Parenting stress in caregivers of children with chronic physical condition—A meta-analysis
Abstract
On the basis of the parenting stress model we compared levels of parenting stress in families with and without a child with a chronic physical condition and analysed correlates of parenting stress in families with a child with a chronic condition. A systematic search through electronic databases identified 547 relevant studies that were included in a random-effects meta-analysis. Parents of children with a chronic condition showed small to moderate elevations of general parenting stress and stress related to the parent–child relationship in particular. They showed moderate to large elevations in health-related parenting stress. Parents of children with cancer, cerebral palsy, HIV infection or AIDS, and spina bifida showed the highest levels of parenting stress. Stress levels also varied by illness severity and duration, child age, parental gender and mental health, marital status, marital quality, and levels of perceived support. Behaviour problems of the child and low parental mental health were the strongest correlates of parenting stress. The present results assist with identifying parents at highest needs for interventions aimed at reducing parenting stress. These interventions should address the reduction of child behaviour problems, the promotion of parental mental health, the increase in marital quality and social support in general, and skills for dealing with stressors.
1 INTRODUCTION
Parenting stress is the psychological distress experienced by parents while trying to meet parenting role demands (Abidin, 1995). It is distinct from other kinds of situational stress, and its effects can be measured separately from assessments of other stressful circumstances (Deater-Deckard, 1998). According to Abidin (1995), stress-evoking factors in the parenting role can be categorized into three domains. Stress in the child domain refers to the child's temperamental and behavioural characteristics, such as high irritability and difficulty with compliance. Stressors in the parent domain involve aspects of parental functioning and personality components, such as feelings of guilt, depression, and a sense of low competence in the parental role. Finally, stress in the parent–child relationship domain refers to negative interactions between parents and their children, such as conflicts and parental dissatisfaction with these interactions. Other authors distinguish between the negative impact of parenting on parental time use (time impact) and emotions (emotional impact; Landgraf, Abetz, & Ware, 1996). High levels of parenting stress have been related to adverse outcomes, such as depression and anxiety in children and their parents (Deater-Deckard, 1998; Fonseca, Nazaré, & Canavarro, 2011), child maladaptive behaviours (Cabrera & Mitchell, 2009; Semke, Garbacz, Kwon, Sheridan, & Woods, 2010), and impaired cognitive development (Grunau et al., 2009; Molfese et al., 2010). Increased parenting stress may also drive parents to use maladaptive parenting practices (Farmer & Lee, 2011). In the case of chronic conditions, parenting stress may interfere with the management of a child's illness (Celano, Klinnert, Holsey, & McQuaid, 2011; Streisand, Braniecki, Tercyak, & Kazak, 2001). Given the negative effects of parenting stress on a large number of parental and child outcomes and the fact that having a child with a chronic condition tends to increase levels of parenting stress (e.g., Golfenshtein, Srulovici, & Medoff-Cooper, 2015), it is important to identify the levels of parenting stress in families with a child with a chronic condition and to identify factors that may increase or decrease these stress levels. Thus, the goal of the present meta-analysis is to quantify the elevated levels of aspects of general and health-related parenting stress in families with a child with a chronic physical condition (physical illness, physical disability, or sensory disability) across a large number of studies and to identify predictors of elevated parenting stress in families with a child with a chronic condition.
Parents of children with a chronic physical condition may experience elevated levels of parenting stress. First, in addition to age-typical caregiver tasks, these parents face further caregiver tasks related to illness management, such as stressful communications with health care providers (Golfenshtein et al., 2015) or counteracting the child's nonadherent behaviour (Streisand et al., 2001). Second, uncertainty and lack of predictability of the course of the child's condition (Lee, Yoo, & Yoo, 2007), as well as concerns about a child's negative prognosis (Streisand et al., 2001), have been identified as sources of parental stress. Third, elevated levels of internalizing and/or externalizing behaviour problems that are often observed in children with chronic physical condition (Pinquart & Shen, 2011) may be another source of elevated levels of parenting stress. Fourth, illness- or disability-related dilution of familial resources (e.g., financial burden) and conflicts between caregiving tasks and other commitments (e.g., the work role) may impair perceptions of competence in the parenting role and the quality of the parent–child relationship (Golfenshtein et al., 2015). Fifth, in the case of hereditary conditions or conditions that result from behaviours of parents or children (e.g., traumatic brain injury), parental feelings of guilt may cause parenting stress (e.g., Hodgkinson & Lester, 2002). Nonetheless, many parents are able to adapt to having a child with a chronic condition (Visconti, Saudinio, Rappoport, Newburger, & Bellinger, 2002).
A narrative review by Golfenshtein et al. (2015) on families with children with congenital heart disease (CHD), cancer, and autism spectrum disorders indicates that parents across these conditions experience increased levels of parenting stress compared to healthy controls. Stress in the difficult child domain seemed to be most prevalent in parents of children with autism spectrum disorders and CHD, whereas parents of children with cancer and CHD often reported stressors related to dysfunctional parent–child interaction. On the basis of 13 studies related to arthritis, asthma, cancer, cystic fibrosis, diabetes, epilepsy, and sickle cell disease, a meta-analysis by Cousino and Hazen (2013) found that caregivers of children with a chronic physical condition reported significantly greater levels of general parenting stress than caregivers of healthy children. Nonetheless, between-group differences were, on average, small (d = 0.40), and the size of these differences varied between studies. The authors did not test whether elevated levels of parenting stress would be more common in parents of children with some conditions rather than in others. Despite the fact that different aspects of parenting stress can be distinguished (e.g., Abidin, 1995), the authors also did not compare elevations of parenting stress across these aspects. In addition, Cousino and Hazen (2013) did not meta-analyse factors associated with stress levels of the parents. Knowledge about these topics would be relevant for identifying subgroups of parents who are at highest need for interventions aimed at mitigating their parenting stress, for identifying aspects of parenting stress that should be the main target of intervention, and for selecting stressors and resources that should be addressed in these interventions. Finally, although—given the small number of published studies included in their meta-analysis—Cousino and Hazen (2013, p. 823) acknowledged that publication bias may have affected their results, these authors did not use available procedures for identifying and correcting the potential impact of this bias on their results, such as Egger's test and trim-and-fill analysis (Duval & Tweedie, 2000). These limitations are addressed in the present meta-analysis.
Thus, the first research question of the present meta-analysis asked whether parents of children with chronic physical condition would report higher levels of parenting stress than parents of other children. On the basis of previous reviews (Cousino & Hazen, 2013; Golfenshtein et al., 2015), we expected small to moderate elevations of general parenting stress (Cohen, 1992). The second research question asked whether elevations of parenting stress vary between different aspects of stress. Studies with general measures of parenting stress (not specific to stressors related to child health) were expected to show lower elevations in families with a child with a chronic condition than studies that ask explicitly for parenting stress due to child health. We also expected that the emotional impact of caring for a child with a chronic condition would be larger than the time impact as the latter will be reduced when other persons—such as the spouse—take some responsibility for the illness management.
The third research question asked whether relative levels of parenting stress would vary by the kind of condition and by aspects of the quality of the study. On the basis of sources of parenting stress in families with a child with a chronic condition (e.g., Golfenshtein et al., 2015; Lee et al., 2007; Streisand et al., 2001), we expected finding the highest levels of parenting stress if the children are likely to need help with daily activities due to restricted everyday competence (e.g., in the case of cerebral palsy and spina bifida), if the illness is associated with elevated risk of mortality (e.g., cancer, HIV infection), or if parents care for a child with elevated levels of behaviour problems, as is, amongst others, the case with regard to epilepsy (Pinquart & Shen, 2011).
The fourth research question addressed whether the effect sizes vary by indicators of study quality as low-quality studies can lead to a distortion of the effect size estimate (Lipsey & Wilson, 2001). Sum measures of study quality are not commendable because some criteria of low quality may lead to underestimations of effect sizes whereas others could lead to overestimations (e.g., Higgins & Green, 2009). Thus, we asked whether between-group differences in parenting stress vary by four individual criteria derived from the Modified Quality Index that has been previously used in a review on caring for children with a chronic condition (Ferro & Speechley, 2009). The criteria address external validity (whether the participants were representative of the entire population from which they were recruited), internal validity (valid and reliable assessment of parenting stress; whether parents of children with and without chronic condition did not differ in third variables or whether the analysis adequately adjusted for confounding effects of third variables), and statistical power. With regard to statistical power, we assessed whether the sample size was sufficient for detecting an effect of 0.40 that has been found in the previous meta-analysis by Cousino and Hazen (2013). As nonsignificant effects may be less likely to be published than significant effects (the file-drawer problem; Lipsey & Wilson, 2001), we were also interested in whether elevations of parenting stress would be lower in unpublished studies than in published studies.
2 CORRELATES OF PARENTING STRESS
The final research question focused on correlates of parenting stress because identifying predictors of parenting stress is relevant for designing stress-reducing interventions. In his model of influences on parenting stress, Abidin suggested six predictor variables, namely, (a) child characteristics, such as his or her demandingness; (b) parental role restrictions; (c) parental health; (d) parental attachment; (e) the relationship with the spouse; and (f) available social support (e.g., Abidin, 1995). Aspects of the child's chronic condition can be subsumed under the child characteristics and role restrictions when referring to elevated caregiving demands. Guided by Abidin's model, the final research question asked whether the levels of parenting stress of mothers and fathers of a child with a chronic condition vary by child characteristics, parental mental health, spousal relationship, and available social support. We did not analyse parental role restrictions as a predictor because these restrictions have been assessed as a central domain of parenting stress rather than as a predictor (Landgraf et al., 1996). We also did not include attachment and parental physical health due to widely lacking related studies.
2.1 Child characteristics
The meta-analysis addressed the severity and duration of the chronic condition, child age, and level of behaviour problems. As more severe conditions tend to be associated with more caregiving demands, we asked whether higher severity of the condition would be associated with higher parenting stress. Longer illness duration provides, on average, more time to adapt to the condition. Thus, we expected finding lower levels of parenting stress if the children had been diagnosed for many years. Older children tend to take more responsibility for their illness management (Naar-King et al., 2009), thus decreasing the demands on their parents. Thus, we expected lower levels of parenting stress in parents of older children. On the basis of Abidin's (1995) model and Cousino and Hazen (2013), we expected a positive association of parenting stress with behaviour problems of the child.
2.2 Parental characteristics
Abidin (1995) proposed that impaired parental mental health would increase levels of parenting stress because of difficulties with meeting the needs of the child and that high levels of parenting stress could impair the mental health of the caregiver. Cousino and Hazen (2013) identified negative associations of parenting stress and parental psychological health in families with a child with arthritis, cancer, cystic fibrosis, and diabetes. We tested whether this result can be generalized when including studies on other diseases.
Mothers are the primary caregiver of most children with a chronic condition and often take the main responsibility for illness management (Quittner & DeGirolamo, 1998). Thus, we analysed whether mothers experience higher levels of parenting stress than fathers.
2.3 Marital relationship
Married parents tend to share caregiving responsibilities, which could reduce the caregiving demands for the individual parent (Mullins et al., 2011). Thus, the present meta-analysis tested whether married parents of children with a chronic condition experience less parenting stress than single parents and whether this stress-reducing effect may be stronger in high-quality, supportive marriages.
2.4 Social support
On the basis of the Abidin (1995) model, we tested whether higher availability of social support is associated with less parenting stress.
3 METHODS
3.1 Sample
- They assessed parenting stress in families of children with a chronic physical condition (physical illness or sensory disability or physical disability).
- The studies provided sufficient information for a comparison of levels of parenting stress with established normative data or a similar group of families with healthy children or reported correlations of parenting stress in families with a child with a chronic condition with variables specified in Table 3.
- Mean age of children was <18 years.
- Studies were published or presented before June 2017.
The literature search was completed on June 23, 2017. In regard to the question of whether a condition is defined as chronic, we followed the suggestion by Thompson and Gustafson (1996), stating that a chronic illness can be defined as a condition that is associated with functional impairment and lasts for a considerable period of time, has a sequela that persists for a substantial period, persists for more than three months in a year, and/or necessitates a period of continuous hospitalization for more than a month.
In order to include studies from different regions of the world, we did not limit the included studies to those written in English. Unpublished studies (e.g., dissertations, master theses) were identified as part of the systematic search with the electronic databases PsycINFO, CINAHL, Google Scholar, and PSYNDEX as well as cross-referencing and included if they met the criteria cited above. We identified 1,063 papers. After screening and assessing for eligibility, we were able to include 547 studies in the meta-analysis. A flow chart of the search for studies is provided in Appendix S1, and the studies included are listed in Appendix S2 (see supporting information).
If between-group differences were provided for several subgroups within the same publication (e.g., for different conditions or for girls and boys), we entered them separately in our analysis instead of entering the global association. For computation of an effect size for total parenting stress in studies with the CHQ-PF, the effect sizes for emotional impact and time impact were averaged.
All studies were coded by the author, and a random sample of 150 studies was coded also by a psychologist with experience in meta-analyses. Differences between the two coders were resolved by discussion. An overview of the assessed variables and interrater reliability is presented in Table 1.
| Variable and coding | Interrater agreement |
|---|---|
| Number of parents of children with a chronic condition | ICC = 1.0 |
| Number of parents in the control group | ICC = 1.0 |
| Mean age of parents (in years) | ICC = .96 |
| Mean age of children (in years) | ICC = .99 |
| Percentage of girls | ICC = .96 |
| Percentage of mothers | ICC = .95 |
| Percentage of married parents | ICC = .98 |
| Percentage of members of ethnic minorities | ICC = .98 |
| Year of publication or presentation | ICC = 1.0 |
| Type of illness (1 = allergies, 2 = arthritis, 3 = asthma, 4 = cancer, 5 = cardiovascular diseases, 6 = cerebral palsy, 7 = craniofacial conditions, 8 = cystic fibrosis, 9 = diabetes, 10 = epilepsy, 11 = HIV infection or AIDS, 12 = kidney–liver–renal diseases, 13 = obesity, 14 = sensory impairment, 15 = spina bifida, 16 = traumatic brain injury, 17 = other diseases) | 97% |
| Duration of illness (in months) | ICC = .93 |
| Method of assessing parenting stress (1 = PSI, 2 = CHQ-PF, 3 = SIPA, 4 = Family Stress Scale, 5 = PIP, 6 = others) | 99% |
| Methods of assessing the correlates of parenting stress (categorical variables) | 95% |
| Representativeness of the sample for which it was recruited (1 = yes, 2 = no/not reported) | ICC = .90 |
| Equivalence of patient and control group (use of equivalent groups or statistical control for confounders; 1 = yes, 2 = no) | ICC = .93 |
| Use of valid and reliable parenting stress measure (1 = yes, 0 = no) | ICC = .98 |
| Sufficient test power for detecting an effect of d = 0.40 (1 = yes, 0 = no) | -a |
| Publication status (1 = published, 0 = unpublished) | ICC = 1.0 |
| Standardized size of between-group differences in total parenting stress (d score) | ICC = .96 |
| Standardized size of between-group differences in the five domains of parenting stress (d score) | ICC = .94–.97 |
| Correlations of the total parenting stress score with the child's age | ICC = 1.0 |
| Correlations of the total parenting stress score with illness duration | ICC = 1.0 |
| Correlations of the total parenting stress score with illness severity | ICC = .95 |
| Correlations of the total parenting stress score with level of behaviour problems | ICC = .94 |
| Correlations of the total parenting stress score with parental gender | ICC = .98 |
| Correlations of the total parenting stress score with parental mental health | ICC = .95 |
| Correlations of the total parenting stress score with marital status of the parent | ICC = 1.0 |
| Correlations of the total parenting stress score with marital quality | ICC = .97 |
| Correlations of the total parenting stress score with level of support | ICC = 1.0 |
- Note. CHQ-PF = Child Health Questionnaire-Parent Form; ICC = intraclass coefficient; PIP = Pediatric Inventory for Parents; PSI = Parenting Stress Index; SIPA = Stress Index for Parents of Adolescents.
- a Computation with G*Power based on the sample sizes.
3.2 Measures
For assessment of parenting stress, 300 studies provided data on versions of the PSI (Abidin, 1995), 202 studies used versions of the CHQ-PF (Landgraf et al., 1996), 21 studies applied the Pediatric Inventory for Parents (Streisand et al., 2001), 10 studies used the SIPA (Sheras et al., 1998), and 20 studies provided data on other parenting stress measures. Illness severity was assessed with measures of impairment of physical functioning, such as the Gross Motor Function Classification System (Brossard-Racine et al., 2012; 38 studies) and assessments of symptom severity and/or frequency (e.g., number of seizures; 46 studies). Child behaviour problems were measured with the Child Behavior Checklist (Achenbach, 1991; 32 studies) and related instruments (28 studies). Parental mental health was assessed with anxiety scales, such the State–Trait Anxiety Inventory (Spielberger, Gorsuch, & Lushene, 1970; eight studies), depression scales (nine studies), and sum measures of mental health (eight studies). Marital quality was assessed with the Dyadic Adjustment Scale (Spanier, 1976; four studies) and related instruments (three studies). Finally, perceived social support was measured with the Multidimensional Scale of Perceived Social Support (Zimet, Dahlem, Zimet, & Farley, 1988; seven studies) and related instruments (21 studies).
3.3 Statistical integration of the findings
- For comparisons of families with and without a child with a chronic condition, we computed effect sizes d as the difference in parenting stress between the sample with chronic condition and the control sample divided by the pooled standard deviation. If the authors only provided test scores for parents of children with chronic condition, we used the norms from the test manuals or from national comparative samples for comparison. For analysing associations of parenting stress within families containing a child with a chronic condition, we extracted Pearson correlation coefficients. Outliers that were more than 2 SD from the mean of the effect sizes were recoded to the value at 2 SD, based on Lipsey and Wilson (2001).
- The effect sizes d were transformed to Hedges' g, in order to correct for bias due to overestimation of the population effect size in small samples. The correlation coefficients were transformed using Fisher's r-to-z transformation.
- Weighted mean effect sizes and 95% confidence intervals were computed. The significance of the mean was tested by dividing the weighted mean effect size by the standard error of the mean. To compare the mean z scores with the correlation coefficients reported in the single studies, the mean z scores were later converted to the original metric of product–moment correlations. To interpret the practical significance of the results, we used Cohen's (1992) criteria: Effect sizes of g = 0.20 to 0.49 and r = .10 to .29 are small, g = 0.50 to 0.79 and r = .30 to .49 are medium, and g ≥ 0.8 and r ≥ .5 are large.
- Homogeneity of effect sizes was computed by use of the Q statistic.
- In order to test the influence of moderator variables, we used an analogue of an analysis of variance. A significant Q score indicates heterogeneity of the effect sizes between the compared conditions. If more than two conditions are compared, differences between two conditions are interpreted as significant if the 95% confidence intervals of two effect sizes do not overlap (Lipsey & Wilson, 2001).
- The trim-and-fill algorithm by Duval and Tweedie (2000) was used in order to check whether the results may have been influenced by a publication bias.
4 RESULTS
The 547 studies provided data on 50,370 parents of children with a chronic condition. The children most often had cerebral palsy (n = 7,380), juvenile arthritis or rheumatism (n = 5,412), cancer (n = 3,907), cardiovascular diseases (n = 3,811), visual or hearing impairments (n = 3,345), diabetes (n = 2,759), asthma (n = 2,631), epilepsy (n = 1,978), and cystic fibrosis (n = 1,747). The children had a mean age of 7.64 years (SD = 3.58; based on 504 studies), and 49% were girls (based on 447 studies). Their condition lasted, on average, for 4.62 years (SD = 3.04, based on 248 studies). About 80% of the participating parents were mothers (79.88%, based on 299 studies). Parents had a mean age of 34.99 years (SD = 4.38, based on 186 studies), 80% of them were married (based on 161 studies), and about 27% belonged to an ethnic minority (based on 241 studies).
4.1 Relative levels of parenting stress in families with a child with a chronic condition
For answering the first and second research questions, we compared levels of parenting stress of families with and without a child with a chronic condition. Because the effect sizes differed between studies with global parenting stress measures (mainly the PSI) and measures that address parenting stress related to child health and behaviour (e.g., CHQ-PF), Q(1) = 98.66, p < .001, we carried out separate analyses for general and health-related stress. As shown in Table 2, there were small elevations in the sum score of general parenting stress (g = 0.41) as well as in the parent domain (g = 0.32) and the child domain (g = 0.45) of the PSI. However, moderate to large elevations of parenting stress were observed with regard to dysfunctional parent–child interactions (g = 0.79). The total score of health-related parenting stress showed large elevations in families with a child with chronic condition (g = 0.86). Elevations on the emotional impact scale can also be interpreted as large (g = 0.94), whereas elevations in the impact of parenting on time use can be interpreted as moderate to large (g = 0.76; Cohen, 1992). The nonoverlaps of the 95% confidence intervals indicate that elevations of dysfunctional parent–child interactions were stronger than elevations in the other two domains of the PSI. Similarly, elevations on the emotional impact scale of the CHQ-PF were stronger than elevations in the time impact scale. Thus, caring for a child with chronic illness is more likely to cause elevated negative emotions than restrictions in parental time use. Nonetheless, additional analyses showed similar effect sizes for time impact and emotional impact in the case of spina bifida (g = 1.44, 95% CI [0.95, 1.03], vs. g = 1.14, 95% CI [0.68, 1.59]) and cerebral palsy (g = 1.28, 95% CI [1.12, 1.45], vs. g = 1.09, 95% CI [0.91, 1.27])—two diseases associated with physical impairments and related needs for parental support.
| Kind of stress | k | g | 95% CI | Z | Q | |
|---|---|---|---|---|---|---|
| Lower limit | Upper limit | |||||
| Global parenting stress | ||||||
| Total score | 397 | 0.41 | 0.35 | 0.46 | 14.21*** | 440.49 |
| Parent domain | 267 | 0.32 | 0.25 | 0.38 | 9.23*** | 256.18 |
| Dysfunctional parent–child interaction | 88 | 0.79 | 0.59 | 0.99 | 7.79*** | 96.27 |
| Difficult child | 251 | 0.45 | 0.38 | 0.52 | 12.42*** | 243.46 |
| Health-related parenting stress | ||||||
| Total score | 247 | 0.86 | 0.79 | 0.93 | 24.03*** | 206.29 |
| Parenting impact on emotions | 247 | 0.94 | 0.88 | 1.00 | 29.05*** | 270.97 |
| Parenting impact on time | 247 | 0.76 | 0.69 | 0.83 | 20.94*** | 245.55 |
- Note. g = weighted effect size (positive scores indicate higher stress in parents with ill children than in other parents); 95% CI = lower and upper limits of 95% confidence interval; Q = test for homogeneity of effect sizes; Z = test for significance.
- * p < .05.
- ** p < .01.
- *** p < .001.
Between-group differences in parenting stress varied by the measures used, Q(1) = 100.57, p < .001. Studies with the CHQ-PF report larger effect sizes (d = 0.85, 95% CI [0.78, 0.92], z = 23.97, p < .001) than studies that used the PSI (d = 0.41, 95% CI [0.35, 0.46], z = 13.89, p < .001), the SIPA (an adaptation of the PSI; d = 0.02, 95% CI [−0.41, 0.45], z = .08, n.s.), the Family Stress Scale (d = 0.48, 95% CI [0.00, 0.96], z = 1.97 p < .05), and other measures (d = 0.43, 95% CI [0.14, 0.73], z = 2.89, p < .01).
With regard to the third research question, we found that the elevations of general and health-related parenting stress varied by kind of condition (Table 3). Large elevations of general parenting stress were observed when children had an HIV infection or AIDS (g = 1.22), whereas moderate elevations were found in the case of epilepsy (g = 0.79), cerebral palsy (g = 0.62), cancer (g = 0.59), and traumatic brain injury (g = 0.59). In addition, small elevations of general parenting stress were observed in the case of spina bifida (g = 0.42), asthma (g = 0.35), allergies (g = 0.33), cystic fibrosis (g = 0.29), and sensory impairment (g = 0.20). No statistically significant elevations were observed in the case of juvenile arthritis or rheumatism, cardiovascular diseases, craniofacial conditions, diabetes, kidney–liver–renal diseases, or obesity. Large elevations of health-related parenting stress were observed if a child had spina bifida (g = 1.26), cerebral palsy (g = 1.15), cancer (g = 1.11), juvenile arthritis or rheumatism (g = 0.90), cardiovascular diseases (g = 0.89), and epilepsy (g = 0.84, Table 3). Statistically significant, moderate elevations of health-related parenting stress were found in families with a child with sensory impairments (g = 0.74), kidney–liver–renal diseases (g = 0.73), diabetes (g = 0.74), allergies (g = 0.63), or asthma (g = 0.56), whereas small elevations were found in the case of obesity (g = 0.48) and traumatic brain injury (g = 0.33). Health-related parenting stress in families with and without a chronic condition did not differ significantly in the case of craniofacial conditions, cystic fibrosis, or HIV infections or AIDS. As only one to three studies were available for these analyses, between-group differences in parenting stress may not have reached statistical significance due to restricted test power.
| Global parenting stress | Health-related parenting stress | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| k | g | 95% CI | Z | Q | k | g | 95% CI | Z | Q | |||
| Lower limit | Upper limit | Lower limit | Upper limit | |||||||||
| Illness | 63.29*** | 58.24*** | ||||||||||
| Allergies | 11 | 0.33 | 0.02 | 0.65 | 2.08* | 5.02 | 3 | 0.63 | 0.15 | 1.12 | 2.57* | 0.77 |
| Arthritis | 4 | 0.16 | −0.40 | 0.71 | 0.55 | 0.58 | 45 | 0.90 | 0.78 | 1.03 | 14.29*** | 39.46 |
| Asthma | 19 | 0.35 | 0.11 | 0.60 | 2.83** | 36.18** | 10 | 0.56 | 0.31 | 0.81 | 4.33*** | 2.53 |
| Cancer | 46 | 0.59 | 0.43 | 0.76 | 7.21*** | 96.22*** | 17 | 1.11 | 0.88 | 1.33 | 9.56*** | 22.96 |
| Cardiovascular diseases | 31 | 0.15 | −0.05 | 0.34 | 1.50 | 25.82 | 19 | 0.89 | 0.68 | 1.10 | 8.35*** | 28.95* |
| Cerebral palsy | 39 | 0.62 | 0.45 | 0.79 | 7.15*** | 37.41 | 28 | 1.15 | 0.98 | 1.32 | 13.18*** | 32.28 |
| Craniofacial condition | 23 | 0.12 | −0.11 | 0.34 | 1.03 | 6.27 | 3 | −0.01 | −0.55 | 0.53 | −0.04 | 0.24 |
| Cystic fibrosis | 27 | 0.29 | 0.07 | 0.50 | 2.62** | 13.68 | 2 | 0.55 | −0.04 | 1.13 | 1.81 | 0.05 |
| Diabetes | 22 | 0.14 | −0.09 | 0.36 | 1.19 | 10.57 | 8 | 0.74 | 0.45 | 1.03 | 4.87*** | 3.27 |
| Epilepsy | 21 | 0.84 | 0.60 | 1.08 | 6.90*** | 18.60 | 8 | 0.76 | 0.46 | 1.05 | 5.04*** | 6.57 |
| HIV infection | 5 | 1.22 | 0.72 | 1.72 | 4.78*** | 3.38 | 1 | 0.66 | −0.17 | 1.48 | 1.54 | 0.00 |
| KLR | 8 | 0.16 | −0.27 | 0.58 | 0.72 | 1.62 | 16 | 0.73 | 0.50 | 0.96 | 6.24*** | 17.87 |
| Obesity | 4 | 0.29 | −0.23 | 0.81 | 1.10 | 1.28 | 5 | 0.48 | 0.10 | 0.85 | 2.47* | 2.56 |
| Sensory impairment | 37 | 0.20 | 0.01 | 0.37 | 2.06* | 24.63 | 5 | 0.74 | 0.34 | 1.14 | 3.65*** | 4.76 |
| Spina bifida | 16 | 0.42 | 0.15 | 0.68 | 3.62*** | 15.75 | 8 | 1.26 | 0.92 | 1.60 | 7.25*** | 9.12 |
| TBI | 9 | 0.59 | 0.21 | 0.96 | 3.03** | 2.22 | 10 | 0.33 | 0.04 | 0.62 | 2.16* | 4.82 |
| Other diseases | 75 | 0.46 | 0.34 | 0.59 | 7.15*** | 95.21 | 59 | 0.83 | 0.71 | 0.95 | 13.93*** | 76.65* |
| Representativeness of the sample | 0.05 | 1.07 | ||||||||||
| No or not reported | 371 | 0.40 | 0.34 | 0.46 | 12.90*** | 338.67 | 218 | 0.84 | 0.77 | 0.90 | 24.92*** | 219.06 |
| Yes | 26 | 0.43 | 0.20 | 0.66 | 3.67*** | 51.84** | 29 | 0.94 | 0.76 | 1.12 | 10.22*** | 38.03 |
| Equivalence of patient and control group | 0.03 | 0.62 | ||||||||||
| No or not tested | 311 | 0.40 | 0.33 | 0.47 | 11.81*** | 307.30 | 211 | 0.86 | 0.79 | 0.93 | 25.07*** | 226.57 |
| Yes | 86 | 0.41 | 0.28 | 0.59 | 6.36*** | 83.18 | 36 | 0.79 | 0.63 | 0.95 | 9.68*** | 30.09 |
| Use of valid stress measure | 2.11 | |||||||||||
| No | 8 | 0.11 | −0.29 | 0.51 | 0.54 | 6.39 | 0 | |||||
| Yes | 389 | 0.41 | 0.35 | 0.47 | 13.51*** | 383.81 | 247 | 0.86 | 0.79 | 0.93 | 24.03*** | 206.29 |
| Sufficient test power | 0.00 | 6.86* | ||||||||||
| No | 231 | 0.40 | 0.32 | 0.48 | 9.72*** | 186.35 | 113 | 0.96 | 0.86 | 1.06 | 18.87*** | 139.16* |
| Yes | 166 | 0.40 | 0.32 | 0.49 | 9.24*** | 204.20* | 134 | 0.79 | 0.71 | 0.86 | 19.69*** | 116.44 |
| Publication status | 0.00 | 3.36 | ||||||||||
| Published | 363 | 0.40 | 0.34 | 0.46 | 12.86*** | 366.16 | 239 | 0.84 | 0.78 | 0.90 | 26.28*** | 246.70 |
| Unpublished | 34 | 0.41 | 0.20 | 0.61 | 3.82*** | 24.38 | 8 | 1.17 | 0.83 | 1.52 | 6.60*** | 10.34 |
- Note. g = effect size; k = number of studies; KLR = kidney/liver/renal disease; 95% CI = lower and upper limits of 95% confidence interval; Q = test for homogeneity of effect sizes; TBI = traumatic brain injury; Z = test for significance of g.
- * p < .05.
- ** p < .01.
- *** p < .001.
With regard to study quality, there were no moderating effects of the representativeness of the sample, of sociodemographic equivalence between the patient and control group, and of the use of valid and reliable measures of parenting stress (Table 3). Note that all studies on health-related parenting stress had used valid and reliable stress measures. However, we found a moderating effect of test power, and studies with higher test power (and larger samples) reported smaller group differences in health-related parenting stress than did other studies. Finally, the results of published and unpublished studies did not differ significantly. The trim-and-fill algorithm was applied for testing funnel plot asymmetry and adding possibly missing studies if asymmetry was detected. Although we found some asymmetry for the total parenting stress scores and the subscores, the re-estimated effect sizes were always slightly larger than the original effect sizes.
4.2 Correlates of parenting stress
For answering the final research question, we first tested whether the size of correlations differed between studies with general versus health-related parenting stress measures by using an analogue of the analysis of variance. In the case of significant differences, we report separate analyses for both groups of studies (Table 4).
| Moderators | k | r | 95% CI | Z | Q | |
|---|---|---|---|---|---|---|
| Lower limit | Upper limit | |||||
| Severity of illness | 101 | .19 | 0.15 | 0.23 | 9.73*** | 98.57 |
| Duration of illness | 45 | −.02 | −0.07 | 0.03 | −0.83 | 51.11 |
| Global stress | 28 | .03 | −0.04 | 0.09 | 0.87 | 39.17 |
| Health-related stress | 17 | −.10 | −0.18 | −0.02 | −2.48* | 5.73 |
| Child age | 75 | .01 | −0.02 | 0.04 | 0.76 | 90.28 |
| Global stress | 55 | .05 | 0.01 | 0.08 | 2.38* | 62.04 |
| Health-related stress | 20 | −.08 | −0.14 | −0.02 | −2.50* | 16.90 |
| Behaviour problems | 72 | .45 | 0.41 | 0.49 | 20.95*** | 71.66 |
| Parental mental health | 34 | −.51 | −0.57 | −0.46 | −19.88*** | 41.47 |
| Parental gender (female) | 53 | .07 | 0.03 | 0.10 | 3.78*** | 56.61 |
| Parent married | 15 | −.18 | −0.22 | −0.14 | −08.82*** | 18.85 |
| Marriage quality | 11 | −.29 | −0.34 | −0.24 | −11.99*** | 10.57 |
| Social support | 34 | −.29 | −0.34 | −0.24 | −10.97*** | 36.91 |
- Note. 95% CI = lower and upper limits of 95% confidence interval; Q = test for homogeneity of effect sizes; r = weighted effect size; Z = test for significance of r.
- * p < .05.
- *** p < .001.
With regard to child characteristics, we found that higher severity of the condition was associated with higher levels of parenting stress (r = .19, p < .001). Longer duration of the condition was associated with lower levels of health-related parenting stress (r = −.10, p < .05), whereas no such association appeared for general parenting stress. Higher levels of behaviour problems were associated with more parenting stress (r = .45, p < .001). Finally, parents of older children reported lower levels of health-related parenting stress (r = −.08, p < .05); the reverse association was found for general parenting stress (r = .05, p < .05).
Parental mental health showed a strong negative correlation with parenting stress (r = −.51, p < .001). In addition, mothers reported slightly higher levels of parenting stress than fathers (r = .07, p < .001). Furthermore, parents reported less parenting stress if they were married or cohabiting (r = −.18, p < .001) and if they perceived the quality of the marriage more positively (r = −.29, p < .001). Higher perceived social support was also associated with lower levels of parenting stress (r = −.29, p < .001). The nonoverlaps of the 95% confidence intervals indicate that associations of parental mental health and child behaviour problems with parenting stress were stronger than the other associations. Trim-and-fill analysis found no evidence for a possible publication bias in studies on correlates of parenting stress.
5 DISCUSSION
The present study is the first meta-analysis that compared levels of parenting stress across a broad range of physical conditions of the child and that analysed correlates of parenting stress in families with a child with a chronic physical condition. We found that parents of children with a chronic condition reported higher levels of parenting stress in general as well as health-related parenting stress in particular. Stress levels varied by kind of condition, child age, severity and duration of the condition, levels of behaviour problems in the child, parental mental health and gender, marital status and marital quality, and perceived availability of social support.
5.1 Comparisons of different aspects of parenting stress
The mean elevations of global parenting stress in families with children with a chronic condition were similar to those reported by Cousino and Hazen (2013). However, on the basis of broader inclusion criteria—such as adding more chronic conditions (+210 studies), studies published between July 2012 and May 2017 (+122 papers), unpublished studies (+38 papers), and studies that were not published in English (+29 papers)—we were able to include 42 times more studies than the previous meta-analysis, thus increasing the statistical power of our analyses.
The present meta-analysis provides a more detailed picture as we found stronger elevations of stress related to dysfunctional parent–child interactions than in the other domains of the PSI and stronger elevations of health-related parenting stress as compared to global parenting stress. Negative parent–child interactions refer to, for example, the perceptions of less than expected positive child behaviour towards the parent and the feelings that the parental efforts are not much appreciated by the child (Abidin, 1995). In some cases, high demands for illness management may not leave enough time for positive emotion-focused interactions with the child. Negative parent–child interactions also arise from conflicts about the distribution of illness management and about nonadherence of the child (e.g., Miller & Drotar, 2003) or from child resistance against elevated levels of parental overprotection (Pinquart, 2013).
The present meta-analysis showed that the impact of a chronic illness on the emotional lives of the parents tends to be larger than the impact on parental time use. Thus, many parents tend to be emotionally stressed even if they do not (or do no longer) have to invest much time into the illness management of their child. Qualitative studies show that many of these parents worry, among others, about possible future deterioration of the health of their child, problems that (may) emerge at school and in peer relations, or about the effect of the chronic condition of their child on healthy siblings (Coffey, 2006; DeVet & Ireys, 1998).
Only one of the criteria of study quality that was derived from Ferro and Speechley (2009) showed a significant moderating effect, thus indicating that our results were robust with regard to the other assessed criteria. Larger group differences in health-related parenting stress that were found in smaller samples might be based on publication bias as studies with smaller samples find more often nonsignificant effect that may remain unpublished (Lipsey & Wilson, 2001). Nonetheless, we did not find significant differences between effect sizes of published and unpublished studies, and the trim-and-fill analysis (Duval & Tweedie, 2000) indicated that the true effect sizes may even be larger than those reported in our meta-analysis. This result cannot be interpreted as a file-drawer problem in the classical sense, as studies with null results rather than those with above-average effect sizes would be expected to remain unpublished and unidentified (Lipsey & Wilson, 2001). The funnel plot asymmetry identified in the trim-and-fill analysis was probably based on other factors, such as larger average sample sizes of studies that assessed health-related parenting stress rather than general parenting stress.
5.2 Comparisons across chronic conditions
The present meta-analysis included a relatively large number of children with chronic conditions that have been understudied in the past, such as cerebral palsy (Britner, Morog, Pianta, & Marvin, 2003). We found the strongest elevations of parenting stress within families with a child with an HIV infection or AIDS, spina bifida, cancer, and cerebral palsy. Although two of these diseases are, in particular, associated with a reduction of life expectancy (AIDS and cancer), the other diseases are associated with persistent impairments of physical and, in part, cognitive functioning and with high support needs (Kliegman, Stanton, St. Geme, & Shor, 2015). Thus, our results indicate that different paths may lead to elevated levels of parenting stress in families with a child with a chronic condition.
In contrast to most assessed conditions, parenting stress was not enhanced in families with a child with craniofacial conditions, such as cleft lip. This could be due to the fact that such conditions are usually treated early in life, and parenting stress has often been assessed several years after the successful treatment.
5.3 Correlates of parenting stress in families with a child with a chronic condition
Whereas parental gender, child age, and duration of the illness could be conceptualized as predictors rather than outcomes of parenting stress, the other correlations may, in principle, indicate bidirectional associations as, for example, behaviour problems of the child, low mental health of the parent, and a low quality of the marital relationship may increase parenting stress whereas elevated stress levels could also negatively affect behaviour problems, mental health, and marital quality (e.g., Abidin, 1995; Deater-Deckard, 1998; Fonseca et al., 2011; Lavee, Sharlin, & Katz, 1996).
Most of the correlational results were in line with Abidin's (1995) model and will not be further discussed. However, comparisons of the size of the correlations and of our results with the review by Cousino and Hazen (2013) need some attention. Parental mental health and behaviour problems of the child were found to be the strongest correlates of parenting stress, thus indicating that measures aimed at improving mental health and behaviour problems in the parent and the ill child, respectively, could make parenting less stressful. Although mothers of children with a chronic condition reported higher parenting stress than fathers, gender differences were very small. As only about 20% of the participants were fathers, fathers with low engagement in the parenting role and related low levels of childcare-related stressors were probably underrepresented, thus reducing the size of observed gender differences in parenting stress.
Whereas the review by Cousino and Hazen (2013) had suggested that parenting stress is not related to the severity of the chronic condition, we found such an association, although the size of the association was small in a statistical sense. Most of the available studies assessed levels of functional impairment as an indicator of the severity of the condition. Higher impairments indicate higher needs for parental support and may, therefore, cause more parental stress.
6 LIMITATIONS AND CONCLUSIONS
Some limitations specific to the present study have to be mentioned. First, only few studies were available for some diseases, such as HIV infection or obesity. Similarly, few studies were available containing associations of marital status and marital quality with parenting stress. A larger number of studies would lead to more robust effect sizes. Second, fathers were underrepresented in most available studies on parenting stress. Third, the present analysis of concurrent associations of parenting stress with chronic condition of the child does not allow for testing causal relationships. Although there are good arguments for the suggestion that a child's chronic condition leads to elevated parenting stress, in some cases, parenting stress can also have an effect on the course of a child's condition when stress interferes with effective illness management (Celano et al., 2011; Streisand et al., 2001). Fourth, we limited the analysis of predictors of parenting stress to variables related to the model of Abidin (1995). Other factors may also play a role, such as parental socioeconomic status or ways of coping with stressors (Cousino & Hazen, 2013). Associations of parenting stress with measures of family functioning could also be addressed as some families tend to report strengthened bonds after a child is diagnosed with a chronic condition (Cowen et al., 1985).
Finally, moderator analyses of correlates of parenting stress were limited to the use of generic versus health-related stress measures and aspects of study quality. Nonetheless, as the average effect sizes were homogeneous, moderator effects of other study characteristics would be unlikely to be found.
Despite these limitations, several conclusions from a research, policy, and clinical perspective can be drawn. From the research perspective, more studies are needed on parenting stress pertaining to chronic conditions that were rarely addressed in the present meta-analysis and on effects of rarely assessed correlates of parenting stress, such as marital quality. As concurrent associations do not inform us about the causal direction of associations, we recommend longitudinal research on predictors of change in parenting stress in families with a child with a chronic condition. Given the differences in the effect sizes between studies with general and health-related measures, we conclude that both kinds of measures should be used in order to get a comprehensive view on parenting stress in families with a child with a chronic condition. Whereas health-related measures provide important insights into stress related to the child's condition, generic measures also address other stressors, such as those related to academic achievement.
With regard to policy perspective, many countries cut the costs of paediatric health care by reducing the frequency and duration of inpatient care—a trend that is likely to increase caregiving demands of the parents and related stressors (e.g., Shaw & McCabe, 2008). Given the observed elevations of parenting stress in families with a child with a chronic condition, cutting financial costs of paediatric care is likely to increase psychological costs for the parents, as indicated, among others, by elevated parenting stress. Thus, society should invest resources for reducing parenting stress and make these resources available for families in need. For example, Hotchkiss and Biddle (2009) suggested a variety of services aimed at reducing parenting stress, such as parenting training and support services, workshops on coping with parenting stress at hospitals, a parenting hotline giving parents immediate access to services and resources, and the creation and maintenance of parenting support groups.
With regard to conclusions for clinicians, the observed highest rates of parenting stress in families with a child with cancer, cerebral palsy, an HIV infection or AIDS, and spina bifida indicate that these parents should be, in particular, screened for parenting stress and receive psychosocial services aimed at reducing parenting stress, if needed. Practitioners should, first, offer measures that reduce stressors—such as community services for children with chronic condition that reduce their behavioural and psychological problems as well as the severity of the child's physical condition and related care demands. Available studies found small intervention effects on the psychological adjustment of children with a chronic condition (Pai, Drotar, Zebracki, Moore, & Youngstrom, 2006), whereas the effect of these interventions on parenting stress still has to be proven. Second, interventions for parents should increase their resources for dealing with stressors that may be difficult to change. Resource-building interventions with parents should train skills for dealing with stressors of parenting a child with a chronic condition (e.g., problem-solving skills, parenting skills training aimed at improving the communication with the child). In addition, respite child care services could reduce parental caregiving demands. A review of interventions for reducing parenting stress in families with children with heterogeneous paediatric conditions by Golfenshtein, Srulovici, and Deatrick (2016) found significant improvement of parenting stress in most of the included studies, although effects on parenting stress varied between individual interventions. Interventions focusing on the improvement of the parent–child relationship and skill-building interventions had the most consistent short-term effects on parenting stress. As long-lasting effects of these interventions were less often found than short-term effects (Golfenshtein et al., 2016), more efforts are needed for increasing the persistence of intervention effects.
