Early intervention program reduces stress in parents of preterms during childhood, a randomized controlled trial

This study is a part of the Tromsø Intervention Study on Preterms (TISP); a randomized, controlled study of preterm infants with birth weight (BW) <2000 g, recruited between March 1999 and September 2002 (Rønning, Ulvund, Dahl & Kaaresen, 1998, unpublished research protocol). Preterm infants were randomized into blocks of six by using computer-generated numbers, to form an intervention group (PI, N =72) and a preterm control group (PC, N =74), and stratified according to gestational age (GA) <28 and GA ≥28 weeks. Healthy newborns (GA ≥37 weeks) were also recruited from the neonatal nursery to form a term reference group (TR, N =75). Written, informed consent was received from all participants before inclusion. Preterm controls (PC) followed the neonatal intensive care unit (NICU) guidelines for discharge of preterm infants, while term controls (TR) were routinely examined once by a pediatrician on their third day of life. Baseline data for each study group have previously been described in detail elsewhere[22, 23], and are shown in Table 1.

The intervention program was a modified version of the MITP[4] aimed at 1) enhancing parents’ understanding of their child’s expressions, and 2) promoting a sensitive, positive and practical transaction between parents and child. Eight nurses were trained to perform the intervention and each family was guided by the same nurse during all the sessions. Each intervention consisted of 7 hour-long sessions with parents and their baby during the last week before discharge, and 4 home visits at 1, 2, 4, and 12 weeks post-discharge[4]. The modification of the MITP included an initial session during which parents could vent their feelings about their preterm child. Mothers participated in all sessions while the fathers’ mean participation rate was 6.5 sessions (SD =3.4), which constituted 54% of the intervention program. In the first session the parents and the intervention nurse investigated the child’s capacities, focusing on the baby’s readiness and social communication abilities. During the following sessions, the parents were helped to recognize and be sensitive to behavioral cues, signs of disturbed regulation and stress in the child’s physiological, motor and state organization. The guidance was given while they observed the infant together, and all comments, questions and suggestions from the parents were appreciated. Finally, this understanding was applied to daily care by helping parents to make adjustments to their child’s strengths and vulnerabilities, in order to reduce stress levels and maximize the parents’ social engagement with their babies. During the four home visits, these topics were revisited and fine-tuned to individual needs, especially in connection with the child’s temperament, which was one of the main topics of the third home visit. The families had no other contact with the intervention nurses. All sessions were documented by logbooks written by the interventionists, and implementation according to the intervention manual[4] was ensured by logbook reviews carried out by the study director (JAR).

At the ages of 6 months, 1, 2, 3, 5 and 7 years, parents completed the Parenting Stress Index-Full Form (PSI-FF, third edition) while the Parenting Stress Index-Short Form (PSI-SF) was used when the children were 9 years old[6]. The PSI-FF consists of 120 questions covering three main dimensions of stress (child, parent and life stress) while the PSI-SF consists of 36 questions extracted from the parent- and child-related dimensions. A five-point Likert scale ranging from ‘strongly agree’ to ‘strongly disagree’ made up the response alternatives on both questionnaires. At 6 months, only one parent reported (mostly mothers) while mothers and fathers reported separately on all the subsequent occasions.

The PSI-FF consists of two main dimensions: Child Domain (47 items covering the subscales: Distractibility, Adaptability, Reinforces Parent, Demandingness, Mood and Acceptability), and Parent Domain (54 items covering the subscales: Perceived Competence, Isolation, Attachment, Health, Role Restriction, Depression and Relation to Spouse). A Total Stress (TS) score was also computed on the basis of all items except the life-stress questions. The PSI-SF is reported as a Total Stress score and by three subscales, each of which consists of 12 items: Parental Distress (PD), Parent-Child Dysfunctional Interaction (P-CDI) and Difficult Child (DC). Some questions in both questionnaires are used to calculate a Defensive Responding score, which indicates the degree of possible inconsistent/denial reporting from respondents.

Both PSI-FF and PSI-SF are frequently used in research[26, 27], and the correlation between Total Stress scores on these two measures is described as high (0.87)[6, 28]. The PSI-SF, DC subscale consists solely of items from the Child Domain in PSI-FF, and the Parental Distress subscale items from the Parent Domain. The P-CDI subscale includes items from both the Child and Parent Domains and focuses on the parental perception of transactions with their child and their expectations about the child’s behavior[6]. The Norwegian versions of both PSI-FF and PSI-SF were translated by Rønning and Abidin, and were used in this study with the permission of Abidin and Psychological Assessment Resources, Inc. (PAR). The questionnaires have some literal differences, in that questions in the PSI-SF may be perceived as more negative and definitive than those in the original PSI-FF format. The Life Stress (LS) questionnaire is part of the PSI-FF and was also used at age nine. The LS questionnaire consists of 22 items covering major life events in the family that are assumed to be challenging, even though they not are directly associated with child or parental challenges.

All the participating children received the same medical, developmental, and psychosocial assessments with recommendations about contacting other services if needed (age 6 months, 1, 2, 3, 5, 7, 9 years). Questionnaires were sent to the families approximately two weeks before each assessment. TISP was approved by the Regional Committee for Medical Ethics (2010/2153/REK nord) and the Norwegian Data Inspectorate on three occasions (in 1999, 2005, and 2010).

Because of repeated measures and the clustering effects of twin pairs, all longitudinal and cross-sectional analyses were performed by multilevel modeling (linear mixed models (LMM), SPSS statistics, version 20, SPSS Inc., Chicago, IL, USA). In the longitudinal analysis, time was treated as a continuous variable. In the cross-sectional analysis, predicted mean group differences with 95% confidence intervals (CI) were calculated. These analyses were also based on a longitudinal model, but in these cases time was treated as a categorical variable[29]. By varying the reference time point in the analysis, predicted group differences could be calculated. To assess agreement between parents, intraclass correlations (ICC) were computed, and the difference between the two independent intraclass correlation coefficients for the PI and the PC groups was tested as described by Alsawalmeh and Feldt[30]. The impact of variable intervention participation by fathers was analyzed by LMM and adjusted to take into account the clustering effects of twin pairs, and effect sizes in this case was given by Pearson correlations. Effect sizes (ES) created by the use of Hedges’ g are reported on predicted cross-sectional differences in mean scores between the PI and PC groups[31]. A P-value < .05 was considered significant. Randomization and inclusion criteria resulted in well-balanced study groups with one exception. Mothers in the PI group had an average of one more year of education at inclusion time (Table 1). The response rates were good throughout the study, still reaching 85% among mothers and 72% among fathers across all groups at age 9 (Figure 1).

No group by age interactions were uncovered on PSI, TS as reported by mothers or fathers from age 6 months until 7 years (Figure 2). Mean scores in all three groups were low compared to the American normative mean score (222 points) reported by Abidin[6].

In PSI-Child Domain a group by age interaction was reported by mothers from age 6 months until 7 years (F(5,642) =2.7, P = .02). While PI mothers reported child-related stress as being at its highest at 6 months and decreasing until age 7, PC mothers reported increasing levels from age one until 5 years (Figure 3).

The interaction concerning child-related stress in maternal reports may primarily be a result of two similar interactions in the subscales Adaptability (F(5,654) =3.3, P = .006) and Mood (F(5,663) =3.2, P = .007). All group by age interactions continued to be significant when controlled for maternal education (Table 1). No group by age interactions were reported either by mothers in PSI-Parent Domain or by fathers in either child- or parent-related stress.

Cross-sectional differences (at age 6 months, 1, 2, 3, 5, 7 and 9 years) in parental reports of child- and parent-related stress are reported first (stress reported at age 6 months, 1 and 2 years has been reported earlier[22, 23], but now predictions are based on a longitudinal model). Next, significant differences in different aspects of parenting stress (PSI, subscales) are reported. Lastly, agreement between parents in the PI and the PC group are compared.

Mother’s reports of child, parent and total stress are displayed in Table 2. Differences between preterm groups were mostly around 0.5 SD, and ESs were at their highest at 5 years in CD (ES =0.62). Mothers in the PI group reported less total stress than mothers in the PC group at every follow-up from age one until nine. Similarly, they reported significantly less child-related stress from age two until nine and less parent-related stress at ages two, three and five.

Fathers in the PI group reported significantly less child-related stress (PSI, CD) than fathers in the PC group at 2, 3 and 5 years and less total stress at age 5 (Table 3). Significant correlations were uncovered between reported stress and the number of interventions in which PI fathers had participated. Fathers who had participated less reported more stress at age 3 in: Total Stress (t(57) =2.2, P = .03, r = -0.32) and child-related stress (t(58) =3.0, P = .004, r = -0.37). A similar result was found at age 9; Total Stress (t(58) =2.5, P = .02, r = -0.33); Parent-Child Difficult Interaction (P-CDI) (t(52) =3.0, P = .01, r = -0.38) and DC (t(50) =2.44, P = .02, r = -0.32).

The subscales of child- and parent-related stress, in which significant differences between the PI and the PC group emerged, are displayed in Table 4. For all differences (whether significant or not) less stress was reported in the PI group. At 6 months, outcomes on one single subscale (Attachment) showed a significant difference between the preterm groups (t(278) =2.9, P = .004, ES =0.56). Fathers reported a similar difference on this subscale at age 1 (t(256) =2.8, P = .006, ES =0.55). More positive feelings were reported by the PI group at age 1 on the PSI subscale ‘Reinforces parent’ by both mothers (t(405) =2.3, P = .02, ES =0.46) and fathers (t(328) =2.0, P = .05, ES =0.44). Mothers in the PC group reported more stress related to lack of competence at age 1 (t(220) =2.3, P = .02, ES =0.47) and this difference between preterm groups persisted until age 7, with the largest effect reported at age 5 (ES =0.67).

A pattern of PI parents perceiving their child as being happier than did PC parents emerged in fathers’ reports at age 1 (t(364) =2.2, P = .03, ES =0.45) and in mothers’ reports at age 2 (t(443) =2.6, P = .01, ES =0.51). This difference continued to be reported by fathers until age 5 and by mothers from age 2 until 7 with increasing ES, reaching 0.60 at age 7. Mothers in the PI group also reported less distractibility/hyperactivity, better adaptability to everyday challenges and a higher acceptability, indicating that infants in the PI group matched their parents’ expectations in a more appropriate way than those in the PC group. At age 5, both parents reported these differences (Table 4). Lastly, a significant difference in the subscale Spouse (t(235) =2.0, P = .05, ES =0.37), as reported by fathers, emerged at age 5 between the preterm groups. Analyses of the questions in this subscale indicated that fathers in the PC group spent less time with their partners than those in the PI group.

The level of agreement between mothers and fathers were computed separately for the PI and the PC groups. Where significant differences in agreement between groups occurred, intraclass correlations in the PC and PI groups are reported and supplemented with P-values[29]. At age 2: Child Domain (ICC_PC =0.25; ICC_PI =0.69; P < .001), Parent Domain (ICC_PC =0.31; ICC_PI =0.64; P = .01), Total Stress (ICC_PC =0.24; ICC_PI =0.71; P < .001); age 3: Parent Domain (ICC_PC =0.26; ICC_PI =0.59; P = .01); age 7: Total Stress (ICC_PC =0.43; ICC_PC =0.65; P = .05) and age 9: DC (ICC_PC =0.36; ICC_PI =0.61; P = .04), Parent-Child Difficult Interaction (ICC_PC =0.20; ICC_PI =0.65; P < .001). Similar tendencies were reported on all other main outcomes except Child Domain at ages three and five, where agreement was at the same level.

Stress reported by the PI and PC groups was compared with reports from the TR group in separate longitudinal and cross-sectional analyses (Figure 2).

Group by age interactions were found in total and child-related stress as reported by mothers and in child-related stress as reported by fathers (Table 5). These three interactions are characterized by similar trajectories, as the TR group reported decreasing levels of stress from age one whereas the PC group reported stress at a higher and stable level across pre-school ages. Similar differences were found in several sub-dimensions (Table 5).

No differences were found in the longitudinal report of total, child- or parent-related stress. In the distractibility/hyperactivity and mood sub-scales, PI fathers reported lower means before school-age and higher means at age 7 compared with TR fathers (Table 5). No significant cross-sectional differences between the PI and the TR groups emerged in reports of parenting stress from age 6 months until 9 years.

A major strength of this study is the high participation rates that were maintained throughout the study period, reaching 85% among mothers and 72% among fathers across groups even at age nine. Although randomization generated a high degree of equality between preterm groups in aspects of birth, medical and socio-economic variables, PI mothers did have an average of one year more education than the PC mothers. Maternal education has previously been reported to be negatively correlated with parenting stress[44] but in the latest meta-analysis by Schappin et al.[8] maternal educational levels were not found to influence any aspect of parental stress. Nevertheless, all our analyses controlled for the difference of one year in mean maternal education. A limitation related to the construction of the study lies in the nature of self-reported questionnaires. Data collected by the PSI questionnaire are a result of parents’ subjective perception of stress on a specific day. An inclusion of biological parameters, such as the measurement of cortisol, may have safeguarded against faulty conclusions. Data may also be influenced by the way questions are asked in the two questionnaires. In the PSI-SF, questions are expressed more directly (more directly problem-orientated formulations), which may have amplified differences between respondents in their perceptions of greater or less stress.

We have previously reported interventional influences on the longitudinal trajectories and cross-sectional differences on children’s behavioral outcomes[25]. Parenting stress is known to be closely correlated with children’s behavioral development[45] and relationships between parenting stress and child behavior outcomes will be reported in papers to follow. This study demonstrates how an early child-centered and family-focused intervention may reduce parenting stress across childhood. This is a finding, not only concerning families taking care of prematurely born children but possibly also for other children and families at risk.