High prevalence of depression in parents of children with Type 1 diabetes in a meta-analysis of data from five continents

2.1 Selection of studies

Papers from peer-reviewed, English-language journals that were published between 1980 and May 2022 were considered for inclusion. This timeframe was chosen due to the publication of the Third Edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-III) in 1980, which first introduced depression as a unitary category.³⁵ It was hoped that this would reduce heterogeneity in measures used within depression studies following this diagnostic criterion. Relevant studies were identified through systemic searches in three electronic databases: MEDLINE, PsycINFO and CINAHL. Animal studies were excluded from the searches. The search was conducted on 16th May 2022. The Cochrane database and PROSPERO register were searched to ensure no similar reviews were either in progress or had been published. This review was registered on PROSPERO (ID Number CRD42022317995).

The following search terms and combinations were used: (Parent* OR carer* OR caregiver* OR ‘care giver’ OR mother* OR father* OR Maternal* OR Paternal*) AND (Depress* OR ‘mood disorder*’ OR ‘low mood’ OR ‘dysthymic disorder’) AND Diabet*. The following MeSH terms were also applied to the searches: (MeSH Diabetes Mellitus, Type 1+) AND (MeSH Depression).

2.2 Inclusion and exclusion criteria

Screening and selection of studies were conducted by the first author (HR). To ensure relevant papers were included in the meta-analysis, strict inclusion and exclusion criteria were applied. Studies were included if they reported prevalence rates of depression in parents of children aged between 0 and 18 years with T1D. The term ‘parents’ is used for the child's primary caregiver and the terms ‘mother’, ‘father’, or ‘other parent descriptor’ are used to describe gender differences. Studies were excluded if: they did not use a validated or reliable measure of depressive symptoms; prevalence rates of depression were not reported; if the children died before parental depression was assessed; if the aim of the study was to investigate the efficacy of treatment; or if the sample used were biased (e.g., only recruiting parents or children with depression). Studies were included if they were: cross-sectional design studies; case-controlled studies; and longitudinal studies. Studies were excluded if they were: randomised controlled trials; treatment or intervention studies; review articles; systematic reviews; meta-analyses; theses and dissertations; book chapters; purely qualitative research; single case reviews; or case studies.

2.3 Data extracted from each study

The first author (HR) screened all studies and extracted information using a database. Several study variables were examined: author; year of publication; country; setting; inclusion and exclusion criteria; data collection method; recruitment method; and sample size. The following participant data was extracted: age, gender and ethnicity of parents; assessment methods and measures; and the number of parents meeting the cut-off for depression. The majority of studies only collected data at one timepoint (k = 15). However, for the remaining studies (k = 3) that reported depression prevalence at multiple time points, only baseline prevalence was extracted. Data extracted from each study is displayed in Table 1. The extracted data for all studies were reviewed by an independent researcher (AH) to reduce the likelihood of error.³⁶ Any queries were discussed until consensus was reached. If consensus could not be reached by the two authors (HR and AH), a final decision was made by the senior researcher (RM-S).

The majority of studies reported levels of depression according to published cut-offs for each measure. For example, no depression, and mild, moderate and severe depression. In the current study, severity of depression was then divided into two categories: parents who scored in the mild range of depression and above (mild plus); and parents who scored in the moderate range of depression and above (moderate plus) on the measures. Two studies only categorised parents as either below or above the clinical cut-off for depression and placed them in ‘moderate plus’ category. One study used a category defined as ‘doubtful cases’, and these were placed into the ‘mild plus’ category in the current study.

The studies included in this meta-analysis collected data at various timepoints following a diagnosis of T1D, ranging from the acute period to 5 years post-diagnosis (see Table 1).

2.4 Risk of bias

The quality of each study was analysed by two researchers (HR and AH). An adapted version of a risk-of-bias tool used in a recent meta-analysis of post-traumatic stress disorder (PTSD) prevalence was used,³⁷ using common quality assessment questions developed by Munn et al.³⁸ The risk-of-bias tool included six questions that assessed quality according to the description of sample characteristics, non-response rates and reasons provided, representativeness, recruitment procedures and reporting of inclusion and exclusion criteria. As prevalence was the only outcome extracted from each study, the type of analyses used were not assessed for quality. A qualitative descriptor of risk-of-bias was applied to each study (Low, Medium and High) and the scoring was adapted to reflect the number of questions assessing quality (9–12 = low risk, 5–8 = medium risk, 0–4 = high risk). All studies were inter-rated by the two researchers (HR and AH). Analysis was then conducted to assess inter-rater reliability using MAVIS (version 1.1.3).

2.5 Statistical analysis

The meta-analyses were conducted using R (version 4.1.3), using the metafor package (version 3.8-1).³⁹ The prevalence of parents with depression was extracted from each paper. Parents were classed as having depression if they met cut-off on a validated depression measure. To account for likely between-study heterogeneity, a random effects model was used, as it provides a more conservative 95% CI around the estimate of prevalence.

The estimates of prevalence underwent an arcsine square-root (angular) transformation to ensure that CIs did not fall below zero for samples where the prevalence estimate was low.⁴⁰ The results were then back transformed to enable ease of interpretation.

The heterogeneity of studies was assessed by visual inspection of forest plots, in addition to conducting a Cochran's Q test.⁴¹ The Cochran's Q test indicates whether heterogeneity is significant between the studies. The I² statistic was used to determine the percentage of total variation in sample estimates that is due to between-study heterogeneity.⁴² I² between 30% and 60% indicates moderate heterogeneity, 50%–90% represents substantial heterogeneity and 75% and above indicates considerable heterogeneity.⁴³ Prediction intervals (PI) were also reported, enabling clinical interpretation of the heterogeneity.⁴⁴ A 95% PI estimates where the true effects are to be expected for 95% of similar studies conducted in the future.⁴⁵

Publication bias was assessed using funnel plots.⁴⁶ However, due to the lack of assessment of clinical significance in prevalence studies, the risk-of-bias in levels of acceptance to journals is reduced.⁴⁷ Funnel plot asymmetry was tested using Egger's test.⁴⁸

Sensitivity analyses were conducted to assess the impact of risk-of-bias on the pooled prevalence. This was achieved by repeating the meta-analysis, excluding samples where there was a high risk of bias. A meta-analytic regression was used to test for any clinically significant differences in the sensitivity analysis.

Moderator analysis was not conducted for certain variables due to the homogeneity of the variables, for example the type of depression measure used (all used self-report measures) and the income status of the country (majority recruited from high income countries). Moderator analysis was conducted to explore differences in prevalence rates between mothers and fathers.

In clinical research, researchers may choose not to write up and publish studies with uninteresting findings, for example those with smaller or nonstatistically significant effect sizes.⁴⁹ However, this can lead to publication bias. With prevalence research, it is unclear whether researchers would be more likely to publish studies reporting either low or high prevalence rates, and therefore less is known about the risk of publication bias within this research. Nevertheless, publication bias was assessed in the current review for robustness.

3.1 Characteristics of the studies

The characteristics of the included studies can be found in Table 1. Participants ranged in age from 20 to 61. The estimated mean age of parents across all studies was 40.4. Seven studies were not included in this overall mean age calculation as they either did not report mean age or they reported it according to gender or severity of depression rather than the total sample. A range of inclusion and exclusion criteria were applied across the studies. Participants were frequently excluded if their child had another serious mental or physical comorbidity and if the diagnosis of T1D was less than 6 months. To assess depression prevalence in their samples, six studies used the Centre for Epidemiologic Studies Depression Scale (CES-D)⁵⁰; three studies used the Beck's Depression Inventory (BDI)⁵¹; three studies used the Patient Health Questionnaire (PHQ-9)⁵²; four studies used the Hospital Anxiety and Depression Scale (HADS)⁵³; and two studies used the Hamilton Depression Rating Scale (HDRS).⁵⁴

3.2 Risk-of bias assessment

Fourteen studies were deemed to be at medium risk-of-bias and four were deemed to be at low risk-of-bias. No studies were deemed to be at high risk-of-bias. Inter-rater reliability was assessed for all ratings on all studies (n = 18) by the two raters (HR, AH). They achieved an intra-class correlation of 0.68 for risk-of-bias, indicating a good correlation on all items [interclass correlation = 0.68, 95% confidence intervals (CI) 33.5–86.6].⁵⁵ This suggests that the quality assessment tool was robust, with good inter-rater reliability.

3.3 Depression prevalence

The pooled prevalence of depression estimates and heterogeneity statistics for all studies can be seen in Table 2. A total of 18 studies reported prevalence rates, which resulted in a pooled prevalence of parental depression following their child's diagnosis of T1D. However, not all 18 studies categorised depression into mild and moderate or reported depression prevalence in mothers and fathers separately. Therefore, the number of studies included in each meta-analysis varies. Please refer to Table 2 for a breakdown of the studies included in each meta-analysis.

Six meta-analyses were conducted: total participants scoring in the moderate depression and above range (moderate plus); total number of mothers scoring in the moderate depression and above range; total number of fathers scoring in the moderate depression and above range; total participants scoring in the mild depression and above range (mild plus); total number of mothers scoring in the mild depression and above range; total number of fathers scoring in the mild depression and above range.

3.3.1 Moderate plus depression

The overall pooled prevalence was 18.4% of the parents from the studies (k = 17) who had a child diagnosed with T1D developed depression in the moderate and above range (see Figure 2). For mothers, the pooled prevalence was 17.3% (k = 12). The degree of heterogeneity for these two meta-analyses was considerably high. For fathers, the pooled prevalence was 9% (k = 6), with a low degree of heterogeneity.

3.3.2 Mild plus depression

This overall pooled prevalence was 32.7% of the parents from the studies (k = 8) who had a child diagnosed with T1D developed depression in the mild and above range (see Figure 3). For mothers, the pooled prevalence was 29.4% (k = 4). The degree of heterogeneity for these two meta-analyses was considerably high. For fathers, the pooled prevalence was 13.6% (k = 2), with a low degree of heterogeneity.

3.4 Sensitivity analysis

On visual inspection of the forest plot, the study of Streisand et al. appeared to be an outlier.⁵ The two meta-analyses estimating the prevalence of mild (mild plus) and moderate (moderate plus) depression and above in the total sample were run again, excluding this study, to analyse the impact of this study on the overall pooled prevalence of depression. Streisand et al.⁶ did not report depression prevalence for mothers and fathers separately. Therefore, sensitivity analysis was only conducted for the total sample. For moderate plus, the prevalence of depression reduced to 16.2%; for mild plus, the prevalence of depression reduced to 27.4%. This sensitivity analysis showed that the inclusion of the Streisand et al. study resulted in little change in the pooled prevalence of depression, and it was therefore deemed appropriate to include it in the analysis.

Sensitivity analysis was also conducted, excluding studies that recruited parents during the acute stage of their child receiving a T1D diagnosis and one study that did not report this information (k = 6; see Table S1). The acute stage refers to studies that included data from the time of diagnosis to 1-year post-diagnosis. The remaining studies that recruited parents at least 6 months post-diagnosis or who had a mean diabetes duration of at least 4.5 years were pooled, and a meta-analysis was conducted (k = 12). For moderate plus, the prevalence of depression reduced only slightly to 16.1%; for mild plus, the prevalence of depression reduced slightly to 27%. This sensitivity analysis showed that the inclusion of studies that recruited parents around the time of receiving a diagnosis of T1D resulted in little change in the pooled prevalence of depression.

3.5 Moderator analysis

To consider whether parent sex moderated the prevalence of moderate plus depression, where possible, single-sex samples were identified and included in a moderator analysis. This analysis comprised of 18 samples drawn from 12 studies (12 mother samples [n = 1106] and 6 father samples [n = 199]). This analysis showed no statistically significant difference in prevalence as a function of gender (Q(df = 1) = 3.73, p = 0.054; mothers pooled estimate = 17.3%, fathers pooled estimate = 9.0%).

3.6 Publication bias

Publication bias was only reported for categories with 10 studies or more⁵⁶: moderate plus – total sample; and moderate plus – mothers only.

For the moderate plus threshold, including all studies, a visual inspection of the funnel plot suggested that the distribution of studies was symmetrical (k = 17). This was confirmed by Egger's test, which was non-significant (p = 0.67). The trim-and-fill analysis did not highlight null or weaker studies as missing, indicating little to no publication bias.

For the moderate plus threshold including only data for mothers, visual inspection of the funnel plot suggested that the distribution of studies was symmetrical (k = 14). This was confirmed by Egger's test, which was non-significant (p = 0.41). The trim-and-fill analysis highlighted a lack of studies on the left side of the funnel plot, indicating a small degree of potential publication bias; including two additional estimated studies, which reduced the pooled prevalence to 15.5% However, this should be interpreted with caution due to the small number of studies.⁵⁶

4.1 Comparison to prevalence of depression in the general population

The World Health Organisation (WHO) estimates that approximately 5% of adults are affected by depression worldwide.⁵⁷ The prevalence of depression is higher in women compared to men, with a global prevalence of 5.5% and 3.2%, respectively, in 2010, among those aged 3 years and above.⁵⁸ Between 6% and 17% of women experience depression in their lifetime, with elevated rates in mothers.⁵⁹ Using the PHQ-9, the National Health and Nutrition Examination Survey (NHANES) found that 8.1% of American adults aged 20 and over had depression in a given 2-week period.⁶⁰ Women (10.4%) were almost twice as likely as men (5.5%) to score above the cut-off for depression. The prevalence of depression in parents of children with T1D in the current study was higher than in the general population; however, the finding that the prevalence is higher in women relative to men is comparable.

4.2 Comparison to prevalence of depression in parents of healthy children

One study analysed depressive symptoms in adults across numerous parent-child relationships, revealing that parents with children currently residing within the home may have lower psychological well-being when compared with those without children.⁶¹ Furthermore, it is important to consider the prevalence of depression among parents of healthy children in order to make comparisons to the results of the current study. Cohn et al. compared rates of depression in parents of children with and without chronic illness, defined as those suffering from a physical, developmental, behavioural or emotional condition lasting at least 3 months. Data were collected from nine countries. The paper looked only at rates of parents scoring above the clinical cut-off for depression and, in contrast with the current paper, did not categorise into ‘mild plus’ and ‘moderate plus’ depression. The paper looked only at rates of parents scoring above the clinical cut-off for depression and, in contrast with the current paper, did not categorise into ‘mild plus’ and ‘moderate plus’ depression. Parents of affected children had higher rates of depression (35%) compared to parents of healthy children (19%).³⁴ Similarly to the current study, this study also had a paucity of data on fathers compared with mothers. This highlights that the prevalence rate of mild and above depression found in the current study is considerably higher than rates in parents of healthy children. However, limited studies in recent years have reported on the prevalence of depression in parents of healthy children and adolescents, and further research is needed to support this claim. Moreover, it appears little is known about the relationship between parenting and depression. Despite these gaps in the literature, our findings support previous studies demonstrating that parents of children with T1D have elevated rates of depression.^{5, 6, 24}

4.3 Clinical implications and suggestions for future research

This meta-analysis suggests that depression is common in a significant minority of parents of children with T1D, and therefore should be routinely assessed in clinical settings. It is recommended that this screening is conducted as early as possible following a child's diagnosis. Interventions should then also be offered to support these parents. According to recent guidelines from the National Institute for Health and Care Excellence (NICE), this could involve anything from guided self-help to individual cognitive behavioural therapy (CBT).⁶² Additionally, recommendations from the International Society for Paediatric and Adolescent Diabetes (ISPAD) state that family-based behavioural interventions should be offered, which could include goal setting, problem-solving and the use of behavioural contracts.⁶³ This could prevent severe depression developing in these parents and reduce the risk of complications in diabetes management for the child.

Studies did not report whether parents had a history of depression. Therefore, it is unclear whether parents had depression before their child's diagnosis or whether they developed it following the child's diagnosis. Future research could explicitly explore this. Furthermore, research could consider the prevalence of depression in parents before their child's diagnosis and how this is associated with the child's diabetes management. Research is needed to explore differences in the prevalence of depression among parents of children with T1D compared to those with healthy children. As previously discussed, parenting has been associated with lower psychological well-being. Therefore, it is unclear to what extent the high levels of depression found in the population of parents in the current study is a result of the child having T1D or the demands of parenting in general. There is a scarcity of data on the prevalence of depression in fathers of children with T1D. In the current study, of the 17 studies that reported the proportion of mothers, 12 reported that over 85% of the included participants were mothers. Further research is required to further explore the rates of depression in fathers of children with T1D. Finally, potential risk factors associated with the development of depression in parents of children with T1D have been explored. For example, systemic factors, disease management, traumatic memories, or appraisals. However, further research exploring the processes involved in the development of depression in this population is needed.

4.4 Strengths and limitations

This study is strengthened by minimal publication bias and by being the first meta-analysis, to our knowledge, to report the prevalence of depression in parents of children with T1D. However, several limitations should be considered. Although some level of heterogeneity is expected in a meta-analysis, the level of heterogeneity in the current study was considerable, making it challenging to make comparisons across studies. Heterogeneity may exist due to differences in measures, culture, age of children, study design, measures and cut-offs used, and whether measures were administered online or face-to-face. Furthermore, all studies were considered non-representative, with limited diversity in race and ethnicity, and recruitment within each study was restricted to a particular geographical locality. However, overall, this meta-analysis represents several geographical areas: Europe; Asia; North America; Northern Africa; and South America. Although the overall prevalence of depression was found to be high in the moderate plus category, the range of depression scores was wide, ranging from 17.9% to 46.6%, making it challenging to make inferences from these results. The majority of participants in the included studies were mothers. With a limited sample size, less is known about the prevalence of depression in fathers of children with T1D. Furthermore, with a limited sample size of fathers, the current sample may not be representative of all fathers of children with diabetes. Finally, all studies included in this review used self-report measures of depression, which are at risk of response bias.

The period 1980–2022 was chosen as the search criteria due to the implementation of the DSM-III. However, the treatment of T1D has developed in recent years, and therefore the experience of families is likely to be different, making it problematic to make comparisons between the oldest and youngest studies included in this meta-analysis. Fortunately, all studies except one included in the current paper were published after the year 2000. However, it is important to note this as a potential limitation, as T1D management is continually developing.

It is important to consider the variation in time since diagnosis as a potential limitation, which ranged from the acute period to 5 years post-diagnosis. The experience of families is likely to vary from the immediate aftermath of diagnosis compared to several years post-diagnosis, and therefore making comparisons between these studies is problematic. However, our sensitivity analysis showed little difference in the prevalence of depression when studies that recruited parents in the acute period following T1D diagnosis were excluded.