One-year follow-up of The Incredible Years Parents and Babies Program: A pilot randomized controlled trial

The foundation of a healthy life begins in pregnancy and early adversity can have detrimental long-term consequences for affected children (1–4). This knowledge has led to increased investment in supporting early childhood development through interventions such as parenting programs (5).

Infants remain dependent upon their parents for many years after birth and are at risk of developmental harm if they suffer neglect or maltreatment during this time because brain plasticity is at its highest in the early years (6). In addition, infants are exposed to more potentially damaging experiences such as child neglect and violent abuse than older children (6–10). Compared to older children, infants are also more sensitive to disruptions in parental care. Infants with mothers suffering from depression or severe stress can show biochemical, physiological, and behavioral dysregulations beginning shortly after birth, and may be at increased risk of mental and behavioral problems (11–13). These infants often show high levels of negative emotion and distress and have a higher risk of developing an insecure attachment to their caregivers (14, 15).

Early onset of behavioral or emotional problems increases the risk of later adverse outcomes such as violence, delinquency, teen pregnancies, substance abuse, mental health problems, school dropout, and long-term unemployment (1, 3, 16–23). The nature of early parent-child interactions is a central predictive factor of many developmental outcomes; supporting parents in developing and applying positive parenting skills can stimulate healthy child development and prevent future problems (4, 5, 31–33, 23–30). The Incredible Years Parents and Babies (IYPB) program aims to support these child development goals.

The Incredible Years series (IY) offers prevention and treatment programs for parents, children, and teachers. The goal is to prevent and treat young children’s behavior problems and promote their social, emotional, and academic competence. The IY parenting programs have shown effect sizes ranging from d = 0.13 for preventive studies to d = 0.50 for treatment studies on parent-reported child conduct problems in a meta-analysis (34) and an overall effects size of d = 0.35 for parent-reported conduct problems in a meta-analysis using individual participant data pooled from 14 European IY trials (32). The effectiveness of the IY intervention is retained when transported to other countries (35) and is consistent across socioeconomic and ethnic background (36–38). The IY Toddler BASIC program for children 1–3 years old has shown positive results (39–42), however, we know less about the effectiveness of IY programs for the youngest children (43). Also, as many of the IY trials are wait-list designs we also know less about the long-term effects of the programs than the short-term outcomes (44).

The Incredible Years Parents and Babies program (IYPB) was developed in 2010 for families with infants from birth to one year. A pre-post evaluation of IYPB in Wales found that parenting competence and mental health significantly improved over time for those who participated in the program (45). A second evaluation conducted in Wales with a control group, found a significant positive effect on observed mother sensitivity, while no differences were found between the two groups on child development, parenting confidence, or parental mental well-being (46). None of these studies, however, included randomization. A trial of both the IYPB and the IY toddler programs (the E-SEE trial) is currently running in the UK, but no results have been published yet (47). The present study is the first randomized controlled trial (RCT) of the IYPB. The trial was conducted in two Danish municipalities from 2013 to 2015. The trial protocol and results at post-intervention are presented in previous papers (48, 49). The sample included 112 families who were randomized to IYPB (n = 76) or usual care (UC) (n = 36). The mothers who participated in the study were relatively representative of the general population, but had slightly higher levels of education than the general population. There were no significant differences in demographic characteristics (including education) between the IYPB group and the UC group at baseline. At post-intervention, when the children were approximately 5.5 months old, we found no differences between mothers who had received the IYPB intervention and those who received UC when examining parent-reported outcomes (N = 104). Moderator analyses, however, revealed that the 25% and 50% lowest functioning mothers who received the intervention reported higher parent stress, lower parenting confidence, and decreased mental health at post-intervention than the corresponding mothers in the control group; this result was statistically significant. In contrast, of the mothers in our sample who were the highest functioning at baseline, the mothers who received the IYPB intervention reported statistically significantly less parent stress than those in the control group (49).

In this paper, we examine the effects of the IYPB as a universal intervention on parent and child well-being and child development one year after the program ended and the children were 18 months old.

The treatment effect is measured in a linear regression framework with a lagged dependent measure:yi=γDi,treatment+μyipre+θXi+ɛi

The model consists of an individual treatment dummy (D_i,treatment), the individual baseline measure of the dependent variable yipre and a vector of individual covariates (X_i).

Categorical data are presented as numbers and percentages, and continuous data as means and standard deviations. The trial includes two sets of twins, both in the IYPB group. To account for the lack of independence between twins, we selected the first twin only for analyses of the parent outcomes, but kept both twins for child outcomes. We performed intention to treat analyses for primary and secondary outcomes with multiple regression analysis, including controls for site and baseline score.

Of the 112 mothers assessed at T1, 11 were lost due to attrition at T3. Attrition can lead to bias in the estimates, and to avoid this we chose to impute the missing observations. To deal with attrition, we first tested the assumption that data were missing completely at random (MCAR) by creating an indicator variable for observations missing at T3 and fitting it with a logit model. All baseline measures of the outcome variables and other covariates, along with the treatment dummy, were included as predictors. We found two significant predictors at the 5% level, indicating that the data is missing at random (MAR). Although the assumption of MAR data is not testable, imputation is still shown to produce less biased results than listwise deletion (69). Furthermore, should the data be MCAR instead of MAR, listwise deletion would lead to unbiased, albeit possibly inefficient, estimates. We generated 50 datasets with multiple imputations (70) using Stata’s multiple imputation (mi) impute procedure on chained equations. We truncated the imputed datasets to fit within the original scales.

As the PSS is the primary outcome we recalculated the PSS scores at T2 according to the two-subscale scoring recommended by Pontoppidan et al. (56) to examine if this changed any of the results at T2. We also used the two-subscale version to calculate PSS scores at T3.

Scores for the CDQ were marked as missing if more than four items were missing (out of a total of 40). For participants with 1-3 missing values, the missing values were replaced by the mean score of the item calculated by the remaining scores. As there were more missing total scores for the CDQ than for the rest of the measures we did not perform any further imputation for this measure and all analyses on the CDQ are based on the not imputed data.

We estimated the model with ordinary least squares (OLS) regression. A two-tailed test α = 0.05 was implemented for all analyses. We calculated effect sizes by dividing the adjusted mean difference between the trial arms by the pooled standard deviation. We used robust standard errors to account for group effects. We applied paired sample t-tests for testing total group change over time. As described in the protocol (48), we examined heterogeneous effects by comparing IYPB to UC in the following subsamples: (1) mothers scoring within the lowest 25% of the distribution at baseline, (2) mothers scoring within the lowest 50% of the distribution at baseline, (3) mothers scoring within the highest 50% of the distribution at baseline, and (4) mothers scoring within the highest 25% of the distribution at baseline. In the moderator analysis, we included an interaction term between a dummy for being in the 25% or 50% lowest/highest scoring group and a dummy for intervention allocation. Analyses were performed with Stata version 14.

Figure 1 presents the flow diagram of participants in the trial. Of the 125 families who had given their initial consent, 13 withdrew consent when the interviewer contacted them to schedule a home visit. Of the 112 randomized families, 76 were allocated to IYPB and 36 to UC. Eight families dropped out before T2 assessment, and a further three families dropped out between T2 and T3 assessment. The three families that dropped out between T2 and T3 all belonged to the UC group. We did not include further covariates in the analyses, as variables such as the mother’s age, education, and parity did not affect the results.

FIGURE 1.

We obtained data on parent participation from one of the two sites. On average, the participating mothers completed 6.3 sessions. Fathers participated less often than mothers (four sessions on average for the fathers who participated). The protocol describes Complier Average Causal Effects (CACE) analyses for two levels of participation: (1) mothers who participated in at least three of the eight sessions, and (2) mothers who participated in at least six of the eight sessions. However, as we could only obtain detailed participation data from one site, we could not perform these analyses.

Table 2 shows means and regression output comparing IYPB mothers and UC mothers at T3. The comparison of IYPB and UC mothers at long-term follow-up shows no significant difference in any of the outcomes.

We examine change over time for the full sample for the measures that were used at both T2 and T3 (PSS-PS, PSS-LPS, MDI, WHO5, and ASQ:SE-2e), to examine how the full sample developed after the intervention ended. There is a significant increase in the level of depressive symptoms (MDI p = 0.04), but no difference over time for parent stress (PSS-PS), parental satisfaction (PSS-LPS), mental health (WHO5), and child socio-emotional development (ASQ:SE-2e). Child socio-emotional development was measured at all three time points, but whereas the same version of the ASQ:SE-2e was used at T1 and T2, a different version with different items was used at T3 when the child was 18 months old. Therefore, the total scores of the two versions are not directly comparable. Because some of the measures were not strictly normally distributed we performed a sensitivity analysis by applying a non-parametric test (Wilcoxon signed-rank test) to the non-imputed data. Results are consistent with the t-test.

In this paper, we investigated the one-year follow-up effects of the IYPB program as a universal intervention for parents with infants aged 0-1 year. We did not find any statistically significant differences between the IYPB and the UC groups on any outcomes at long-term follow up one year after the intervention ended when the children were approximately 18 months old. This finding is consistent with the post-intervention results (49) in which we also found no significant differences for parent outcomes between the intervention and control groups, except for a single item. At post-intervention, mothers in the intervention group had smaller networks that helped them with practical matters than mothers in the control group. We do not find this result at the one-year follow-up, which supports the interpretation of the finding as a spurious effect. At one-year follow-up we also examined the mothers who scored in the lowest functioning 25^th percentile at pre-intervention using moderator analyses; we found that the lowest functioning mothers in the IYPB group reported significantly lower scores on the interest and curiosity subscale of the parent reflective function scale than the lowest functioning mothers who participated in the UC condition. At post-intervention, the lowest functioning mothers who participated in the IYPB group reported statistically significantly higher parent stress, lower parenting confidence, and worse mental health than control mothers. In contrast, the IYPB mothers who were in the highest functioning 25^th percentile reported significantly lower parent stress at post-intervention. These results were not maintained at one-year follow-up.

The lack of effectiveness of IYPB as a universal intervention in the present trial is in line with a recent systematic review of the effects of universal interventions for parents with infants (71), and other trials of universal parenting interventions (72–75). It also corresponds with the most recent research on the IY program where it has been stated that “children with more marked levels of conduct problems tend to benefit most from the intervention, suggesting it is more suitable for high-risk prevention or treatment, rather than universal prevention”(44). Both the short- and long-term results from our trial indicate that the IYPB intervention does not have effects on parent-reported outcomes either immediately following the intervention or one year after the end of the intervention when offered to a universal population in a country where the quality of usual care is high. It is important to note that the IYPB program was not developed as a universal intervention where there is already a high standard of care, but was intended to be used with disadvantaged families with observable difficulties in caring for their infants. In addition, the program was developed in the U.S. where there is a much lower standard of usual care. Our moderator analyses found that the lowest functioning mothers in the IYPB group had worse outcomes for some measures. We do not know the reason for these findings, but the IYPB program may be more useful for families in the context the developer intended with more homogenous groups in environments with a low standard of usual care.

As discussed in the previous paper (49) the mothers in the study have a relatively high education level and standard of living, we therefore would expect most of them to do well caring for their infants without any intervention. Most measures used in the study are developed for at-risk or clinical samples and there is a risk of ceiling or floor effects when applying them to a universal group (76). This can make it difficult to assess improvement over time. Whereas we found that almost all rating scales significantly improved from T1 to T2 when examining all mothers in the trial as a whole, only one out of five changed significantly over time from T2 when the children were approximately 5.5 months old, to T3 when they were about 18 months old. Overall, parental stress, parental satisfaction, parental mental health and child socio-emotional development did not change significantly from T2 to T3. Parental depression levels, however, were significantly higher at T3 compared to T2. This result may be a spurious result but may also be found because the majority of mothers were back to full-time work and most children were attending full-time daycare at the time of the follow-up when the children are 18 months old. This can put more pressure on the family as a whole.

When interpreting the results of this trial, it is important to take into consideration that the control group received a high level of usual care, which is a relatively extensive intervention provided to families in the Danish health care system. Although we do not see a marked improvement from T2 to T3 as we did from T1 to T2, both parents and children have good scores on most outcomes, indicating that the universal intervention that is offered to Danish families is beneficial, or that this population of parents adjust well to their new roles over time. Because of the high level of care provided to our control group, we are not able to conclude whether offering a universal IYPB program would be superior to receiving low-level care or no intervention.

Because the children in the study were a year older at T3 than at T2 the measures we used at T3 were not the same as those administered at T2 to correspond with the developmental stage the child. The analysis of differential effects is therefore not directly comparable from T2 to T3. At follow-up, we still find that most outcomes favor the IYPB group for the highest functioning half of the sample, whereas most outcomes favor the control group for the lowest functioning half. When conducting the moderator analyses, we find that for the interest and curiosity subscale of the parenting reflective functioning measure (PRFQ-1), IYPB mothers in the lowest-scoring quartile show lower interest and curiosity compared to control mothers. There is also a trend for intervention mothers in the highest-scoring quarter to show higher interest and curiosity than control mothers. This result, however, is not statistically significant [CI: -0,03;1,94] (Table 3). As we did 40 tests in all this may be a spurious effect.

Analyses at both post-intervention (T2) and follow-up (T3) indicate that the well-functioning mothers may benefit more from the IYPB intervention than the lowest functioning mothers when offered as a universal intervention. This may be due to the heterogeneity of the parents’ functioning in the IYPB groups when it is offered as a universal intervention. Parents may compare their skills and life circumstances to others of higher or lower functioning and this may affect their assessments and feelings about themselves and their families. It may be that lower functioning families offered IYPB would show more improvement if they did not attend a universally offered group but received the program in a more homogenous setting with other families who have similar needs and backgrounds. This approach may also allow the IYPB group leaders to focus on areas that are more relevant to specific groups.

This is a pilot trial based on relatively small sample size and the results should, therefore, be interpreted with prudence. Furthermore, analysis of the quality of the parent-child relationship at post-intervention has not yet been published. These outcomes are rated by objective observers and could differ from the parent-reported outcomes that are presented in this and the previous paper on short-term outcomes (49).

Limitations of the study include the relatively small sample size and that usual care in Denmark is fairly comprehensive. It is important to note that the IYPB intervention was developed for use with homogenous groups of lower functioning families rather than a universal population. Future research should investigate the effects of the program with disadvantaged families who show observable difficulties in caring for their infants.

When offered to a group of relatively well educated mothers who are already receiving a high level of usual care, we do not find any significant differences between the IYPB group and the intervention group. The lowest functioning 25^th percentile of mothers assigned to the IYPB group at pre-intervention reported significantly lower scores on the interest and curiosity subscale of the parent reflective function scale at one-year follow-up than control mothers, therefore further investigation into the cause of this result should be undertaken before being used with this population in universal groups. The IYPB intervention was developed for use with groups of lower functioning parents who do not have extensive access to other services, rather than a universal population with a high standard of care and it is important to be aware of this in clinical practice.

ClinicalTrials.gov NCT01931917 (registration date August 27, 2013)