Some limitations should be mentioned. First, due to the cross-sectional design of our study, it is not possible to conclude for causal relationships between the identified predictors and the severity of depression. Longitudinal studies are needed to better explore this issue. Second, one should keep in mind that our data were collected during the first French global lockdown due to COVID-19. Previous reports have indicated increased depression rates in general population [
82] and non-clinical samples [
83‐
85] during lockdown. Here, 54.4% of caregivers presented a possible depressive disorder (CES-D total score ≥ 16), thus higher than the ones reported in caregivers of subjects with bipolar disorder or schizophrenia (respectively, 22–33% and 42%) outside lockdown periods [
66,
79‐
81]. This figure is in line with Chiu et al. (2022) [
78] reporting that 56% of family carers of older adults reported mild to severe depression between April and May 2020 (i.e., during lockdown in Hong Kong), a much higher prevalence than in the general population during the same period [
79]. A few studies have shown exacerbated depression and burden during COVID-19, as compared to pre-pandemic levels, among caregivers of people with dementia [
80] or with disability or cognitive decline [
81]. Importantly, one study showed that caregivers had a greater likelihood of somatic and mental health issues than non-caregivers during the first months of the pandemic, even after adjusting for preexisting health status [
82]. Undoubtedly, disruption of healthcare facilities and social restriction measures brought new challenges to caregivers facing an unexpected increase in responsibility and a greater experience of burden. In this context, psychological support interventions using digital solutions could be a useful format to improve the mental health of family caregivers [
83]. Third, our results should be interpreted in the context of the particular scale used to evaluate depression and depressive features. Common depression scales differ substantially in symptom content [
84]. Other depressive symptoms, not featured in the CES-D such as
somatic complaints, might be relevant and should be investigated in future research carried out in caregivers. Fourth, our sample size did not allow for introducing socio-demographics such as gender, age, or marital status (among other characteristics reported in Table
1) into estimated networks. Although we used the bootstrapping methodology introduced by Epskamp et al. (2018) [
35] for gaining insight into the accuracy of estimated parameters—resulting in meaningful and stable edges—, adding more nodes would have sizeably increased the number of estimated parameters, a threat to the accurate estimation of the models. However, caregiver’s factors including age, gender, educational level, income and patients’ factors such as age and clinical symptoms are likely to influence caregiving burden and depression levels on family caregivers [
85]. For instance, higher income would decrease financial problem and stress related to providing care for ill family member [
86]. Taking these covariates into account might impact the network structure of caregiving dimensions and depressive symptomatology estimated in the present study. Therefore, larger studies using a similar (network) approach are warranted to better characterize the inter-relationships between caregiving experiences, health-related outcomes including depression,
and covariates. Finally, 83.9% of the caregivers included in the present study were members of family associations. Therefore, most of participants may have benefited from peer support and/or psychoeducational resources which are associated with lower depression and burden scores [
73]. Studies assessing depression and burden in samples of caregivers who have never benefited from any caregiver interventions are needed to assess whether they are a more vulnerable subgroup. In this regard, surveys focusing on caregivers in early intervention services may be helpful.