality in Northern-Norway were eligible for inclusion. The recruitment period was between October 2015 and December 2017. Pregnant women and partners attending the antenatal clinic were recruited by midwives who gave information about the study. Potential participants agreeing to be contacted were later telephoned by a member of the research team for more information about the study and to plan a meeting for inclusion. The final sample recruited was 220 women (approximately 12 % of pregnant women in the region) and 130 of their partners. All participants gave written informed consent. Data was collected by means of online questionnaires answered during a meeting with a member of the research team (T1), or from home (T4, T5).
The intervention
The NBO consists of 18 neurobehavioural observations focusing on the infant’s behavioural repertoire within the attentional-interactional, autonomic, motor and organisation of states domains (5). This includes observations of responsivity to visual and auditory stimulation, capacity for habituation or sleep protection, amount of crying and ease of consoling, stress responses, reflexes, muscle tone and motor activity. Based on the observations, care giving strategies such as handling, sleep protection, comforting and regulation of social interaction are discussed (41). The NBO takes 15 to 40 minutes to administer and can be used from birth until the infant is three months old. The observations are not performed as a checklist, but tailored to the needs of the individual family and the awake and sleep states of the infant (5). Parents are encouraged to participate actively in the observation of their infant and to share their experiences, and the clinician meets them with a non-didactic and non-judgemental attitude. The overall aim is to provide tailored information and supervision related to parenting strategies based on the individual infant’s signals.
In this study, the NBO-group received three NBOs as an additional component to care as usual: 1. At the maternity ward with a midwife within 2 days post-delivery, 2. At the routine home visit with a public health nurse at 7-10 days post-delivery, and 3. NBO consultation at the well-baby clinic at 4 weeks post-delivery (additional to usual care visits). The comparison group received care as usual at the maternity ward, a home visit with a public health nurse at 7-10 days post-delivery, and had their first meeting at the well-baby clinic at 6 weeks post-delivery. Care as usual also included guidance on topics such as feeding, early social interaction, sleeping patterns, motor development, safe environment, crying, handling and caring for the baby, and the parents’ life situation and mental health (47). In addition, the baby’s weight gain was evaluated. The NBO was integrated as part of the public health nurses’ regular practice. However, a distinction is that in the NBO the guidance is given as part of the observation of the baby and tailored to the unique baby’s state and behavioural communication cues, whereas care as usual may include more general guidance delivered as part of a conversation with the parents. The NBO was administered by certified midwives and public health nurses. They were instructed to keep logs after each NBO session to register the date of the NBO, who were present and which observation elements were performed.
Research measures
Demographic information was collected at T1 and included questions about mental health history, physical health, education, work status before pregnancy, gross annual household income, marital status, number of previous children, social support from family and friends, and whether the pregnancy was wanted. The following four self-report questionnaires have been included for a description of sample characteristics and pre-intervention group differences. Pregnancy related anxiety (fear of giving birth, concerns about one’s appearance related to pregnancy, and fear of bearing a handicapped child ) was measured with the 10-item Pregnancy-Related Anxiety Questionnaire-Revised (PRAQ-R; 48) at T1. Emotional, physical and sexual abuse, and household dysfunction during the parent’s own childhood and adolescence was measured with the 10-item questionnaire Adverse Childhood Experiences (ACE; 49) at T1. Depressive symptoms during the last two weeks was assessed with the 21-item Beck Depression Inventory-II (BDI-II; 50) at T1. The mother’s bonding towards her baby during pregnancy was measured with the 19-item Maternal Antenatal Attachment Scale (MAAS; 51) at T3 (between gestational weeks 31 and 41, median 34.0, mean 34.4, SD 2.23). For more information about these instruments see the study protocol (45). See Table 1 for study design and a short description of the outcome measures.
Parenting stress was assessed with the Parenting Stress Index (PSI; 54). The PSI is a self-report inventory designed to assess stress in the parenting role, in the relationship between the parent and child and related to the perception of the child. The instrument consists of two domains: Parent domain (PD) and Child domain (CD). The PD was measured at T4 and T5, whereas the CD was measured only at T5. The PSI-PD consists of 54 items, divided into seven subdomains: competence, parent-infant bonding/attachment, isolation (e.g., social isolation and lack of social support), health (e.g., parental physical health), spouse (e.g., support from spouse), depression and role restriction. The PSI-CD consists of 47 items, divided into six subdomains: distractibility/hyperactivity, adaptability (e.g., the child’s ability to adapt to changes), parent reinforcement (e.g., experience of being liked by the child), demandingness, mood and acceptability (e.g., parental acceptance of the child). For the present study, we used the total scores from the PSI-PD and the PSI-CD. The Norwegian version of the PSI has been used in earlier research (55). In the present sample, PSI had excellent internal consistency (Table 2).
Mother-infant relationship measures. Maternal bonding to the infant was measured at T4 and T5 with the Maternal Postnatal Attachment Scale (MPAS; 14). The MPAS is a self-report inventory consisting of 19 items, each with 2 to 5 response options, e.g., from “Very incompetent and lacking in confidence” to “Very competent and confident”. All items have a minimum and maximum score of 1 and 5, respective, and some items are reversed. The MPAS measures the mother’s pleasure in interacting with her baby, the mother’s level of irritation towards the baby, and the quality of the maternal bonding, e.g., feeling proud of the baby. The 19 items yield a total score, with higher scores indicating healthier bonding. The current version of MPAS was translated to Norwegian by members of the research team, under the consultation of a professional translator. In the present study, MPAS had good internal consistency (Table 2).
Maternal reflective functioning was assessed with the Parental Reflective Functioning Questionnaire (PRFQ; 56) at T4 and T5. PRFQ consists of 18 items, with response options on a 7-point Likert scale from “strongly disagree” (1) to “strongly agree” (7). The 18 items are equally divided into 3 subscales. The subscales measure different aspects of parental reflective functioning and are analysed separately in previous studies and not summed up to a total score (56). The subscales are: Pre-Mentalizing (PM) modes (e.g., attributing negative intentions to the child and a lack of focus on the child’s inner life as a way of making sense of the child’s behaviour), Certainty about Mental States (CMS; parents’ ability to recognize that the child’s inner experiences are not always apparent), and Interest and Curiosity (IC) in mental states. Higher scores on all scales signals higher capacity for reflective functioning, whereas on the CMS subscale both high and low scores may be less optimal, indicating overconfidence or a too high degree of uncertainty in understanding the child’s states, respectively. The current version of the PRFQ was translated to Norwegian by A. Goksøyr and H. Braarud. In the present study, the PRFQ subscales had low to good internal consistency (Table 2).
Maternal confidence in parenting skills and the mother’s self-reported ability to perceive her child’s needs was assessed with the Maternal Confidence Questionnaire (MCQ; 57) at T4. The MCQ consists of 14 items, with response options on a 5-point Likert scale from “never” (1) to “very often” (5). Examples of items are: “I have all the skills needed to be a good parent” and “When my baby is cranky, I know the reason”. Higher scores on the scale indicate a higher sense of competence. The Norwegian version of MCQ was used in earlier research (58). In the present sample, MCQ had good internal consistency (Table 2). Satisfaction/benefit of the postpartum follow-up measure. The mothers’ experiences of the professional follow-up after birth were assessed at T4 using questions developed for the present study. The first domain which consisted of five questions was: “Through the follow-up you have received after birth from the maternity ward and the well-baby clinic, how much have you learned about the child's signals and needs in relation to:” (1) “the eating situation?”, (2) “Sleep/sleep patterns?”, (3) “Social interaction?”, (4) “Nappy change?”, and (5) “Crying/fuzziness?”. The second domain which consisted of four questions was: “In the follow-up you received after birth from the maternity ward and the well-baby clinic, how much did you feel you could: ”(1) “Share thoughts and concerns?”, (2)“Ask questions”, (3) “Get practical guidance?”, and (4) “Have trust in the health care worker?”. The last domain was: “Overall, to what extent do you feel that the follow-up has supported youand your family in a satisfactory manner?” Participants answered the questions on 5-point Likert scales from 1 (“Nothing”/“To a very small extent”) to 5 (“Very much”/“To a very large extent”). These questions are treated as single items and Cronbach`s alphas are therefore not reported.
Statistical analysis
All statistical analyses were carried out using IBM SPSS Statistics for Windows, Version 25.0 (59). Analysis of missing values revealed that all dependent variables contained missing data with a total missing value frequency of 22.1% and with 97 participants providing data for all measures at each meeting. The majority of participants (40 cases, 18.2%) had only one missing value, whereas 24 cases (10.9%) missed at least 80% of our dependent variables. By excluding these 24 participants, the final analysis sample consisted of 196 participants (NBO-group: N = 82; comparison group: N = 114), with all providing data for each questionnaire on at least one of the post-NBO meetings. In our final sample, 8 participants missed data collection at T4 (NBO-group: N = 6), while data from T5 was missing in 26 cases (NBO-group: N = 9). That is, 82.65% participated on both T4 and T5, even though some did not complete all scales or items. Moreover, data from the PSI was missing in most participants (PSI-PD: N = 52, PSI-CD: N = 46), but even here, the proportion of missing data from both groups were comparable (PSI-PD: NBO-group: N = 22 (26.8%), comparison group: N = 30 (26.3%); PSI-CD: NBO-group: N = 18 (21.9%), comparison group: N = 28 (24.5%)). Finally, data from the EPDS-T4, MCQ-T4 and PRFQ-T4 was collected in most cases (N = 184, NBO-group: N = 76 (92.6%), comparison group: N = 108 (94.7%)). We replaced missing data using the multiple imputation method with 50 iterations, by including data available from all scales and questionnaires as both predictors and predicted variables. Comparisons of participants missing the dependent variables (N = 24) with the final sample (N = 196) showed significant differences with regard to age, education, work status before pregnancy and annual household income. Missing participants were younger (p = .007, 29.0 vs 31.6 years), had a lower level of education (p < .001), lower income (p = .018) and were less often in full-time work before pregnancy and more often students (p = .003). There were no significant differences between the final sample and the excluded group on any other demographic or clinical variables.
By comparing basic demographic variables (e.g., age, well-baby clinic, level of education, gross annual household income, number of previous children, whether the current pregnancy was wanted, presence of any physical health problems, and previous occurrence of depressive symptoms, see Table 3 for all variables), a significant effect of group membership (comparison group vs. NBO) was found for the level of education only (Z = -2.19, p = .028), since the number of participants with 4 or more years spent in higher education was nearly twice in the comparison relative to the NBO group (74 vs. 41 participants, respectively). All other demographic parameters were comparable between the two groups (p’s > .08 for all). Similarly, potential group differences in any of our baseline measures (BDI-II at T1, PRAQ at T1 and T3, ACE at T1, MAAS at T3) collected prior to the intervention were assessed using independent-samples t-tests, revealing no significant effect of group (p’s > .19). Therefore, we added the level of education (3 levels: upper secondary school or lower, <4 years or ≥4 years in higher education) as a covariate for all statistical analyses to control for the potential contribution of education to the observed effects. Given the cluster-controlled design of our study (45), we also included well-baby clinic as another covariate, coded as a dummy variable.
To assess the differences between the NBO-group and the comparison group, the sample size for our study was based on a priori power analysis using multivariate analysis of variance (MANOVA) with an estimated effect size of f2=0.07, a power of 0.8 and an alpha level of .05 (45). This approach enables evaluating the joint effect of the intervention on psychometric scales sensitive to overlapping psychological constructs, accounting for possible covariations between them. We conducted separate MANCOVAs (Multivariate analysis of covariance) for testing the differences between the two groups on our two domains of interest: depressive symptoms/parental stress and mother-infant relationship, while controlling for the level of education and well-baby clinic. Separate analyses were performed at time points T4 and T5 (at week 6 and months 4 postpartum, respectively). We used scores from the EPDS, PSI-PD and PSI-CD for estimating maternal depression/stress (with PSI-CD being available at T5 only), whereas subscales from the PRFQ (PRFQ-PM, PRFQ-CMS, PRFQ-IC), the MPAS and the MCQ were used to assess the mother-infant relationship. Even though some variables showed signs of skewness and/or kurtosis (i.e., values > 1 or < -1) indicative of non-normal distributions, given the relatively large sample size of the current study and that assumptions of homogeneity of variances and covariance matrices were not violated (with the only exception of the PRFQ-PM score collected at T4 with a significant Levene’s test of F(1,193) = 7.45, p = .007), we decided to proceed with our original multivariate approach with reporting Pillai’s trace statistics (V) and calculating 95% bias-corrected and accelerated (BCa) bootstrapped confidence intervals for the contribution of NBO status to each outcome variable (60). For both domains of interest, separate analyses were performed at time points T4 and T5, with scores from the MCQ being available at T4 only. In order to investigate if the differences between the two groups were changing from T4 to T5, scores for the two domains were entered into repeated-measures ANCOVAs (Analysis of covariance) with Time (T4, T5) and Questionnaire (maternal depression/stress: EPDS, PSI; mother-infant relationship: PRFQ, MPAS) as within-subject variables, NBO-group as the between-subject variable, and Education and well-baby clinic as covariates.
Finally, responses to the questions regarding satisfaction with the follow-up from both groups were compared using Mann-Whitney U test. Potential variations in follow-up responses across the six well-baby clinics of intervention were assessed separately for the NBO and comparison groups with Kruskal-Wallis test. All statistical analyses were performed with an alpha value of .05, with Bonferroni correction applied for follow-up ANOVAs evaluating the contribution of each dependent variable to the joint effects revealed by the MANCOVA approach. Effect size is reported using Cohen’s f2.