The burden of caring for adults with depression and suicidal ideation in five large European countries: analysis from the 2020 National Health and Wellness Survey

Respondents included in the current study were aged ≥18 years, could read and write in the primary language of the country in which the study was conducted, and provided informed consent. Using the overall 2020 Europe NHWS sample of 62,319 respondents, we identified caregivers of adults with unipolar depression based on affirmative response to the survey query “Are you currently caring for an adult relative with any of the following conditions = depression?” Excluded were those who reported caring for an adult relative with bipolar disease or schizophrenia, since these patients have different underlying psychopathological factors even though their patterns of depressive episodes may be similar to adults with unipolar depression [29, 30]. The final sample cohort was grouped according to their response to “Has the person you care for with depression had thoughts that they would be better off dead or of hurting themselves in some way or have planned or attempted to end their life /commit suicide in the past six months?” into two categories: caregivers of persons with unipolar depression and SI (positive response) and caregivers of adults with unipolar depression but no SI (negative response) (see Table 1).

The study compared the characteristics and outcomes of caregivers of unipolar depression + SI (CAUD-SI) and caregivers of unipolar depression + no SI (CAUD-nSI). Descriptive statistics included means and SDs for continuous variables and frequencies and percentages for categorical variables. Missing data is minimized by allowing respondents the option of ‘decline to answer’ or ‘prefer not to answer’ for items asking about potentially sensitive topics, such as weight or income. The resultant small amount of missing data are expected to be missing at random (MAR).

Bivariate analyses evaluated differences between caregiver groups using ANOVA for continuous variable and chi-square tests for categorical variables. Multivariate analyses were employed to examine the association between caregiver group and outcome after adjusting for potential confounding factors. Model choices were determined according to the distribution of each outcome. Linear mixed models with gaussian distribution were used for health outcomes (EQ-5D-5L, SF-6D, and SF-12v2 scores) and generalized linear mixed models with negative binomial distribution were used for economic outcomes (WPAI metrics, healthcare resource utilization) after controlling for the confounding effect of sociodemographic (age, education, employment status) and health status (smoking status, exercise, CCI) factors using backward models. Adjusted means, standard errors (SE), regression coefficients, p-values, and 95% confidence intervals were calculated for all the dependent variables. P < 0.05 was considered as statistically significant.

All statistical analyses were conducted using SPSS v23.0, SAS v9.4, and/or R v3.6.0 or higher.

A total of 62,319 respondents (France N = 15,152; Germany N = 15,020; Italy N = 10,026; Spain N = 7074; UK N = 15,047) were included in the study. Of these respondents, 554 (0.89%) were identified as CAUD-SI whereas 837 respondents (1.34%) were identified as CAUD-nSI (Fig. 1).

Unadjusted comparisons showed that all caregivers, regardless of subgroup, were more often women (CAUD-SI = 53.8% vs. CAUD-nSI = 58.2%, p = 0.106), married (CAUD-SI = 65.5% vs. CAUD-nSI = 64.9%, p = 0.831), and reported lower annual household income (CAUD-SI = 76.9% vs. CAUD-nSI = 76.6%, p = 0.906) (Table 1). Differences between groups were observed with CAUD-SI respondents being younger (40.7 ± 15.2 years vs. 43.5 ± 15.2 years, p < 0.001), more educated (university degree: 37.8% vs. 32.5%, p = 0.040) and more likely to be employed (69.0% vs. 61.3%, p = 0.014) compared to CAUD-nSI respondents, respectively. CAUD-SI respondents were more likely to smoke (44.0% vs. 33.0%, overall p < 0.001) and exercise (76.0% vs. 68.2%, p = 0.002) and had a higher average CCI score (0.71 vs. 0.53, p = 0.012).

CAUD-SI respondents reported higher mental health burden in terms of self-reported medical diagnosis (SR-MD) of depression (47.8% vs. 37.5%, overall distribution p < 0.001) and sleep problems (14.8% vs. 10.9%, overall distribution p < 0.05) (Table 1). Additionally, differences in PHQ-9 scores were noted to be significant between CAUD-SI and CAUD-nSI (p < 0.001), with higher proportion of CAUD-SI respondents reporting moderate, moderately severe, and severe depression.

Unadjusted analyses of outcomes related to caregiving for depression with and without SI indicated substantially higher burden among CAUD-SI respondents compared with CAUD-nSI (Table S1).

CAUD-SI respondents reported significantly lower scores for each domain – physical functioning (44.9 vs. 48.7, p < 0.001), bodily pain (41.8 vs. 44.3, p < 0.001), physical role functioning (40.4 vs. 43.7, p < 0.001), emotional role functioning (34.2 vs 38.7, p < 0.001), and social role functioning (38.1 vs. 41.5, p < 0.001), as well as lower PCS (46.2 vs. 48.5, p < 0.001) and MCS scores (38.4 vs. 40.3, p < 0.001).

CAUD-SI respondents also reported lower health status as measured by EQ-5D-5L (0.63 vs. 0.73, p < 0.001) and SF-6D health utilities (0.60 vs. 0.64, p < 0.001) than CAUD-nSI.

In terms of WPAI, CAUD-SI respondents reported higher levels of absenteeism (24.1 vs. 14.6, p < 0.001), presenteeism (48.4 vs. 34.7, p < 0.001), overall work productivity loss (54.7 vs. 39.6, p < 0.001), and activity impairment (49.0 vs. 39.5, p < 0.001) compared with CAUD-nSI respondents.

When assessing HRU, a higher percentage of CAUD-SI respondents reported visit to a psychologists or therapists in the past six months (12.3% vs. 8.5%, p = 0.021), however no difference was noted for visits to a psychiatrist (11.2% vs. 8.6%, p = 0.109) compared with CAUD-nSI respondents. Further, significantly higher traditional provider visits (7.9 vs. 6.5, p = 0.005), ER visits (1.48 vs. 0.59, p < 0.001), and hospitalizations (0.83 vs. 0.32, p < 0.001) were observed in CAUD-nSI respondents than CAUD-nSI.

Multivariable models adjusted after controlling for covariates showed that CAUD-SI versus CAUD-nSI had substantially lower adjusted means for MCS (37.5 vs. 38.9, p = 0.007) and PCS (42.7 vs. 45.0, p < 0.001) (Table S2, Fig. 2). Specific domains of the SF-12v2 with statistically significant marginal mean differences between groups included physical functioning (difference = 3.5), bodily pain (difference = 2.4), physical role functioning (difference = 3.0), emotional role functioning (difference = 4.0), and social role functioning (difference = 3.0).

In the adjusted model, the CAUD-SI group had significantly poorer health status compared to CAUD-nSI for both the SF-6D (0.57 vs 0.60, p < 0.001) and EQ-5D-5L (0.58 vs 0.66, p < 001). The MCID of > 0.07 points was observed for the EQ-5D-5L (Table S2, Fig. 3).

Comparison of WPAI in adjusted models showed no significant differences in absenteeism, presenteeism, or overall work impairment between CAUD-SI and CAUD-nSI respondents (p > 0.05), but differences in activity impairment were observed between the two groups (p = 0.026). (Table S2, Fig. 4).

CAUD-SI respondents reported higher estimated marginal means for provider visits (10.0 vs. 7.9, p < 0.001), ER visits (1.49 vs. 0.73, p < 0.001), and hospitalizations (1.03 vs. 0.52, p < 0.001) compared with CAUD-nSI respondents (Table S2, Fig. 5).

The strengths of the study include self-reported outcomes that are uniformly derived across all five European countries. The sample of caregivers is drawn from a representative sample of the general population of each country in which the study was conducted, which notably differs from many caregiver studies that use convenience samples [27, 43]. Recall bias is minimized since most survey questions are based on current practice and attitude. Also, missing data is minimized by including ‘don’t know’ or ‘decline to answer’ as options for items regarding potentially sensitive topics, such as weight or income.

The present study is not without limitations. The NHWS is a panel-based survey and may not reach segments of populations including those with limited internet access, disabilities, or lack of familiarity with online surveys. Further, caregivers who can read and write in the language of their resident country were included in the NHWS, thereby excluding potential caregivers who do not meet this criteria; although we would assume this potential bias to be minimal as most caregivers of their adult relatives are likely fluent in their country of residence’s language. The study relies on self-reported data and therefore diagnosis of depression and other relevant diagnoses (e.g., those used in the comorbidity index) and report of healthcare visits cannot be confirmed. Also, since only measured variables were accounted for in regression analysis, there is possibility of groups differing on unmeasured variables that may have an impact on outcomes. Finally, the NHWS does not capture caregiving for a patient based on patient characteristics such as severity, medication use, or duration of their disease. This in turn, is likely to affect the assessment of burden as caregiving is impacted by degree of severity of disease and depression is a complex disorder that is modulated by several factors including severity, duration and overall management of disease.