Predictors of depression in young carers: a population based longitudinal study

ABSTRACT

Caregiving in adolescence, specifically when directed towards a parent, is associated with increased depression. This study examines whether young carers caring for their parents are more likely to be depressed than those caring for other family members and non-caring youth. Using data from two waves of the Growing Up in Ireland study (N = 3,312), the above associations were examined cross-sectionally (Aged 17; T1), and longitudinally (Aged 20; T2). The role of parent and peer relationship quality in predicting depression was also examined. Compared to non-carers, carers of parents were 3.67 times more likely to be depressed at T1, but not at T2. Moreover, young carers of parents who were depressed reported lower peer attachments, as well as lower quality relationships with their parents. Logistic regression analyses showed these relationships to be predictive of depression. This research is discussed in terms of supporting young carers from a family-based perspective.

KEYWORDS:

• Caregiving
• depression
• mental health
• relationship quality
• social support
• young carer

Introduction

Worldwide, there are an estimated 349 million people dependent on family members for their care, i.e. family caregivers (World Health Organisation, 2017). These numbers are increasing steadily year on year. In Ireland, for example, there has been a 50% increase in family caregivers since 2006, with 1 in 8 of those aged 15 and above reporting that they care for someone. Throughout Europe, as well as in other developed nations, an ageing population, alongside advanced healthcare prolonging the lives of ill and disabled individuals means children, adolescents, and young adults are becoming increasingly likely to take on caregiving roles (Lacey et al., 2022; Pickard, 2015). These young people are commonly referred to as young carers (<18 years) and young adult carers (18–25 years) (Becker et al., 2000; Chevrier et al., 2022). The hidden nature of youth caregiving makes it difficult to determine how many young people are providing unpaid care, but estimates suggest it could be between 2–8% (Leu & Becker, 2017), or even higher in some countries (Family Carers Ireland, 2020; Untas et al., 2022). This high prevalence, alongside evidence showing poorer health and well-being in adult caregivers (Gallagher et al., 2018) has sparked interest amongst researchers, practitioners, and policymakers into the immediate and long-term health outcomes for these young carer groups (Lacey et al., 2022).

Caring influences people’s lives in positive and negative ways (Joseph et al., 2009). Feelings of pride and competency (Family Carers Ireland, 2023) as well as enhanced maturity and compassion (Fives et al., 2010) have been identified in young carers. Simultaneously, mental health issues, such as depression, have been widely observed in young carer populations, with systematic reviews highlighting an array of poor mental health outcomes in this population (Fleitas Alfonzo et al., 2022; Lacey et al., 2022; Saragosa et al., 2022). However, the relationship between youth caregiving and mental health is poorly understood. With some young carers experiencing mainly positive effects, and others suffering from negative mental health effects, the heterogeneity of young carer populations is being increasingly recognized (Janes et al., 2022). Thus, researchers are becoming increasingly interested in how individual caregiving contextual factors influence mental health in young carers (Joseph et al., 2020).

An issue with the current literature is that there is a lack of theoretically driven research examining mental health outcomes in young carers (Vitaliano et al., 2004). Stress models, which examine how individual contextual factors interact with stress to influence mental health, have been adapted for use within the caregiver literature (Vitaliano et al., 1991, 2004). These models identify caregiving as a chronic stressor negatively impacting physical and mental health (Russo et al., 1995; Schulz & Sherwood, 2008; Vitaliano et al., 2003). However, the extent to which stress impacts health largely depends on individuals’ resources (e.g. social support) and vulnerabilities (e.g. psychosocial risk factors), with those with low resources and high vulnerabilities experiencing negative health effects like depression (Vitaliano et al., 1991). Given the high prevalence of depression in young carers (Fleitas Alfonzo et al., 2022; Lacey et al., 2022; Saragosa et al., 2022), this study aims to identify and examine young carers resources and vulnerabilities, and how these influence depression risk.

Social support is a broad concept that describes the social resources or networks that may be available to people that shows they are cared for, loved, and connected (Cohen & Syme, 1985). Social support is consistently associated with better mental health in carers of all ages (Dharampal & Ani, 2020; Gallagher et al., 2022b; Nap et al., 2020), including support from formal (professional services) and informal sources (family, friends, and neighbours). In adolescence, the two key informal sources of social support are parents and peers. Research suggests that the quality (i.e. positive, and negative aspects) of these relationships are integral to ensure the positive benefits of social support for psychological well-being (Farooqi, 2014; Teo et al., 2013; Uno et al., 2002). The broader literature on adolescence shows that high-quality parent and peer relationships lower depressive symptoms (Delgado et al., 2022; Johnson & Galambos, 2014; Withers et al., 2016). However, there are few studies examining the role of parent and peer relationships in predicting depression in young carers, despite research emphasizing the specific and unique support needs of young people providing care (van der Werf et al., 2022).

In general, in relation to the sources of social support, research shows that high quality parent and peer relationships lower depressive symptoms in adolescence (Delgado et al., 2022; Johnson & Galambos, 2014; Withers et al., 2016). However, it is unclear which aspects of these relationships are most important. Findings from the Orygen Adolescent Development Study show that expressions of aggression and fewer expressions of positivity (i.e. validation and affection) from parents were associated with depression in adolescence (Cano et al., 2016). Moreover, perceptions of parental rejection and negative parental feedback have been strongly associated with depressive symptoms in adolescence (Hankin et al., 2009; MacPhee & Andrews, 2006). Correspondingly, cross-sectional data suggests peer acceptance and dependability significantly reduce depressive symptoms in adolescence (Adedeji et al., 2022). In fact, some research suggests peer relationship quality overshadows that of parental relationships, in terms of its benefits for adolescent health (Delgado et al., 2022), albeit not all studies support these findings (Young et al., 2005). In fact, studies have highlighted that many young carers feel isolated from their peers (Family Carers Ireland, 2023; Fleitas Alfonzo et al., 2022), and report closer relationships with their parents than non-caring youth (Chikhradze et al., 2017). Considering these findings, this study will examine the influence of parent and/or peer relationships on depression risk in young carers.

Another contextual factor worthy of investigation is who the young person is caring for. For example, research has found that carer well-being outcomes can vary across care-recipient types (e.g. caring for a child vs caring for a parent) (Hammersmith & Lin, 2019). For example, caring for a spouse has been found to be more stressful compared to caring for other relatives (Pinquart & Sörensen, 2011). In the context of young carers, children of ill parents are more likely to experience symptoms of depression and anxiety compared to children whose parents are not ill (Korneluk & Lee, 1998). Further, a study on young carers found that, a greater risk of poorer mental health was evident when the care-recipient was a parent compared to other relatives (Pakenham & Cox, 2014a). As such, it is plausible that caring for a parent may be more harmful than caring for other family members, when considered in terms of mental health. However, no studies have compared depression risk in young carers of parents to depression risk in young carers of other relatives, using a non-carer control group. This study aims to address this gap.

Based on the above research, the current study aims to test whether caring for a parent (CP) or other relative (CR) is associated with greater risk of depression both cross-sectionally and longitudinally. A second aim will be to examine the associations between aspects of parent and peer relationship quality and depression in CPs. In terms of hypotheses, the first hypothesis is that CPs will experience more depressive symptoms than CRs and non-caring controls (NC) at T1 and T2. The second hypothesis is that CPs who report better social relationships (i.e. high positive aspects, low negative aspects) with peer/parents at T1 will be less likely to be categorized as depressed at both time points.

Methods

Participants, design and procedure

The current study utilizes data from Waves 3 (T1) & 4 (T2) of the Growing Up in Ireland Study (GUI), Cohort’98. GUI is a nationally representative longitudinal study of children and adolescents living in the Republic of Ireland. The sample design utilized a two-stage selection process, involving the school as the primary sampling unit, and the children within being the secondary unit. Children and their families were recruited in 2008/9, when the children were 9 years old. This study uses the data from when they were 17–18 years old (Wave 3) and 20–21 years old (Wave 4) as these waves specifically asked about caregiving duties. Information was collected from the young person, and where possible information was collected from their main (usually biological mother) and secondary caregivers (usually biological father).

Measures

Caregiving

Caregiving at T1 & T2 was assessed by asking the young people ‘Do you care for or look after another family member on a regular basis? By “caring” I mean things like cooking for them, helping them wash or dress, making sure they take medication, supervising them when there is no one else at home’. This question utilized a binary response format (yes or no). If they responded yes to this question, they were asked how this person was related to them (i.e. grandparent or other elderly relative, a parent or stepparent, a younger sibling, a sibling of the same age or older than you, someone else). We used these variables to group the participants into carers of parents (CPs; i.e. caring for a parent or step parent), carers of other relatives (CRs; i.e. caring for any other relative/someone else) and ‘non-carers’ (NC; i.e. responding no to the caring question at both time points).

As the interest of this study was on the longitudinal effects of caring, it only included those who reported caring for a parent or stepparent at both time-points; For participants who reported caring for a parent as well as other family members(s) at the same timepoint, they were coded as a CP. If a young person reported caregiving activity at T1 and T2, but the care recipient had changed (e.g. caring for a parent at T1 and a sibling at T2) they were excluded from the analysis.

Socio-demographics

Age, gender and education

At T1, information was collected from the young people and their parents on their age, gender, education, ethnicity, social class, and household composition. For analysis purposes, as well as limiting self-identification within the sample (a prerequisite of the data gatekeepers’ criteria), we recoded responses to the education question as ‘in education’ if participants stated they were still in school, studying for a further education course, or in higher education, and all other responses were coded as ‘not in education’ (e.g. unemployed). Again, to protect participants anonymity, we dichotomized ethnicity into White and non-White for analyses purposes. The family’s social class was derived from information relating to the main and secondary caregivers current and previous employment history and occupation. A social class classification was endorsed by the Central Statistics Office (CSO) which encompassed the following nine categories: professional worker, managerial and technical, non-manual, skilled manual, semi-skilled, unskilled, all others gainfully occupied, validly no social class and unknown. We created a social class dichotomy where professional worker and managerial and technical were classed as upper class, and the other categories were collapsed into a lower social classification. We also dichotomized household composition into single parent or two parent households.

Social relationships

Parental relationships

At T1, parental relationship quality was captured using the Network of Relationship Inventory behavioural systems version (NRI-BSV) with Mother and Father (Furman & Buhrmester, 2009), based on an adapted version used in the German PAIRFAM study (Huinink et al., 2011). This scale captures the quality of social relationships (positive and negative) that children have with their mothers or fathers on four dimensions: Intimacy (2 items, α = .82), Appreciation (2 items, α = .80), Conflict (2 items, α = .84) and Fear of Love Withdrawal (FOLW) (3 items, α .65) (The Centre for Effective Services, n.d.).¹ High scores on these scales indicated greater disclosure of personal thoughts and feelings by the young person to their mother/father, greater perceived recognition or appreciation from the mother/father, and increased conflict between the mother/father and the young person, respectively. Higher scores on the FOLW scale signifies the young person worries that their parents may not want to spend time with them or like them as much due to their behaviour. The scale has been adapted for use in several studies examining the impact of social relationships on young people’s mental health (Wang et al., 2021). Each item was scored on a 5-point Likert scale (from 1 = Never/Not at all, to 5 = Always/Completely true). The young person was asked to answer these questions with respect to their mother and father separately, where appropriate.

Peer relationships

The 25-Item Inventory of Peer Attachment (IPA) was used as a measure of social relationship quality with peers at T1 (Armsden & Greenberg, 1987). This self-report scale captures the young person’s perceptions, thoughts, and feelings regarding their relationships with close friends along the following dimensions: Trust (10 items), Communication (8 items), Alienation (7 items), with Cronbach’s alpha scores ranging from .72 to .92 (Armsden & Greenberg, 1987). Items were measured on a 5-point Likert scale (1 = Almost never or never true, to 5 = almost always or always true). High scores on the ‘Trust’ subscale indicate the respondents sense of security that their peers respect and understand their needs/desires. Higher scores on the ‘Communication’ subscale suggest the adolescent feels their peers are both sensitive and responsive to their emotional state and concerns. High scores on the ‘Alienation’ subscale indicate frequent or intense anger, or detachment from the attachment figure (Armsden & Greenberg, 1987). A total peer attachment was computed using the subscales. This scale yielded an internal reliability score of .92 (Armsden & Greenberg, 1987). Moreover, it has been used in several studies examining the impact of peer relationships on the mental health of young people (McMahon et al., 2020; Song et al., 2022).

Depression

Depression was measured using two scales: the Short Mood and Feeling Questionnaire (SMFQ) at T1 and the 8-item version of the Centre for Epidemiological Studies Depression Scale (CESD-8) at T2. The SMFQ has been widely used to measure depression in children and adolescence (Angold et al., 1995; Turner et al., 2014), while the CESD-8 is appropriate for use with young adults (Radloff, 1991). Both have sensitivity cut-off scores that indicate possible depression. For this analysis, scores greater than 12 on the SMFQ suggesting probable depression were used to examine between group difference in depressive symptoms at T1, while scores greater than 8 on the CESD-8 were used to examine between group differences in depressive symptoms at T2.

Short mood and feeling questionnaire (SMFQ)

The 13-Item SMFQ is a validated screening tool for assessing depressive symptomology in adolescence in epidemiological studies (Angold et al., 1995; Turner et al., 2014), correlating highly with other measures of childhood depression (Angold et al., 1995). It has good internal reliability (α = .88; Thabrew et al., 2018). The items were measured on a 3-point Likert scale (0 = True, 1 = Sometimes true, 2 = Not true).² A total score was computed for each participant, ranging from 0–26, with higher scores indicating higher depressive symptomatology, which was then dichotomized into a binary variable using a cut-off of 12 to signify probable depression (≤12 signifying no depression risk, ≥ 12.1 signifying depression risk). Other studies in young adult samples have utilized a value of 11 or higher as a cut-off (Eyre et al., 2021; Turner et al., 2014). In this study, a score greater than 12 was adopted, which is in line with the Child Outcomes Research Consortium approach.

Centre for epidemiological studies depression scale (CESD-8)

The CESD-8 is a widely used self-report measure designed for use as a depression screening tool in the general population (Radloff, 1997) and in young adults (Radloff, 1991). It has good internal reliability (α = .89; Seth et al., 2011) and has been previously used to assess depressive symptomology in young adult carers (Gallagher et al., 2022a). Respondents rate how often during the past week they felt a certain way on a 4-point Likert scale (0 = rarely or none of the time (less than 1 day) 1 = Some or a little of the time (1–2 days), 2 = Occasionally or a moderate amount of the time (3–4 days), 3 = most or all the time (5–7 days).³ A total score is then computed, ranging from 0–24. We utilized a cut-off of 8, which has a sensitivity of 99% and specificality of 71% (Briggs et al., 2018).

Statistical analysis

Prior to analysis, data was screened and cleaned and weights⁴ were applied to the data to account for attrition and the sampling procedure and design. Initially, descriptive statistics and tests of differences across the groups (CP, CR, NC) are reported (see Table 1). Histograms and Q-Q plots were examined to assess normality of scale data. Concludingly, non-parametric tests of difference (Mann-Whitney U tests and Kruskal Wallis H tests) were used to assess for differences between carer groups on continuous variables. Chi-square tests assessed differences between carer groups on categorical variables. A series of hierarchical logistic regressions were then performed. In each of these regressions, covariates were entered in Step 1, with the primary predictor variables in Step 2 or Step 3. For hypothesis tests, a 5% level of significance was used. To reduce the family wise error rate, a Bonferroni adjustment was applied as a post-hoc procedure. Odds Ratios with associated 95% confidence intervals are reported for the measure of effect size.

Table 1. Weighted means and standard deviations and kruskal-wallis H test of difference on parent and peer relationship quality variables.

Variable	NC (3033) (Mean, SD)	CP (98) (Mean, SD)	CR (181) (Mean, SD)	Tests of difference
Peer attachment	98.17(14.94)	94.67(16.72)	97.13(14.17)	H(2) = 4.59, p = .101
Maternal intimacy	6.15(1.93)	6.57(2.37)	6.39(1.98)	H(2) = 8.699, p = .013
Maternal appreciation	8.03(1.63)	7.96(2.12)	7.91(1.64)	H(2) = 3.65, p = .161
Maternal conflict	5.2 (1.54)	4.77 (1.7)	5.29(1.61)	H(2) = 6.73, p = .035
Maternal fear of love withdrawal	5.96 (2.56)	6.72(3.51)	5.98 (2.52)	H(2) = 2.78, p = .249
Paternal intimacy	5.02 (1.96)	5.22 (2.40)	5.05 (2.28)	H(2) = .504, p = .777
Paternal appreciation	7.54 (1.95)	6.87 (2.65)	7.44(1.83)	H(2) = 1.76, p = .415
Paternal conflict	4.94 (1.77)	4.15(1.78)	5.07(1.81)	H(2) = 17.703, p < .001
Paternal fear of love withdrawal	5.57 (2.88)	5.06 (2.95)	5.68 (2.85)	H(2) = 7.43, p = .024

Results

Data from 3,312 young people were included in the analyses.

Sample characteristics

There were no significant differences between groups on household composition (χ² (2) = .801, p = .67) or gender (χ² (2) = .02, p = .99). There were approximately equal numbers of boys and girls in each group, with CPs being 50.5% female, CRs being 49.7% female, and NCs being 49.8% female. Approximately 25% of the entire sample reported belonging to a single parent household. Finally, 95.1% of NCs were in full time education, and there were no significant differences in education status between the groups (χ² (2) = .98, p = .61).

There were significant differences between groups on their main caregiver’s gender (χ² (2) = 8.39, p = .015), and ethnicity (χ² (2) = 29.9, p < .001); CPs were more likely to have a male main caregiver, and NCs were the most likely to have a main carer who was white compared to CRs and CPs. Over 40% of the NC’s had parents whose professions placed them in the higher social classification, followed by 33.7% of CRs, and < 30.6% of CPs (χ² (2) = 9.84, p = .007). NC’s had older parents (M = 48.5, SD = 5.6), compared to CPs (M = 46.12, SD = 5.62), and CRs (M = 46.43, SD = 6.52), (H(2) = 42.15, p < .001). Finally, NCs were slightly younger (M = 17.18, SD = .38), followed by CRs (M = 17.2, SD = .4), with CPs being the oldest (M = 17.3, SD = .46), (H(2) = 10.31, p = .006).

Primary predictors

Means and standard deviations of the primary predictors can be seen in Table 1. No significant differences were observed between groups on total peer attachment. However, differences were observed between groups on maternal intimacy, maternal conflict, paternal conflict, and paternal FOLW. Interestingly, CPs scored highest out of the three groups on maternal intimacy, and lowest on maternal conflict, paternal conflict, and paternal FOLW. Whereas, CRs scored highest on both maternal and paternal conflict, and highest on paternal FOLW.

Do CPs differ in depression status from CRs and NCs cross-sectionally and longitudinally?

As can be seen in Table 2, CPs were significantly more likely to be categorized as being depressed at T1 (Cramer’s V = .078, p < .001). At T2, despite a notably higher percentage of CPs (<30.6%)⁵ and CRs (23.8%) compared to NCs (18.2%) being categorized as depressed, these were not statistically different from each other. Further, between T1 and T2, the number of CPs who were in this depressed category fell, from 30.6% to < 30.6%. Interestingly, the number of CRs in the depressed category increased from T1 to T2, from < 16.6%⁶ to 23.8%.

Table 2. Depression status of young people at wave 3 and wave 4 and tests of difference.

Variable	NC (3033)	CP (98)	CR (181)	Test of difference
SMFQ (CO12, % depressed)	14.3	30.6	<16.6*	X(2) = 19.97, p < .001
CESD-8 (CO8, % depressed)	18.2	<30.6*	23.8	X(2) = 4.63, p = .099

*Statistics have been suppressed due to regulations imposed by the CSO.

Between-subjects analyses

Cross-sectional associations between social relationships and depression

We controlled for group differences in socio-demographics (Ethnicity, main caregiver’s gender and age, social class, young person’s age) in Step 1 of the logistic regression model, to see if any of these were contributing to the caring groups differences in depression status. The step 1 model was non-significant, [X²(5) = 10.21, p = .069]. In step 2, carer status was significant, [X² (7) = 22.59, p = .002] whereby CPs were seen to have a 2.6-fold increased risk of depression when compared to NCs (OR = 2.55, 95% CI 1.56, 4.17). CRs were not seen to be at an increased risk of depression in comparison to NCs. Hosmer and Lemeshow tests revealed this model to be an excellent fit to the data (Chi-Squared = 5.312, df = 8, p = .724), explaining between .8% and 1.4% of the variance in depression. As such, the previous finding that CPs are more at risk of depression remained after controlling for sociodemographic variables.

The final model (see Table 3) investigated whether parental relationship variables previously seen to differ between carers and non-carers explained any of the variance in depression. This model was significant, X² (11) = 275.16, p < .001, with caring for a parent being the strongest predictor of depression, showing that CPs are 3.67 times more likely to be depressed than their non caring peers. Hosmer and Lemeshow tests suggest that this model is a good fit (Chi-Squared = 13.51, df = 8, p = .096), explaining between 9.3% and 16.8% of the variance in depression. Results show that young people reporting higher maternal intimacy, are 15% less likely to be depressed at baseline. Similarly, those reporting high maternal conflict at baseline are 37% more likely to be depressed, while those reporting higher paternal fear of love withdrawal have a 21% increased relative risk of having depression. However, adding the relationship predictors did not abolish the caring status effects observed previously. Thus, young people’s relationships with their parents are predictive of depression status regardless of their carer status.

Table 3. Binary logistic regression predicting depression at T1 (SMFQ) using demographics, carer status, and parental relationship quality predictors.

Variable	Odds ratio	95% confidence interval	p - value
Step 1
MC gender	1.579	.762, 3.270	.219
MC age	.990	.968, 1.012	.362
MC ethnicity	.355	.114, 1.111	.075
Social class	.999	.791, 1.261	.992
YP age	1.263	.95, 1.68	.110
Step 2
Non carer	ref	ref	<.001
1) carer of parent	3.665	2.139, 6.279	<.001
2) carer of other	1.099	.669, 1.807	.710
Step 3
Maternal intimacy	.846	.796, .899	<.001
Maternal conflict	1.372	1.267, 1.486	<.001
Paternal Conflict	.960	.898, 1.027	.234
Paternal Fear of Love Withdrawal	1.213	1.171, 1.257	<.001

Longitudinal associations between T1 social relationships and depression in caring youths

We conducted a similar set of hierarchical logistic regressions with depression at T2 as the outcome (CESD-8 > 8), where we controlled for socio-demographics in step 1, T1 depression status in Step 2, carer status at step 3, and parental relationship variables at step 4. As can be seen in Table 4, having older parents was protective against depression status [X²(5) = 15.95, p = .007]. Hosmer and Lemeshow tests show this model fit the data well [X²(8) = 15.06, p = .058]. In step 2, depression status at T1 positively predicted depression status at T2 such that those who were depressed at T1, had an almost five-fold risk of being depressed at T2 (OR = 4.63, 95% CI 3.66, 5.85). In step 3, being a carer was not predictive of depression status, while in step 4, maternal conflict (OR = 1.14, 95% CI 1.06, 1.23) and paternal FOLW (OR = 1.05, 95% CI 1.01, 1.08) at T1, positively predicted depression at T2, such that greater conflict with mothers and a higher paternal FOLW was associated with an increased risk for all young people. Further, those young people reporting a high level of maternal intimacy at T1 had a 6% reduced risk of having depression at T2. Overall, this model was significant, X²(12) = 208.37, p < .001, explaining between 7.2% and 11.9% of the variance in depression at T2. Together, these suggest that the key predictors of depression at T2 were having older parents, depression at T1, and mothers’ intimacy and conflict as well FOLW from their fathers and not caregiving per se.

Table 4. Binary logistic regression predicting depression at T2 using demographics, depression at T1, carer status, and parental relationship quality predictors.

Variable	Odds ratio	95% confidence interval	p - value
Step 1
MC gender	.823	.473, 1.432	.490
MC age	.968	.95, 99	.001
MC ethnicity	1.026	.478, 2.204	.947
Social class	1.030	.835, 1.271	.781
Step 2
SMFQ (Dep, not dep)	3.67	2.86, 4.71	<.001
Step 3
Non carer	ref	ref	.730
1) carer of parent	1.008	.56, 1.801	.980
2) carer of other	1.19	.77, 1.828	.428
Step 4
Maternal intimacy	.94	.89, .99	.018
Maternal conflict	1.14	1.06, 1.23	<.001
Paternal fear of love withdrawal	1.05	1.01, 1.08	.010
Paternal conflict	1.04	.98, 1.101	.247

Hosmer and Lemeshow test [X(8) = 33.94, p < .001].

Within carers of parents analyses

The next set of analyses focused on associations with depression status within CPs. For this, differences among depressed and non-depressed CPs were examined. At T1, Fischer’s Exact Test revealed no differences between depressed and non-depressed CPs on main carer gender (Two-tailed, p = .267) or main carer ethnicity (Two-tailed, p = .309). Similarly, Chi-squared tests observed no differences between depressed and non-depressed CPs on household composition (χ² (1) = 1.39, p = .238), social class (χ² (1) = .647, p = .476) or education status (χ² (1) = 2.32, p = .127). Interestingly, Mann Whitney U tests revealed significant differences between groups on main carer age (U = 382.5, p < .001) and young person age (U = 706.00, p = .019) with the depressed group (M = 41.58, SD = 5.95) having younger parents compared to the non-depressed group (M = 48.1, SD = 4.16). This depressed group were also older (M = 17.47, SD = .51), than their non-depressed counterparts (M = 17.23, SD = .43). Moreover, the depressed group had more than twice the number of females than males, (χ² (1) = 5.63, p = .018).

In Table 5, we have presented means and standard deviations of parental relationship quality variables for the groups (depressed vs non-depressed CPs). Interestingly, depressed CPs had significantly lower mean scores on maternal appreciation, paternal intimacy, and paternal appreciation. Depressed CPs also had lower scores on paternal conflict, and higher scores on maternal FOLW. It is recommended that for each predictor in a model, there should be between 10–20 events, to avoid model overfit (Stoltzfus, 2011). Therefore, due to the reduced sample size in this sub-sample (N = 98), we added the primary predictor variables (i.e. relationship quality variables) separately into step 2 (for cross sectional) or step 3 (for longitudinal) of the models.

Table 5. Weighted means and standard deviations and test of difference (Mann-Whitney U test) of parental and peer relationship quality predictors.

Variable	Depressed (N = 30)	Non depressed (N = 68)	Tests of difference
Maternal intimacy	6.11(2.44)	6.78(2.32)	U = 680, p = .142
Maternal appreciation	7.09(2.69)	8.36(1.68)	U = 524.5, p = .003
Maternal conflict	4.67 (1.76)	4.82(1.69)	U = 821, p = .853
Maternal fear of love withdrawal	9.32 (4.22)	5.55(2.37)	U = 408.5, p < .001
Paternal intimacy	3.29 (1.88)	5.99 (2.15)	U = 243.5, p < .001
Paternal appreciation	4.55 (2.74)	7.79(1.98)	U = 247.5, p < .001
Paternal conflict	3.72 (2.16)	4.33 (1.59)	U = 524.00, p = .028
Paternal fear of love withdrawal	5.58 (3.6)	4.86(2.64)	U = 667.00, p = .424
Peer attachment	82.85 (17.71)	99.84(13.41)	U = 402.00, p < .001
Alienation	15.84(5.48)	15.2 (3.29)	U = 848.5, p = .496
Peer communication	24.56 (10.39)	31.17(5.66)	U = 595.00, p = .005
Peer trust	32.12 (12.26)	41.88 (6.36)	U = 471.00, p < .001

Parental relationship quality as cross-sectional predictors

Logistic regressions analyses, including T1 depression status (SMFQ scores > 12) as the outcome variable, with covariates (main carer age, young person age and gender) entered in Step 1 and primary predictors in step 2 were carried out. The step 1 model significantly predicted depression status, [X²(3) = 34.55, p < .001], with having an older parent reducing the risk of depression by 24% (OR = .76, 95 CI .66, .87, p < .001). Being female also conferred an increased risk. Paternal intimacy (OR = .601, 95% CI .43, .85, p = .003), paternal appreciation (OR = .63, 95% CI .46, .86, p = .003), maternal FOLW (OR = 1.32, 95% CI 1.08, 1.6, p = .006) and maternal appreciation (OR = .69, 95% CI .51, .92, p = .012) were significant predictors of depression. Paternal intimacy and appreciation were associated with a 40% and 37% reduced risk of depression, respectively. Similarly, maternal appreciation reduced risk by 31%. Maternal FOLW increased risk by 32%.

Peer relationship quality as cross-sectional predictors

Depressed carers and non-depressed carers differed in total peer attachment, peer communication, and peer trust (see Table 5). Depressed CPs reported significantly lower levels of overall peer attachment, as well as lower levels of peer communication and peer trust. Three hierarchical logistic regressions, controlling for covariates (main carer age, young person age and gender) in step 1 and adding peer trust, communication and total peer attachment individually in step 2, showed that increased attachment (OR = .951, 95% CI .92, .987, p = .008), communication (OR = .909, 95% CI .842, .982, p = .016), and trust (OR = .92, 95% CI .86, .98, p = .012) reduce the relative risk of depression (SMFQ). For every unit increase in attachment, the risk of depression is reduced by 4.9%, with trust and communication reducing the risk by 8% and 9%, respectively. In each of these models, main carer age was a significant predictor of depression, with younger main carer age increasing the risk of depression.

Longitudinal regressions

Using similar methods as above, we examined if, after accounting for covariates and T1 depression status, the cross-sectional predictors of depression status, would predict depression status at T2 (CESD-8). First, the model with covariates (main carer age, young person age and gender) in Step 1, and T1 depression status in step 2 was significant (X²(4) = 13.77, p = .008), accounting for between 13.2% and 20% of the variance in depression. In step 3, we added peer attachment, and found that peer attachment was not a longitudinal predictor of depression status (OR = .99, 95% CI .951, 1.021, p = .42). Previous depression status remained a significant predictor of depression (OR = 5.88, 95% CI 1.51, 22.87, p = .011) with an almost 6-fold increased risk of being depressed at T2 if you were depressed at T1.

In a similar set of regressions, controlling for covariates in Step 1 and depression in step 2, paternal appreciation (OR = .66, 95% CI .49, .89, p = .008) and maternal appreciation (OR = .71, 95% CI .54, .95, p = .019) significantly predicted depression, decreasing one’s relative risk by 34% and 29% respectively. However, depression remained a strong longitudinally predictor, increasing one’s relative risk by up to 5 times (OR = 5.01, 95% CI 1.19, 20.98, p = .028).⁷

Discussion

To our knowledge, this is the first study to examine comparatively the differential influence of caring for a parent, vs caring for another family member, on depression in adolescence and young adulthood. Here we confirmed that caregiving, when directed towards a parent, was associated with a higher risk of depression. Moreover, while maternal intimacy was associated with a reduced depression risk, maternal conflict and paternal FOLW increased this risk for all young people. However, the strongest predictor of depression after controlling for these factors was caring status, with CPs being 3.67 times more likely to be categorized as depressed at T1, compared to CRs and NCs.

In the longitudinal analysis, there were no group differences in depression status at T2. We also found that after applying controls, depression at T1 was the strongest predictor of depression at T2. For all young people, negative relationship quality with parents increased the risk of depression, while positive aspects reduced this risk. Findings from the CP subgroup analysis found that both parent and peer relationship quality were predictive of T1 depression status. Specifically, paternal intimacy and appreciation were associated with a 40% and 37% reduced risk of T1 depression, respectively. Similarly, maternal appreciation reduced risk by 31%. Maternal FOLW increased risk by 32%. For peer relationships, overall attachment reduced risk by 4.9%, with follow-up analyses showing that trust and communication specifically reduced risk by 8% and 9%, respectively. Importantly, for CPs, prior depression was the strongest longitudinal predictor of depression, incurring a 6-fold increased risk. Moreover, it was positive parental relationships, and not the peer relationships, that were important for depression at T2. Here, results showed that paternal and maternal appreciation were associated with decreasing CPs relative risk of depression by 34% and 29%, respectively.

These findings extend previous research on young and young adult carers (Pakenham & Cox, 2015) by showing that carers (aged 17/18) are more vulnerable to depression if they are caring for parents compared to caring for a different relative. Although there is a paucity of research on the impact of caring for parents on young people (Pakenham & Cox, 2014b), others studies on young people living with parents who have a mental/physical illness do demonstrate a higher risk of mental health issues (Aldridge, 2006; Anderson & Hammen, 1993; Barkmann et al., 2007). One explanation for why CPs are more at risk for depression than CRs may be because they are taking on a primary carer role within the family (Charles, 2011). While the findings of this study don’t clarify why CPs are more at risk, it is the first to demonstrate that the nature of the carer-care recipient relationship is an important factor influencing depression risk in young carers.

Interestingly, these results showed that caregiving did not predict depression over time, whether for a parent or a different family member. These findings are in contrast from other longitudinal studies showing increased psychological distress (Nakanishi et al., 2022) and poorer mental health in young carers over time (Brimblecombe et al., 2020). However, these differences may reflect diversity in age ranges for young carers between studies (Brimblecombe et al., 2020) or a focus on COVID-19 effects (Nakanishi et al., 2022). Another explanation for this could be the differential use of depression measures at T1 and T2 in this study, or the use of a categorical measure of depression as opposed to a continuous measure. Moreover, the findings of this study could reflect the focus being on CPs as opposed to young carers in general. If this study examined young carers as a homogenous group, a different pattern may have emerged. Interestingly, while there were no statistically significant differences in depression status between the groups at T2, the results show that the number of depressed CPs fell, whereas the number of depressed CRs and NCs increased (see Table 2).⁸ Concludingly, more longitudinal research is needed to examine the long-term effects of caregiving on depression, and mental health in general, in adolescence and young adulthood.

A crucial component of this study was to examine the role of parent and peer relationships on depression risk. For the young people in general, including young carers, high maternal conflict, low maternal intimacy, and high paternal FOLW predicted depression in adolescence and young adulthood. However, in adolescence, caring for a parent was the strongest predictor of depression cross-sectionally. Over time, the strongest predictor of depression for all young adolescents was prior depression status, suggesting that treating depression early on, or preventing its onset, will have health protective effects over time. When looking within CP, cross-sectional analyses showed both parent and peer relationships to be predictive of depression. However, results from the long-term analyses showed that positive parental relationships, specifically characterized by expressions of recognition and appreciation from the parents to the young carer, were the strongest social predictors. Further, in these same analyses, those CP, who were depressed at T1, were five-times more likely to be categorized as depressed at T2, again highlighting the importance of addressing this vulnerability early on. These findings also support and extend a stress model of youth caregiving (Vitaliano et al., 1991), whereby increased vulnerability (i.e. caring for a parent, depression) and low resources (i.e. low positive aspects and high negative aspects of parent and peer relationship quality) result in increased risk of psychological distress (i.e. depression).

In relating the social relationship findings to previous research, they are similar to that found elsewhere, whereby social factors are important for the health and well-being of young carers (Alfonzo et al., 2020; Family Carers Ireland, 2023). In those studies, a risk of isolation is damaging, whereas in this study we demonstrate that positive and negative social relationship quality with both peers and parents have protective and damaging effects, respectively. Other studies on young carers have reported a lack of understanding amongst peers of their role, and a lack of empathy for the challenges they face, to be associated with poor mental health (Stamatopoulos, 2018), and we extend on those, by showing that low communication and trust amongst peers directly influences depressive symptoms in adolescence. Thus, peer-based interventions or support groups promoting effective and responsive communication and trust for young carers of parents could be effective in reducing depression risk. However, given the findings for parental relationships and the risk of depression, they also underscore the importance of using a whole family approach to intervention work. These approaches should promote positive relations (e.g. appreciation building) amongst parents of CPs, as these were the strongest mitigating factors against depressive symptoms. In fact, researchers have highlighted that when young carers don’t feel appreciated in their caring role they are at increased risk of depression and reduced risk of positive outcomes such as competency and self-esteem (Chojnacka & Iwański, 2022; Janes et al., 2022). This study substantiates and expands on these findings to show that appreciation and recognition from both parents, not just the care receiver, significantly reduce depression risk in CPs, and that these aspects of parental relationship quality may be particularly important for CPs mental health. These findings may be particularly important for CPs caring for parents with mental health issues, as these relationships may incur additional strain (Van Loon et al., 2015). On a final note, this within group analysis also found that the strongest predictor of depression in young adulthood is depression in adolescence, increasing risk by up to 6 times for CPs.

These findings confirm that CPs are an extremely vulnerable group. Notably, young carers are largely overlooked in policy and legislation in many developed nations such as the United States and Ireland (Leu & Becker, 2017). Evidence shows that a robust, reliable, country specific research base helps to advance national responses to young carers needs, largely through non-governmental organizations (NGO) utilizing research evidence to lobby for young carer recognition in policy, law, and practice (Leu & Becker, 2017). The findings of this study can be used by NGOs, such as Family Carers Ireland, to advocate for the needs of CPs, who evidently confer a particular mental health risk. Additionally, health leaders and health practitioners should use this research as evidence of a need to quickly identify young carers of their patients and offer supports to reduce the risk of depression (Lacey et al., 2023).

Strengths, limitations and recommendations for future research

Despite the strengths of the study (e.g, population-based research, non-biased sampling, longitudinal design), there are some study limitations. Here, we did not examine young people caring for multiple persons, i.e. a proxy measure of the intensity of caregiving responsibilities. While the lack of a measure of caregiving intensity could be considered a limitation, previous research has emphasized an overreliance on data from young people providing ‘substantial’ care, and accessing support services (Janes et al., 2022). Thus, this study is representative of all young Irish people who provide care, regardless of the intensity or extent of caregiving responsibilities. Another, limitation of the current study is the differential use of depression measures at T1 and T2. Although depression assessments were adjusted to account for the changing life stage (i.e. late adolescence to emerging adulthood) of the sample, using different depression measures makes it difficult to adequately assess changes in depression overtime. To minimize this effect, we utilized established cut off scores. However, future research in this area should aim to use the same validated measure of depression at multiple timepoints to address changes in depression longitudinally. Moreover, due to the database gatekeeper criteria, we were unable to generate several important demographic descriptive statistics that may shed light on the characteristics of the young CPs. However, the main finding concurs with several others in the field and have extended on those findings, particularly around the importance of social relationships on the health of CPs. The bidirectional influence of depression on social relationships, and vice versa, should be considered in follow-up studies. Finally, an issue within the youth-caregiving literature is the lack of consensus surrounding age ranges for young and young adult carers (Hendricks et al., 2021). Thus, as this sample is first examined in adolescence and is looked at again in young adulthood, it was difficult to align these age groups with previous research on youth providing care.

In summary, the present study confirmed that caring for a parent is associated with an increased risk of being depressed relative to caring for other relatives. However, this risk is not conferred over time. Nonetheless, the within CP analysis showed that depression at T1 increased the risk of T2 depression by up to six times, implying a greater vulnerability with implications for early intervention. Importantly, the quality of their close relationships can offer protective effects against depression onset and reoccurrence. Here, results found both peer and parent relationship quality to be important contributors to risk of depression for these CPs. In this group, interventions which promote and encourage peer relationships based on communication and trust can significantly reduce depression risk for young carers of parents. Moreover, whole family approaches to interventions and support services for young carers of parents, which aim to encourage positive relations between parents and their teenagers, can reduce depression risk in adolescence, and have lasting effects into adulthood. Importantly, depression in adolescence is likely the best predictor of depression in early adulthood. This finding highlights the need for tailored support and early interventions for adolescent CPs to prevent them embarking on a trajectory of depressive symptoms that pervades into adulthood.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

The research conducted in this publication was funded by the Irish Research Council under grant number [GOIPG/2023/2478]

Notes on contributors

Aoife Bowman Grangel

Aoife Bowman Grangel is a PhD student with an interest in the health and well-being of young carers

Jennifer McMahon

Dr Jennifer McMahon is a Senior Lecturer in Psychology who specialises in youth and adolescent mental health

Nikki Dunne

Dr Nikki Dunne is Research Manager with Family Carers Ireland

Stephen Gallagher

Stephen Gallagher is an Associate Professor of Health Psychology with a special interest on health and well-being of family caregivers

Notes

1. An example of an item used to assess intimacy is ‘How often do you tell them what you’re thinking?’. One item used to assess appreciation is ‘Does the following person express recognition for things you’ve done?’. An item assessing conflict asked, ‘How often are you and this person angry or annoyed at one another?’. FOLW as assessed using items such as ‘I’m afraid they think I’m silly or stupid if I make a mistake’.

2. Participants are asked to rate how often in the past 2 weeks they felt a certain way. One item used in this scale is ‘I felt miserable or unhappy’.

3. One item used in this scale is ‘I felt sad’.

4. The sample is weighted to account for the young person’s gender, level of education of the primary caregiver, family type, number of children in the household at age 9, household social class, household equivalized income quintile, work status of the primary caregiver, and the young person’s score on the Drumcondra reading test at 9 years old.

5. The exact statistic has been suppressed due to the database gatekeepers’ criteria on data confidentiality.

6. The exact statistic has been suppressed due to the database gatekeepers’ criteria on data confidentiality.

7. Growing Up in Ireland (GUI) is funded by the Department of Children, Equality, Disability, Integration and Youth (DCEDIY). It is managed by DCEDIY in association with the Central Statistics Office (CSO). Results in this report are based on analyses of data from Research Microdata Files provided by the Central Statistics Office (CSO). Neither the CSO nor DCEDIY take any responsibility for the views expressed or the outputs generated from these analyses.

8. Due to regulations regarding data confidentiality, the percentage of CPs who were depressed at T2 has been suppressed.