Abstract

Meta-analysis was used to synthesize findings from comparative studies of depression in mothers of children with and without developmental disabilities. Effect sizes were determined for 18 studies conducted between 1984 and 2003. A weighted effect size of .39 indicated an elevated level of depression in mothers of children with developmental disabilities. Planned comparisons found that age of child and disability category moderated effect sizes. Results show that mothers of children with developmental disabilities are at elevated risk of depression compared to mothers of typically developing children. Depression in mothers of children with developmental disabilities is a condition that is presently not being addressed on a wide scale, although promising interventions are available.

For over 4 decades, researchers have identified maternal depression as one problematic response to parenting children with disabilities (Cummings, Bayley, & Rie, 1966; Gath, 1977; Glidden & Schoolcraft, 2003; Nixon & Singer, 1993; Wolfensberger, 1967). Several researchers comparing mothers of children with and without disabilities have reported significantly higher levels of depressive symptoms in mothers of children with developmental disabilities (Amaral, 2003; Beckman, 1991; Breslau & Davis, 1988; Breslau, Staruch, & Mortimer, 1982; Kazak, 1987; Kazak & Marvin, 1984; Miller, Gordon, Daniele, & Diller, 1992; Roach, Orsmond, & Barratt, 1999; Scott, Atkinson, Minton, & Bowman, 1997; Witt, Riley, & Coiro, 2003; Wolf, Noh, Fisman, & Speechley, 1989). This evidence, however, is more complex because in several comparative studies, researchers have also found no significant differences between groups (Barakat & Linney, 1992; Bristol, Gallagher, & Schopler, 1988; Capelli, 1990; Gowen, Johnson, Goldman, & Appelbaum, 1989; Kazak & Marvin, 1987; Harris & McHale, 1989; Seltzer, Greenberg, Floyd, Pettee, & Hong, 2001; ,Walker, Ortiz-Valdes, & Newbrough, 1989).

These mixed and seemingly contradictory findings are consistent with an emerging understanding that there is a wide range of responses to parenting children with developmental disabilities (Glidden, Kiphart, Willoughby, & Bush, 1993), including positive outcomes (Hastings & Taunt, 2002). An earlier assumption of virtually universal and unvarying negative impacts on families (Wolfensberger 1967) has been displaced by a more complex understanding of family adaptation (Fergusen, 2001; Glidden et al., 1993; Singer & Irvin, 1991; Turnbull, Turbiville, & Turnbull, 2000) and of long-term resilience (Seltzer et al., 2001).

Despite the acknowledgement of variability in outcomes, there are reasons why it is important to determine whether there is, on average, an elevated level of depression in mothers of children with developmental disabilities and, if so, the magnitude of the difference in depression levels between mothers of children with and those without disabilities. Foremost among these reasons is the fact that both minor and major depression are linked to detrimental individual, familial, and societal outcomes (Glidden & Schoolcraft, 2003; Zimmer & Minkovitz, 2003). Even mild depression is a cause for concern. People with minor depression have shown considerable limitation and distress in social, work, and physical functioning. Hays, Wells, Sherbourne, Rogers, and Spritzer (1995) reported that elevated levels of depressive symptoms were associated with lowered well-being, impaired role function, impaired social function, and poor general health. In their longitudinal study of 1,790 persons, they found that people with minor depression had lower levels of functioning than people with serious chronic illnesses. Major depression is associated with longer lasting and more severe limitations in functioning (Kessler, Zhao, Blazer, & Swartz, 1997). If a group of people are at greater risk of depression, they should be an important social concern, deserving of preventive and ameliorative efforts.

Another reason for addressing the question of whether or not mothers of children with disabilities experience more depression on average is that results may help researchers to better understand the etiology of depression in women in general. It is well-established that women of child-bearing age in the general population are at elevated risk for depression (Kessler, 2003). In studies using self-report questionnaires to assess prevalence of depression among large samples of women, approximately 20% have mild to major depression (Lin, Dean, & Ensel, 1988). Investigators using standardized clinical interviews to determine whether women experience depression have found prevalence rates ranging from 8% to 20% (Kessler, 2003). One hypothesis about the causes of depression is that it is, in part, the result of elevated stress in daily life. For example, Breslau and Davis (1986) used a sample of mothers of children with developmental disabilities and chronic illnesses to assess the impact of presumed chronic stress. Their findings were reported in the context of contributions to the larger theoretical question of the relationship between chronic stress and depression.

The question is also important for applied reasons, specifically, the design of developmental disability service system and related social policy. If mothers of children with developmental disabilities are at higher risk for depression than those of children without disabilities, the existing social service system for children with developmental disabilities and their families may need to specifically target this problem. Several randomized studies have shown that depressive symptoms in mothers of children with developmental disabilities can be effectively reduced with certain kinds of interventions (Hastings & Beck, 2004). The weight of the evidence suggests that cognitive behavioral treatments with additional support services are highly promising as ways to reduce maternal distress (Bristol, Gallagher, & Holt, 1993; Hawkins, Singer, & Nixon, 1993; Nixon & Singer, 1993; Pelchat, Bisson, Ricard, Perreault, & Bouchard, 1999; Singer et al., 1994; Singer, Irvin, & Hawkins, 1988; Singer, Irvin, Irvine, Hawkins, & Cooley, 1989; Singer et al., 1993). We note that all of these interventions brought parents together in groups as sources of mutual support and included structured content presented by a professional facilitator, often psychoeducational material aimed at improving coping skills. These interventions, however, are not widely available in the United States. They also may not be sufficient for some mothers and, consequently, the service system may need to make available more intensive forms of psychological support to some mothers. Singer et al. (1993) noted that in their intervention studies, a small number of mothers did not respond sufficiently to psychoeducational interventions and needed additional treatment, either marital counseling or individual psychotherapy for depression. The need for these services can be partly determined via meta-analysis.

Meta-analysis is a way to synthesize a body of research literature statistically. It standardizes the difference between groups in each individual study, allowing them to be combined across studies. It has been particularly useful in medicine, education, and psychology when a body of studies has yielded inconsistent findings when reviewed by the traditional method of comparing the number of studies with significant and nonsignificant findings (Hedges & Olkin, 1985). This traditional approach is flawed because statistical significance is dependent, in part, on sample size. Studies with small samples may fail to detect important differences, even though large differences exist in the populations and, conversely, those with large samples may find significance for very small differences between populations. Further, simple counts of the number of studies with and without significant findings do not give a sense of the magnitude of the difference between populations.

Meta-analysis uses a standardized indicator of the magnitude of difference between groups, effect size. One commonly used statistic in research syntheses is d (Hedges & Olkin, 1985). Two recent meta-analyses illustrate the utility of this method in synthesizing family research in the developmental disabilities field. Risdall and Singer (2004) reviewed 13 studies, comparing marital adjustment in families of children with and without developmental disabilities. The effect size, d = .21, was small, indicating that the impact of raising children with developmental disabilities on marital relationships is negative but relatively minor. Another meta-analysis by Rossiter and Sharpe (2001) synthesized comparative studies on the impact on siblings of having a brother or sister with mental retardation. They found an overall effect size of d = −.06, indicating that the overall impact across 79 comparisons was also negative but very small. These results help to clarify contradictory outcomes between individual studies, and they make it possible to generally characterize the strength of the impact of children with developmental disabilities on their families. Both of these results contradict earlier historical assumptions of more substantial negative impacts of children with developmental disabilities on family relationships while also indicating that there is legitimate cause for concern about some couples and, to a lesser extent, siblings in these families (Risdall & Singer, 2004).

My purpose in conducting this study is to use meta-analysis to answer questions about the prevalence of depression in mothers of children with and without disabilities. My first research question is, Across studies, what is the difference in the average level of depressive symptoms in mothers of children with developmental disabilities compared to mothers of nondisabled children as indicated by the effect size, d?

Effect sizes are difficult to understand intuitively. In order to provide more information about the differences between the two groups, one way to give a sense of the size of these differences is to look at simple percentage comparisons. For the most commonly used measures of depressive symptoms, cut-off scores have been established such that scores over a specified number indicate the likely existence of either minor or major depression, permitting a comparison of the percentage of parents with depression in the two groups. The second question is, What percentage of mothers in the two groups have depression when it is measured as a dichotomous variable (depressed or nondepressed)?

The third question concerns whether there are moderating variables that partly account for differences in effect sizes between studies. Based on theory and previous research, I identified three variables as possible moderators for planned comparisons: (a) year of publication, (b) age of child, and (c) disability label. Year of publication was chosen as a possible moderator because service systems in the United States and Canada have increasingly focused on family support, so that it is possible that maternal depression may have declined in prevalence between 1982 and 2003. I selected child age as a possible moderator because recent research suggests that maternal depression declines over time (Glidden & Schoolcraft, 2003) and disability category because there is some evidence that parents with children who have autism have unusually high levels of stress (Moes, Koegel, Schreibman, & Loos (1992).

The fourth and final question is a technical one that is important in meta-analytic studies. There is concern that studies showing significant differences are more likely to be published than are studies with nonsignificant findings, possibly biasing the outcome. It is possible that there are a number of unpublished studies in which researchers did not find any significant differences, and these studies are languishing in file drawers. A statistical method, the fail-safe statistic, has been developed to determine how many such unpublished studies would be needed to reduce the average effect size between groups in published studies to a negligible level (Hedges & Olkin, 1985). The higher the fail-safe number of studies, the less likely it is that the effect size misrepresents the data because of publication bias. Therefore, the fourth question is, how many unpublished studies would need to exist to reduce the effect size to a negligible number?

Method

Search Procedures

I used the four approaches recommended by Cooper (1989) to search for studies. Journals were searched electronically using the databases PsychINFO, PubMed, and ERIC and unpublished dissertations were sought from the Dissertation Abstracts International database. The search was limited to studies published in the United States and Canada because one research question concerned whether there have been significant changes over time in effect sizes that might be associated with the increase in early intervention and family support services in these two countries during the past 20 years. My working definition of developmental disability was a modification of the definition in the Developmental Disabilities Act (PL 98–527). Developmental disabilities were defined as occurring before age 21, likely to continue indefinitely, require professional services of lifelong duration, and result in functional limitations in at least three major areas of life activities. I narrowed this broad definition by including only those developmental disabilities associated with mental retardation or a combination of intellectual and physical impairments (i.e., physical disabilities and chronic illnesses not associated with cognitive disabilities were excluded). Consequently, search terms for all databases included all possible combinations of the terms disability, developmental disability, handicap, autism, mental retardation, cerebral palsy, traumatic brain injury, and spina bifida with stress, depression; or distress with parents, mothers, and families. Databases were searched for all years available. When a journal was identified with articles that met the eligibility criteria, a title search was conducted for all issues of that journal. Titles and abstracts were read for all studies that included the search terms.

All promising articles, dissertations, and book chapters were obtained and reviewed. References from these studies were searched, in turn, to identify other studies. References from reviews of the literature published as chapters in books were also searched (Singer & Irvin, 1991). Finally, authors who had published two or more studies on the topic were contacted by e-mail and asked whether they had additional published or unpublished data sets or knew of other studies. In addition, authors were contacted by letter or e-mail when further data were needed in order to determine an effect size or clarify possible errors in published data.

To be included in the meta-analysis, researchers had to have collected data on depressive symptoms or general psychological distress, including depression, by using published standardized self-report measures with well-established psychometric properties. Studies with researcher-created measures with unknown psychometric properties were not included.

Studies had to include mothers of children with developmental disabilities and a comparison group of mothers of children without disabilities. Studies in which researchers compared one group of mothers of children with disabilities with population norms on standardized measures were excluded.

Studies also were excluded if (a) sufficient statistical data were not available in order to calculate an effect size, (b) if the samples combined did not disaggregate data from mothers and fathers, and (c) if they were later determined to be outliers because of highly unusual effect size values (Hedges & Olkin, 1985).

In the kind of meta-analysis used in this paper, namely, fixed-effect modeling, an assumption is that effect sizes are independent. Some investigators reported on the means and SDs for more than one disability comparison group. In these instances, the depression scores were averaged across groups with different developmental disability labels. Similarly, data from longitudinal studies were averaged across time to yield one between-groups comparison.

Statistical Analysis

I calculated effect sizes for the differences between the two groups for the individual studies and an overall mean effect size, d, was produced, as recommended by Hedges and Olkin (1985). Studies reporting means and SDs were converted via the equation

 
formula

where d is the effect size, Xt is the mean of the disability group, Xc is the mean of the nondisability group, and s is the pooled SD of the two groups. Hedges and Olkin (1985) modified this effect size slightly by multiplying by a small correction factor that reduced bias. In the kind of effect size analysis used in this study, a fixed effects model, individual ds are averaged to determine an overall effect size. If researchers reported frequencies or proportions, those were converted to an effect size via the equation d = (PePc)/ spooled, where d is the study effect size, Pe is the proportion in the experimental group (mothers of children with disabilities), Pc is the proportion in the comparison group (mothers of children without disabilities), and spooled is the pooled SD of the two groups. I calculated effect sizes for ANOVA F tests using the formula d = {F[(ne + nc)/(nenc)]}1/2 where F is the ANOVA outcome, ne is the number of subjects in the disability group, and nc is the number of subjects in the nondisability group (Hedges & Olkin, 1985).

Studies with larger samples are likely to produce more accurate results than those with smaller samples. Thus, individual effect size statistics need to be refined by giving more weight to larger studies (Hedges & Olkin, 1985). Further, as discussed below, studies by researchers using better sampling procedures and designs and those reporting key variables are also likely to better approximate the population effect size. Consequently, a weighted effect size was calculated to account for both sample size and quality (Shadish & Haddock, 1994). The following formula was used to calculate an overall effect size weighted by both sample size and quality:

 
formula

where Te is the estimate of the population effect size, wi is the inverse of the variance for each effect size, Wi is a score on a quality index, and Ti is the individual effect size for each study (Shadish & Haddock, 1994). The inverse of the variance w was calculated using the sample sizes ne and nc for each study.

To determine whether all samples (effect sizes) were drawn from a common population of effect sizes with a shared mean, I performed a test of homogeneity. This Q statistic was calculated using Hedges and Olkin's (1985) formula:

 
formula

The Q statistic is tested via traditional inferential methods. If this homogeneity test produces a result that is significant at the .05 level (of chi-square at k − 1 dfs), the individual studies are considered to be more variable than expected due to sampling error alone; thus, if the Q is significant, α = .05, the effect sizes are heterogeneous and cannot be assumed to have been drawn from the same population of studies. It is important to note that the homogeneity of variance refers to the differences between studies, not subjects within studies. In effect, homogeneity across studies indicates that the participants in all of the studies can be treated as one large pooled sample.

Planned comparisons

Rosenthal, Rosnow, and Rubin (2000) presented a procedure for conducting planned comparisons in meta-analysis in order to identify significant moderating variables. Three comparisons were planned based on theory or previous research findings. I hypothesized that the year of publication would be associated with a downward trend, such that effect sizes would be smaller the more recent the publication date. A second planned comparison was based on the hypothesis that studies with samples of parents of children with autism would have larger effect sizes than studies of parents of children with mental retardation or spina bifida. Finally, I conducted a planned comparison to test the hypothesis that parents of younger children would have higher levels of depressive symptoms than mothers of adults with developmental disabilities. The formula I used to conduct planned comparisons as recommended by Rosenthal, Rosnow, and Rubin (2000) is

 
formula

in which Mi is the mean for the ith group of effect sizes, λi is a weight assigned to each group of effect sizes based on the hypothesis or previous research findings, MSwithin is the mean sum of squares within groups, and ni is the number of effect sizes that make up a group. For example, a planned comparison can be developed based on the hypothesis from previous research findings that studies of parents of children with autism will have significantly higher effect sizes than will studies of parents of children with mental retardation or studies of parents of children with spina bifida. The effect sizes are then aggregated into three groups, one for each disability category. In order to model the hypothesis, lambda weights of +2, −1, and −1 were assigned to the groups, respectively, thereby weighting the autism group higher than the other two groups, which are assumed to be equivalent. Fcontrast is based on traditional ANOVA and is analogous to it.

Fail-safe statistic

The formula used to calculate the fail-safe number is k0 = k(ddc)/dc in which k0 is the number of studies with null results needed to render the average effect as negligible and k the number of studies in the synthesis, d = the overall average effect size, and dc is the effect size at a negligible level; in this analysis it was set at .01 (Hedges & Olkin, 1985).

To conduct the effect size analyses for this study, I utilized a commercially available computer program, METAWIN (Rosenberg, Adams, & Gurevitch, 1997) specifically designed for performing meta-analyses. METAWIN does not include the procedure for calculating the quality weighted effect size, so a program developed with the assistance of a statistician was used.

Weighting effect sizes for quality

To deal with variability in the quality of studies, I assigned quality weightings. The process of developing and applying these weights is, in fact, a detailed effort to analyze and summarize methodological rigor. Following a validity framework recommended by Lipsey (1994), quality weights were determined for each effect size based on (a) suggestions from narrative reviews and other relevant literature, (b) conceptual and methodological considerations, and (c) key variables identified in the literature that might confound the relationship between groups and the dependent measures. Using these criteria, I rated several variables for quality.

The first concern was the extent of demographic information provided about the mothers in both groups in each study. Several demographic variables correlated significantly with the prevalence of elevated levels of depressive symptoms in large population studies (Lin et al., 1986). The variables were (a) family income, (b) employment, (c) race or ethnicity, (d) marital status, and (e) educational attainment. Each identified demographic variable was assigned a weight of either 1 if reported or 0 if not reported.

A similar methodological concern had to do with the amount of descriptive information reported about the children in both groups and the extent to which the samples had homogeneous diagnostic categories. The following variables were judged as important based on a reading of the literature: (a) child's age, (b) child's primary diagnosis, and (c) an indicator of the severity of the condition. Studies were assigned a weight of 1 for each reported variable or 0 if not reported. Child characteristics were coded by assigning a 0 for samples that combined children who had developmental disabilities with children who had chronic illnesses and a 1 for studies with samples solely made up of children with developmental disabilities. Quality also varied according to the method of obtaining the child samples; namely, whether they were convenience groups or based on larger population-sampling methods. If studies drew both samples of children from large population samples using methods of random selection, they were coded with a 1 and those with convenience sample were assigned a 0.

Another important quality dimension had to do with whether researchers tested for demographic differences between samples. If significant differences were identified, studies were rated for whether or not these differences were statistically controlled. For studies in which researchers either established equivalence of the two groups or statistically controlled for differences in demographic variables a 1 was assigned; if not, a 0.

Additional quality variables had to do with the level of peer review applied to each study and the dependent measures. Doctoral dissertations and presentations were assigned a 0 and publications in peer-reviewed journals, a 1. The last quality variable was the construct validity and psychometric maturity of the measures used in these studies. Widely used and psychometrically “mature” instruments were assigned a weight of 1. These measures were the Beck Depression Inventory (Beck, Steer, & Garbin, 1988); the Center for Epidemiology Depression Scale (Radloff, 1977); the Langer Symptom Checklist (Langer, 1962), the Brief Symptom Inventory (Derogatis & Melisaratos, 1983), and the Symptom Checklist 90 Revised (Derogatis, 1977). Researchers using instruments that have been challenged for their construct validity as a measure of depressive symptoms or those with psychometric properties that are unknown were assigned a 0. The Parent Stress Index Depression subscale (Abidin, 1995) was assigned a 0 because the author of the instrument asserted in the manual that this subscale was best understood as a measure of a parent's ability to marshal energy for parenting tasks rather than as traditional depression. In a study by Witt et al. (2003), an index consisting of three questions from the National Health Information Survey was also rated as 0 because no psychometric properties of this index were available.

Percentage comparisons

To aid interpretation of the effect sizes, I also used an additional approach, namely, comparisons of the percentages of mothers scoring above or below clinical cutoff scores for depression. In order to understand this approach, it is important to note that depression is sometimes reported as a continuous variable and sometimes as a dichotomous one (depressed or not depressed). Two of the measures most commonly used in survey studies of depression, the Center for Epidemiology Depression Scale and the Beck Depression Scale, have established clinical cutoff scores for determining whether or not a person falls into the category of having depression. These cut-off scores are based on studies comparing scores on self-report measures with results of psychiatric interviews. Percentage of participants in each group falling at or above the clinical cutoff score of 16 and 10, respectively, were calculated and compared. Because most of the researchers in this synthesis did not report on depression as a dichotomous variable, it was necessary to extrapolate from the data on means and SDs by first standardizing depression scores with the standardized z statistic and then finding the percentage of subjects above and below these cutoff scores using a z distribution table. The resultant percentages are approximations based on the assumption that scores in the studies were normally distributed. Although percentages provide only a rough indicator, they may help to make intuitive sense of the effect sizes. In two studies (Blacher, Lopez, Shapiro, & Fusco, 1997; Scott et al., 1997), the researchers did report on the percentages of subjects above and below clinical cutoff scores, and the published data rather than z scores were used.

Reliability

To determine interobserver agreement on quality weightings, a second reader coded 100% of the studies. Percentage of interobserver agreement was calculated by dividing the total number of agreements by agreements plus disagreements and multiplying by 100. In instances of disagreement, the author and the observer discussed the texts and arrived at an agreed upon correct data point for final calculation of the overall weighted effect size.

Results

Sample of Studies

In an initial search of the categories that yielded the largest number of studies (family and disability as title words), I found 276 articles in the PsychInfo database, 39 in the Dissertation Abstracts database, and 43 in ERIC. Of these 358 articles, 19 met the inclusion criterion. Sixteen were published in peer-reviewed journals, 1 was a paper presented at a conference (Seltzer et al., 1989), and 2 were unpublished dissertations (Amaral, 2003; Cappelli, 1990). I rejected Amaral's study as an outlier because Amaral reported an effect size eight times larger than the highest effect size in the other 18 studies (Hedges & Olkin, 1985). Subsequently, 18 studies were included in the analyses. Table 1 lists each of the studies alphabetically. It includes data on the authors and publication dates and the measure of depression, the means and SDs of the depression scores or the percentage over a clinical cut-off score, quality weightings, effect sizes, and 95% confidence intervals. As noted in the footnotes, when mean scores and SDs were not available, I calculated effect sizes from percentages, F tests, t tests, or chi-square tests as reported in the studies.

Table 1.

Outcomes of Studies Comparing Mothers of Children With and Without Developmental Disabilities Including Effect Size and Quality Ratings

Outcomes of Studies Comparing Mothers of Children With and Without Developmental Disabilities Including Effect Size and Quality Ratings
Outcomes of Studies Comparing Mothers of Children With and Without Developmental Disabilities Including Effect Size and Quality Ratings

Table 2 presents demographic information, including the mothers' education levels, mothers' age, ethnicity or race, family income or SES, and marital status. For the children age and disability categorical diagnosis were included.

Table 2.

Descriptive Information on Mothers and Children With Disabilities

Descriptive Information on Mothers and Children With Disabilities
Descriptive Information on Mothers and Children With Disabilities

Homogeneity of Variance

When the 18 studies were analyzed for homogeneity of variance, the Qw statistic was not significant, indicating that all of the effect sizes could be combined and the participants treated as one pooled sample consisting of 6,641 parents of children with disabilities and 26,438 parents of children without disabilities.

Effect Size Estimates

The average unweighted d for the 18 studies was .35, with the 95% confidence interval of +.29/+.39. When weighted by quality indicators and the reciprocals of the variances, the weighted effect size, d, was .39,. with a 95% confidence interval of +.31/+.47. Effect sizes of this magnitude are, as a rough rule of thumb, considered between small and moderate (Cohen, 1988).

Fail-Safe Number

The fail-safe calculations indicated that 34 studies with null results would need to exist in order to reduce the average d to a negligible level.

Percentage With Elevated Levels of Depressive Symptoms

A common sense, albeit crude, indicator of the difference between categorical variables is a simple contrast of percentages. Table 3 presents the percentage of mothers in each group whose scores fell on or above the clinical cut-off of 16 from the studies utilizing the Center for Epidemiology Depression Scale and 10 on the Beck Depression Inventory. Table 3 shows that, on average, 29% of mothers of children with developmental disabilities had elevated levels of depressive symptoms compared to 19% of parents in the comparison group, a difference in prevalence of 10%.

Table 3.

Mothers With and Without Children With Developmental Disabilities Scoring Above Clinical Cut-Off Scores on the Depression Measures

Mothers With and Without Children With Developmental Disabilities Scoring Above Clinical Cut-Off Scores on the Depression Measures
Mothers With and Without Children With Developmental Disabilities Scoring Above Clinical Cut-Off Scores on the Depression Measures

Planned Comparisons and Interobserver Reliability

Three planned comparisons were conducted in order to identify moderating variables. Sufficient data and a rationale for planned comparisons warranted comparisons for date of publication, child's age, and disability category. The comparison for year of publication yielded a nonsignificant result, indicating that the year in which the study was conducted was not a moderator of effect size differences between studies. The comparison of effect sizes by child age was significant, F(2,15) = 11.99, p = .001, indicating a significant difference between effect sizes in studies of parents of adult children and effect sizes from studies of parents of children in middle and early childhood. The planned comparison for disability category indicated that in studies focused on parents of children with autism, researchers found higher effect sizes compared to studies of parents of children with mental retardation and studies of parents of children with spina bifida, F(2, 15), p = .005.

Interobserver agreement on the quality weighting was 93.6% (range = 80% to 100%).

Discussion

It is clear from these data that mothers of children with disabilities are at elevated risk of depression compared to the population of mothers of nondisabled children sampled in these studies. Small to moderate effect sizes were relatively consistent over the period from 1982 through 2003. When examined in terms of percentages of women with scores over clinical cutoffs, there was a relatively consistent increase in prevalence of roughly 10%. Although women of child-bearing age in general are at high risk of depression, the addition of extra challenges associated with caregiving further raises the risk for psychological distress. The prevalence of elevated levels of depressive symptoms in the general population of mothers of all children in the United States has been recognized as an important public health problem (National Institute on Child Health and Human Development, 1999). The additional prevalence in mothers parenting a child with a developmental disability is cause for further concern. Almost one third, 29%, of the mothers of children with developmental disabilities in these studies experienced depression. These results are consistent with research suggesting that increased environmental stress is linked to depression (Witt et al., 2003).

The finding that effect sizes did not decrease over the period from 1982 to 2003 suggests that increases in services during this period of time did not impact depression, although they are likely to have had other benefits. Future researchers should include measurements of service utilization and satisfaction in order to better understand the relationship between services, their stated missions, and maternal satisfaction with them. These variables were not assessed in any of the studies under review.

These meta-analytic findings help to support the argument that there is need for more widely available interventions designed to prevent and treat depression in mothers of children with developmental disabilities. The fact that a large majority of mothers in these studies were not depressed suggests that depression is not an inevitable consequence of parenting children with developmental disabilities. Hastings and Beck (2003) reviewed the literature on interventions for stress in this population of mothers and concluded that group interventions using cognitive–behavioral stress management methods were promising. Singer et al. (1989) reported that half of the parents in their intervention study of stress management combined with additional support services showed clinically significant improvements in mental health. Promising interventions have all involved professionally led psychoeducational groups in which parent-to-parent support was encouraged. Attention should be given to replicating intervention studies for the purpose of establishing treatments of choice and confirming efficacy and generalizability. The problem appears to be widespread, suggesting that intervention studies should be tested on a large scale as part of commonly available service systems. Group stress management interventions may not be sufficient for some parents who are likely to require more individualized and intensive forms of treatment (Singer et al., 1989).

Using a planned comparison approach, I found significantly higher effect sizes in the two studies that disaggregated data on parents of children with autism. Higher levels of depression associated with this condition again raise questions about the need for specific services targeted at this population as well as further research to identify more specifically why there is more distress in this group. The finding that parents of adult children had a significantly lower effect size than did parents of children in early and middle childhood is consistent with research suggesting that maternal distress gradually decreases over time (Glidden & Schoolcraft, 2003). It is important to note that there was only one available comparative study with a group of parents of adults, and there were no studies of adolescents that could be included in the planned comparison. The possibility of an age-related trend should be investigated.

The quality of these studies was highly variable. In the rating system I used to develop a quality index for each study, the highest possible score was a 13. The average quality rating across the 18 studies was 8.9 (range = 3 to 12). Authors of several studies failed to report important demographic information that should be routinely included in future research along with means and SDs of dependent variables. Only two studies drew from large population samples (Seltzer et al., 2001; Witt et al., 2003), whereas the others relied on samples of convenience. Although difficult, future investigators should draw from population samples that avoid the problems of external validity raised by use of convenience samples.

Further, future researchers should compare depression in mothers from minority ethnic and linguistic groups, including recent immigrants. With the exception of a few studies, the subjects were predominantly middle-class White women. Authors who did use more diverse samples did not disaggregate the data by ethnicity. Blacher et al. (1997) found high levels of depression in Latina mothers of children with developmental disabilities, suggesting that further study of minority populations, particularly recent immigrants, is warranted.

Finally, fully 71% of mothers of children with developmental disabilities in the studies analyzed in this paper did not experience problems with depression. It is remarkable how little is known about the large majority of mothers of children with developmental disabilities who are faring as well, on average, as the rest of the women of child-bearing age in the United States. The finding that roughly 29% of mothers experience elevated levels of depressive symptoms should serve as a caution against blanket assumptions about negative impacts of parenting a child with disabilities while making it clear that a serious problem needs to be addressed.

Table 2.

Continued

Continued
Continued

References

References
Abidin
,
R. R.
1995
.
Parenting Stress Index professional manual (3rd ed).
Odessa, FL: Psychological Assessment Resources
.
Amaral
,
R.
2003
.
How do children with developmental disabilities impact their parents' parent satisfaction, self-esteem, symptoms of stress, ways of coping, marital satisfaction and family support.
Dissertation Abstracts International
64
/
04
:
1183
. (
UMI No. 3089126
).
*Barakat
,
L. P.
and
J. A.
Linney
.
1992
.
Children with physical handicaps and their mothers: The interaction of social support, maternal adjustment, and child adjustment.
Journal of Pediatric Psychology
17
:
725
739
.
Beck
,
A. T.
,
R. A.
Steer
, and
M. G.
Garbin
.
1988
.
Psychometric properties of the Beck Depression Inventory: Twenty-five years of evaluation.
Clinical Psychology Review
8
:
77
100
.
*Beckman
,
P. J.
1991
.
Comparison of mothers' and fathers' perceptions of the effect of young children with disabilities.
American Journal on Mental Retardation
95
:
585
595
.
*Blacher
,
J.
,
S.
Lopez
,
J.
Shapiro
, and
J.
Fusco
.
1997
.
Contributions to depression in Latina mothers of with and without children with retardation: Implications for caregiving.
Family Relations
46
:
325
334
.
*Breslau
,
N.
and
G. C.
Davis
.
1986
.
Chronic stress and major depression.
Archives of General Psychiatry
43
:
309
314
.
*Breslau
,
N.
,
K. S.
Staruch
, and
E. A.
Mortimer
.
1982
.
Psychological distress in mothers of disabled children.
American Journal of Diseases of Children
136
:
682
686
.
Bristol
,
M. M.
,
J. J.
Gallagher
, and
K. D.
Holt
.
1993
.
Maternal depressive symptoms in autism: Response to psychoeducational intervention.
Rehabilitation Psychology
38
:
3
10
.
*Bristol
,
M. M.
,
J. J.
Gallagher
, and
E.
Schopler
.
1988
.
Mothers and fathers of young developmentally disabled and nondisabled boys: Adaptation and spousal support.
Developmental Psychology
24
:
441
451
.
*Cappelli
,
M.
1990
.
Marital interaction of couples with children with spina bifida: A case-control study (UMI No. NN68849).
Dissertation Abstracts International
53
/
03
:
4366
.
Cohen
,
J.
1988
.
Statistical power analysis for the behavioral sciences (2nd ed).
New York: Academic Press
.
Cooper
,
H.
1989
.
Integrating research: A guide for literature reviews.
Newbury Park, CA: Sage
.
Cummings
,
S. T.
,
H. C.
Bayley
, and
H. E.
Rie
.
1966
.
Effects of the child's deficiency on the mother: A study of mentally retarded, chronically ill, and neurotic children.
American Journal of Orthopsychiatry
36
:
395
408
.
Derogatis
,
L.
1977
.
SCL-R-90 Version: Manual-I.
Baltimore: Johns Hopkins University
.
Derogatis
,
L. R.
and
N.
Melisaratos
.
1983
.
The Brief Symptom Inventory: An introductory report.
Psychological Medicine
13
:
595
605
.
Fergusen
,
P.
2001
.
Mapping the family: Disability studies and the exploration of parental response to disabilities.
In G. Albrecht, K. Seelman, & M. Bury (Eds.), Handbook of disability studies (pp. 373–395). Thousand Oaks, CA: Sage
.
Gath
,
A.
1977
.
The impact of an abnormal child upon the parents.
British Journal of Psychiatry
130
:
405
410
.
Glidden
,
L. M.
,
M. J.
Kiphart
,
J. C.
Willoughby
, and
B. A.
Bush
.
1993
.
Family functioning when rearing children with developmental disabilities.
In A. P. Turnbull, J. M. Patterson, S. K. Behr, D. L. Murphy, J. G. Marquis, & M. Blue-Banning (Eds.), Cognitive coping, families, and disability: Participatory research in action (pp. 183–194). Baltimore: Brookes
.
Glidden
,
L. M.
and
S. A.
Schoolcraft
.
2003
.
Depression: Its trajectory and correlates in mothers rearing children with intellectual disabilities.
Journal of Intellectual Disability Research. Special Issue on Family Research
47
/
4–5
:
250
263
.
*Gowen
,
J. W.
,
N.
Johnson-Martin
,
B. D.
Goldman
, and
M.
Appelbaum
.
1989
.
Feelings of depression and parenting competence of mothers of handicapped and nonhandicapped infants: A longitudinal study.
American Journal on Mental Retardation
94
:
259
271
.
*Harris
,
V. S.
and
S. M.
McHale
.
1989
.
Family life problems, daily caregiving activities, and the psychological well-being of mothers of mentally retarded children.
American Journal on Mental Retardation
94
:
231
239
.
Hastings
,
R. P.
and
A.
Beck
.
2004
.
Practitioner review: Stress intervention for parents of children with intellectual disabilities.
Journal of Child Psychology and Psychiatry
45
:
1338
1349
.
Hastings
,
R. P.
and
H. M.
Taunt
.
2002
.
Positive perceptions in families of children with developmental disabilities.
American Journal on Mental Retardation
107
:
116
127
.
Hawkins
,
N. E.
,
G. H. S.
Singer
, and
C.
Nixon
.
1993
.
Short term behavioral counseling for families of persons with disabilities.
In G. H. S. Singer & L. E. Powers (Eds.), Families, disability and empowerment (pp. 317–342). Baltimore: Brookes
.
Hays
,
R. D.
,
K. B.
Wells
,
C.
Sherbourne
,
W.
Rogers
, and
K.
Spritzer
.
1995
.
Functioning and well-being outcomes of patients with depression compared with chronic general medical illnesses.
Archives of General Psychiatry
52
:
11
19
.
Hedges
,
L. V.
and
I.
Olkin
.
1985
.
Statistical methods for meta-analysis.
New York: Academic Press
.
*Kazak
,
A. E.
1987
.
Families with disabled children: Stress and social networks in three samples.
Journal of Abnormal Child Psychology
15
:
137
146
.
*Kazak
,
A. E.
and
R. S.
Marvin
.
1984
.
Differences, difficulties and adaptation: Stress and social networks in families with a handicapped child.
Family Relations: Journal of Applied Family and Child Studies
33
:
67
77
.
Kessler
,
R. C.
2003
.
Epidemiology of women and depression.
Journal of Affective Disorders
74
:
5
13
.
Kessler
,
R. C.
,
S.
Zhao
,
D. G.
Blazer
, and
M.
Swartz
.
1997
.
Prevalence, correlates, and course of minor depression and major depression in the national comorbidity survey.
Journal of Affective Disorders
45
/
1–2
:
19
30
.
Langer
,
T. S.
1962
.
A twenty-two item screening score of psychiatric symptoms indicating impairment.
Journal of Health and Social Behavior
3
:
269
276
.
Lin
,
N.
,
A.
Dean
, and
W.
Ensel
.
1986
.
Social support, life events, and depression.
San Diego: Academic Press
.
Lipsey
,
M. W.
1994
.
Identifying potentially interesting variables and analysis opportunities.
In H. Cooper & L. Hedges (Eds.), The handbook of research synthesis (pp. 111–124). New York: Russell Sage Foundation
.
*Miller
,
A. C.
,
R. M.
Gordon
,
R. J.
Daniele
, and
L.
Diller
.
1992
.
Stress, appraisal, and coping in mothers of disabled and nondisabled children.
Journal of Pediatric Psychology
17
:
587
605
.
Moes
,
D.
,
R.
Koegel
,
L.
Schreibman
, and
L. M.
Loos
.
1992
.
Stress profiles for mothers and fathers of children with autism.
Psychological Reports
71
:
1272
1274
.
National Institute of Child and Human Development Early Child Care Research Network (NICHDECCRN).
1999
.
Chronicity of maternal depressive symptoms, maternal sensitivity, and child functioning at 36 months.
Developmental Psychology
35
:
1297
1310
.
Nixon
,
C.
and
G. H. S.
Singer
.
1993
.
A group cognitive behavioral treatment for excessive parental self-blame and guilt.
American Journal on Mental Retardation
97
:
665
672
.
Pelchat
,
D.
,
J.
Bisson
,
N.
Ricard
,
M.
Perreault
, and
J. M.
Bouchard
.
1999
.
Longitudinal effects of an early family intervention programme on the adaptation of parents of children with a disability.
International Journal of Nursing
36
:
465
477
.
Radloff
,
L.
1977
.
The CES-D scale: A self-report depression scale for research in the general population.
Applied Psychological Measurement
1
:
385
401
.
Risdall
,
D.
and
G. H. S.
Singer
.
2004
.
Marital adjustment in parents of children with disabilities: A historical review and meta-analysis.
Research and Practice for Persons with Severe Disabilities
29
:
95
103
.
*Roach
,
M. A.
,
G. I.
Orsmond
, and
M. S.
Barratt
.
1999
.
Mothers and fathers of children with Down syndrome: Parental stress and involvement in child care.
American Journal on Mental Retardation
104
:
422
436
.
Rosenberg
,
M.
,
D. C.
Adams
, and
J.
Gurevitch
.
1997
.
MetaWin.
Sunderland, MA: Sinauer
.
Rosenthal
,
R.
,
R. L.
Rosnow
, and
D. B.
Rubin
.
2000
.
Contrasts and effect sizes in behavioral research: A correlational approach.
Cambridge: Cambridge University Press
.
Rossiter
,
L.
and
D.
Sharpe
.
2001
.
The siblings of individuals with mental retardation: A quantitative integration of the literature.
Journal of Child and Family Studies
10
:
65
84
.
*Scott
,
B. S.
,
L.
Atkinson
,
H. L.
Minton
, and
T.
Bowman
.
1997
.
Psychological distress of parents of infants with Down syndrome.
American Journal on Mental Retardation
102
:
161
171
.
*Seltzer
,
M. M.
,
J. S.
Greenberg
,
F.
Floyd
,
Y.
Pettee
, and
J.
Hong
.
1998, November
.
Life course impacts of parenting a child with a disability.
Paper presented at the Annual Scientific Meeting of the Gerontological Society of America, Philadelphia
.
Shadish
,
W. R.
and
C. K.
Haddock
.
1994
.
Combining estimates of effect size.
In H. Cooper & L. Hedges (Eds.), The handbook of research synthesis. New York: Russell Sage Foundation
.
Singer
,
G. H. S.
,
A.
Gang
,
L.
Powers
,
E.
Cooley
,
C.
Nixon
,
K.
Kerns
, and
D.
Williams
.
1994
.
A comparison of two psychosocial interventions for parents of children with acquired brain injury: An exploratory study.
Journal of Head Trauma Rehabilitation
9
:
38
49
.
Singer
,
G. H. S.
and
L. K.
Irvin
.
1991
.
Supporting families of persons with disabilities: Emerging findings, practices, and questions.
In L. H. Meyer, C. A. Peck, & L. Brown (Eds.), Critical issues in the lives of people with severe disabilities (pp. 271–312). Baltimore: Brookes
.
Singer
,
G. H. S.
,
L. K.
Irvin
, and
N.
Hawkins
.
1988
.
Stress management training of parents of children with severe handicaps.
Mental Retardation
26
:
269
277
.
Singer
,
G. H. S.
,
L. K.
Irvin
,
B.
Irvine
,
N. J.
Hawkins
, and
E.
Cooley
.
1989
.
Evaluation of community based support services for families of persons with developmental disabilities.
Journal of the Association for the Severely Handicapped
14
:
312
323
.
Singer
,
G. H. S.
,
L. K.
Irvin
,
B.
Irvine
,
N. E.
Hawkins
,
J.
Hergreness
, and
R.
Jackson
.
1993
.
Helping families adapt positively to disability: Overcoming demoralization through community supports.
In G. H. S. Singer & L. Powers (Eds.), Families, disability, and empowerment (pp. 67–84). Baltimore: Brookes
.
Turnbull
,
A. P.
,
V.
Turbiville
, and
H. R.
Turnbull
.
2000
.
Evolution of family professional partnerships: Collective empowerment as the model for the early twenty-first century.
In J. P. Shonkoff & S. J. Meisels (Eds.), Handbook of early childhood intervention (pp. 630–650). New York: Cambridge University Press
.
*Walker
,
L. S.
,
J. A.
Ortiz-Valdes
, and
J. R.
Newbrough
.
1989
.
The role of maternal employment and depression in the psychological adjustment of chronically ill, mentally retarded, and well children.
Journal of Pediatric Psychology
4
:
357
370
.
*Witt
,
W. P.
,
A. W.
Riley
, and
M. J.
Coiro
.
2003
.
Childhood functional status, family stressors, and psychological adjustment among school-aged children with disabilities in the United States.
Archives of Pediatric Adolescent Medicine
157
:
687
695
.
*Wolf
,
L. C.
,
S.
Noh
,
S. N.
Fisman
, and
M.
Speechley
.
1989
.
Brief report: Psychological effects of parenting stress on parents of autistic children.
Journal of Autism and Developmental Disorders
19
:
157
166
.
Wolfensberger
,
W.
1967
.
Counseling the parents of the retarded.
In A. A. Baumeister (Ed.), Mental retardation: Appraisal, education, and rehabilitation (pp. 329–400). Chicago: Aldine
.
Zimmer
,
K. P.
and
C. S.
Minkovitz
.
2003
.
Maternal depression: An old problem that merits increased recognition by child healthcare practitioners.
Current Opinion in Pediatrics
15
:
636
640
.

Requests for reprints should be sent to George Singer; University of California, Santa Barbara, Department of Education, Santa Barbara, CA 93106. singer@education.ucsb.edu

References with asterisks at the beginning are the primary studies synthesized in this meta-analysis.