Mental health-related quality of life in mothers of children with surgically repaired congenital heart disease: a 13-year longitudinal study

This prospective longitudinal cohort study was conducted at the University Children’s Hospital Zurich, Switzerland. Children with CHD who underwent CPB surgery between 2004 and 2009 and were assessed at six measurement timepoints: at discharge after the first CPB surgery and at the child’s age of 1, 4, 6, 10 years and in adolescence between 11 and 15 years of age (mean age = 13.7 years). At each measurement timepoint, children underwent a neurodevelopmental assessment and parents completed questionnaires on their children’s behavior and HRQoL, family functioning, and their own mental health and HRQoL (see [16, 17] for details).

For the current analysis, families were included if their child underwent CPB surgery before the one-year assessment (conducted when the child was between 1 year 0 months and 1 year 11 months of age). Further, families were included if their child had no genetic or chromosomal disorder because these children were not assessed after 6 years of age. Therefore, of the 300 families initially included in the prospective cohort study, 177 families were eligible for the current analysis. Of these 177 families, 125 mothers completed a questionnaire about their own m-HRQoL at one or more measurement timepoints and were included in this analysis. The detailed recruitment procedure is displayed in Fig. 1.

Analyses were conducted with two statistical software packages, MPlus (Version 8.4; Muthén & Muthén, 2017) and R [24]. Missing data were estimated with imputation by chained equation [25]. Single imputation (m = 1) with 5 iterations was conducted on a dataset including the following data: m-HRQoL, social support (F-SozU-K14) and all demographic data presented in Table 1. Results were validated with an independently imputed dataset with a different seed. All p-values below 0.05 were considered statistically significant.

M-HRQoL in mothers of children with CHD was compared at each measurement timepoint to the matched subset of the Swiss normative sample with a linear regression model that controlled for migration background, parental education, and age. The linear regression was additionally weighted with the weight coefficients reported by [20] by applying the svyglm function in R to match the distribution of sex, age and nationality of the general Swiss population.

Latent class growth analysis (LCGA) was conducted with data from the mothers of children with CHD using maximum likelihood estimation with MPlus to identify subgroups of mothers’ m-HRQoL trajectories over time. Unconditional models with increasing numbers of class solutions up to four classes were compared according to lowest BIC, entropy > 0.8, and significant Lo–Mendell–Rubin and bootstrap likelihood ratio tests (p-value < 0.05) [26]. Only class solutions with classes that included at least 20 subjects were considered to maintain power for further analyses. For the LCGA, we standardized the SF-12 mental summary score with the Swiss normative sample. To do this, z-scores were calculated from the means and standard deviations of two age groups of the Swiss normative sample. Age groups were (1) from 19 to 45 years of age (n = 108) and (2) from 46 to 58 years of age (n = 144). Forming these age groups enables us to account for higher self-reported m-HRQoL in women > 46 years than in younger women in the Swiss normative population (see [20]).

To analyze predictors of low m-HRQoL trajectories, logistic regression analyses were conducted in R. The subgroups identified by LCGA were included as the dependent variable, and the following factors were included as predictors: social support, parental education, migration background, univentricular CHD, and child’s IQ. Because social support was assessed at each measurement timepoint, LCGA was conducted to identify two groups of good or poor social support over time using the same procedure described above (see details in Supplementary Tables 1and 2). Thus, social support was included in the model as a binary variable. Odds ratio (OR) and McFadden R² were reported as effect sizes for predictors and model fit, respectively. A McFadden R² between 0.2 and 0.4 indicates a good model fit [27].

A total of 125 mothers of children with CHD were included in these analyses. Sociodemographic characteristics of the mothers and clinical characteristics of their children with CHD are reported in Table 1. CHD types are presented in Supplementary Table 3. Mothers who completed at least one questionnaire about their own m-HRQoL had significantly higher education than mothers who did not complete (Mann–Whitney–U test: W = 1603, p < 0.001). The children’s CHD complexity (i.e., univentricular vs. biventricular CHD) did not differ between mothers who completed questionnaires and those who did not complete (X²(1) = 0.039, p = 0.843). The median time difference between the first CPB surgery and the one-year assessment was 10.9 months.

Comparison with the normative sample revealed significantly lower m-HRQoL in mothers of children with CHD at discharge. Some 42% had clinically relevant low m-HRQoL (≥ 1SD below normative mean). Subsequently, m-HRQoL improved. By the time the children were 1 year old, there was no significant difference between m-HRQoL in mothers of children with CHD and the Swiss normative sample. M-HRQoL in mothers of children with CHD was significantly higher than Swiss norms when the children were 4 years and 6 years old. There was no difference by the time the children had reached 10 years and 13 years. For all models, higher parental education was significantly associated with better m-HRQoL at each measurement timepoint, whereas age and migration background were not significantly associated with m-HRQoL. See statistical estimates in Table 2.

The fit indices for the class solutions of the LCGA are reported in Supplementary Table 4. LCGA confirmed the best model fit for the two-class solution separating m-HRQoL scores of mothers of children with CHD across time. This two-class solution was supported by significant likelihood ratio tests and moderate to good entropy. BIC was comparable to the three-class solution, and class sizes were adequate. The three-class solution was rejected due to an insignificant Lo-Mendell-Rubin likelihood ratio test and too small class sizes. For the four-class solution, the best likelihood value could not be replicated.

The patterns observed in the two-class solution are meaningful and consistent with a priori hypotheses. The first class was labelled “normal m-HRQoL” (n = 94, 75%) and is characterized by average m-HRQoL (mean z-score between −0.76 and 0.62) with no significant linear change over time (slope: β (SE) = 0.00 (0.01), p = 0.722). The second class was labelled “low m-HRQoL” (n = 31, 25%) and is characterized by m-HRQoL below average (mean z-score between −1.32 and −0.10) with a significant linear decrease over time with a small effect size (slope: β (SE) = −0.03 (0.02) per year, p = 0.040). See means and standard deviations for each class (class 1 = Normal m-HRQoL, class 2 = Low m-HRQoL) and measurement timepoint in Table 3. Average trajectories per class and individual trajectories are displayed in Fig. 2.

Logistic regression was conducted to investigate predictors for classes. The low m-HRQoL class was significantly associated with receiving low social support over time. There was no significant association between class assignment and parental education, migration background, number of CPB surgeries, having a child with an univentricular CHD, or low IQ – also when social support was excluded from the model (data not presented). Model fit was low (McFadden R² = 0.095). Statistical estimates are presented in Table 4.

We found mothers of children with CHD to be at risk for low m-HRQoL when their child undergoes open-heart surgery. This is in line with previous studies [10‐13]. During the period of hospitalization and surgery, parents face several psychological strains, including the fear of losing their children, bonding issues, simultaneous care for other family members, and financial burdens [3, 4]. Beside low m-HRQoL, previous studies also reported symptoms of posttraumatic stress, anxiety, and depression during the child’s hospitalization and shortly after [6]. Another study showed that 16% of mothers and 13% of fathers assessed with a self-rated screening questionnaire met diagnostic criteria for a posttraumatic stress disorder after the child’s first open-heart surgery [28].

In our study, low-maternal m-HRQoL at discharge improved to a normal or even above-normal range at the child's age of 1 to 13 years. However, although an overall improvement in maternal m-HRQoL occurred, m-HRQoL varied substantially over time. Studies with short-term follow-ups have reported mixed findings. Longitudinal studies found that initially low parental HRQoL improved to normal levels 6 months after discharge, and increased anxiety during hospitalization eventually improved to normal levels > 12 months after discharge [10, 11]. In contrast, other studies have found persistently lower well-being and increased depressive and anxiety symptoms during early childhood [14, 15, 29]. A cross-sectional study including parents of children with CHD attending preschool showed substantial variation in parental psychological distress, with 27% of mothers reporting high levels of stress and 19% reporting defensive low levels of stress [30].

Improvement of parental well-being after a traumatic and stressful event, such as having a hospitalized child undergoing open-heart surgery, could result from a range of psychological processes. First, parents could develop and adopt coping strategies to adjust to the cardiac condition of their child and associated stressors. Indeed, Demianczyk et al. used semi-structured interviews to identify frequently used coping strategies in parents of children with CHD. Coping strategies include amongst others (1) self-care to help reduce stress, termed active coping, (2) seeking emotional support among their social network, (3) accepting the situation and trying to find peace, and (4) being spiritual and having faith. In Demianczyk et al.’s study, parents also reported coping strategies that are unique to having a critically ill child, such as self-education, trust in the medical team, connecting with other parents of children with CHD, ensuring presence at the bedside, and spending time away from the bedside for self-care [31]. Another reason for improved m-HRQoL over time could be that parents may develop “posttraumatic growth”: positive personality changes after a life-changing event [32]. Indeed, a study recently demonstrated that parents of children with CHD report moderate to strong posttraumatic growth arising from their experience with CHD [33]. However, associations between posttraumatic growth and improved well-being have not been investigated and therefore warrant further research. Lastly, a “response shift” could be another psychological process explaining improved parental well-being. A response shift indicates that someone has adapted their internal standards and values of well-being after experiencing a traumatic event; the phenomenon has previously been introduced as accompanying improved HRQoL in parents of children with CHD after open heart surgery [10].

Interestingly, in our study, mean m-HRQoL in mothers of children with CHD was even higher than the Swiss norms when the child was in preschool (i.e., 4 and 6 years of age). One explanation for such a “honeymoon phase” could be that in this phase, most children are in good cardiac condition and the burden of potential neurodevelopmental problems may only manifest later at school age, when cognitive and social demands increase. Therefore, parents might feel relieved to return to a normal life and therefore feel exceptionally well during this period in comparison to the stressful time during their children’s hospitalization and open-heart surgery. Again, adaptive coping strategies, posttraumatic growth, and response shift may foster positive well-being during this phase. However, this hypothesis has not been tested, and future qualitative studies may usefully explore age-related changes in the subjective well-being of parents of children with CHD.

We found substantial variation in m-HRQoL among mothers of children with CHD. Indeed, LCGA identified two qualitatively distinct subgroups of m-HRQoL trajectories: 75% of mothers belonged to a normal m-HRQoL subgroup, and 25% of mothers belonged to a low m-HRQoL subgroup. In the low m-HRQoL subgroup, m-HRQoL significantly decreased over time with a relatively small effect size. This finding underlines the need to identify those at risk for low and decreasing m-HRQoL to provide early psychosocial support. We found that persistent low m-HRQoL was associated with sparse social support. This is in line with previous studies that identified social support as one of the most consistent and strongest predictors of parental well-being and resilience [34]. Another study found that a combination of family-related predictors, which include dyadic adjustment, social support, parenting stress, and posttraumatic stress symptoms, explained approximately 75% of the variance in QoL outcome 4 months after discharge [35]. A systematic review summarized how the family factors of family strain, parental perceptions, and coping were stronger predictors of parental well-being than the illness severity and physical limitations of the child [8]. In agreement with their finding, we did not observe an association between low m-HRQoL over time and CHD severity measured as number of surgeries and univentricular CHD. Lotto and colleagues suggest that the parental perception of CHD severity differs from the CHD severity defined by the cardiologists. They found poor agreement between parental risk estimation and objective measures of surgical risk [36]. Indeed, a parent’s subjective appraisal of their child’s health may affect their well-being more strongly than the actual clinical risk. Future studies should consider this difference when evaluating predictors of parental well-being.

25% of mothers of children with CHD reported chronically low m-HRQoL. Further studies are needed to evaluate how these mothers can be supported and investigate the impact of family-based interventions on their long-term well-being. The importance of parental well-being has been underlined by research showing that low maternal mental health is associated with neurodevelopmental impairments in children with CHD [37‐39]. Few studies have investigated interventional approaches to supporting mental health in parents of children with CHD. A systematic review demonstrated that parent interventions during the children’s stay on the ICU may improve maternal coping, mother–infant attachment, parenting confidence, satisfaction with clinical care, and infant mental health (see systematic review by [40]). Another study demonstrated that skin-to-skin care during the stay on the ICU can mitigate the negative effects of parental role alteration by reducing maternal physiological and psychological stress responses. This was observed in reduced salivary cortisol levels and anxiety symptoms [41]. Two randomized controlled trials showed that educational intervention programs after hospitalization improved maternal QoL and anxiety symptoms [42, 43]. These studies demonstrate promising interventional approaches to promoting parental mental health during and after hospitalization. Nevertheless, health policies that prioritize parental mental health care for parents with critically ill children remain lacking [40]. Further, reasons why parents do not proactively seek mental health support can be manifold and include (a) the belief that psychological support is outside the care team’s scope of practice, (b) the inclination to “stay strong” and fear negative judgment, and (c) the desire to focus care resources on the child [44]. Thus, substantial effort and standardized screening procedures may be required by the clinical care team to facilitate mental health support for parents of children with CHD.

Our study has shown that many mothers report critically poor m-HRQoL at the time of hospitalization. Although some of these mothers recover and report normal m-HRQoL throughout their child’s development, others continue to have low m-HRQoL. We have shown that CHD complexity may not adequately predict the trajectory of maternal m-HRQoL, but perceived social support can. Screening for psychosocial resources during hospitalization may be suitable for early identification of mothers at risk for persistent low m-HRQoL. Future studies are needed to investigate screening tools for identifying mothers at risk to target psychological interventions on those who need them most.