Gender differences in the risk of depressive disorders following the loss of a young child: a nationwide population-based longitudinal study

Major depressive disorder (MDD), a leading cause of economic burden in many parts of the world, accounted for 8.2% of global years lived with disability in 2010 [1]. A recent review indicated that approximately 4.7% of the world population currently has MDD, and the prevalence varied widely across region and demographic stratum [2]. In Taiwan, the incidence of treated MDD was on the notable rise over the past decades. For the individuals aged 45 to 54 years — the age group with the incidence peak, the estimated rate for men and women was 2.15 and 4.51 per 1, 000 person-year, respectively [3]. Cumulative evidence has pointed out several individual-level health and social characteristics associated with increased risk of onset and recurrence of MDD, including chronic diseases (both mental and physical one), disadvantaged socioeconomic status, low social integration, and stressful life events [4‐8].

Having a loved one die was the most frequently reported life event in adulthood [9]. Experiencing the death of a beloved one (especially a family member) often causes extremely complicated persistent grief despite coping mechanisms [10]. An increased risk of adverse health outcomes has been reported among the bereaved spouses, parents, and children [11‐13], with the severity of bereavement often depending on the closeness and bonding between the living and the dead. Since the parent-child relationship has been regarded as one of the closest ties that a middle-aged individual may have, the grief after losing a child is often intense and chronic. Studies have consistently demonstrated that the bereaved parents may experience greater physiological and psychological disorders, including cancer, diabetes, depression, anxiety, suicidal attempts, and even death through health-compromising behaviors, dysregulated neuroendocrine/immune systems, and intertwined connections with prior vulnerabilities [14‐17]. The emergence of these disorders not only fluctuates over time but also varies by children’s characteristics (e.g., age and unexpectedness of death) [7, 18, 19]. In the case of suicide, the risk was reported to emerge abruptly in the first months of child loss and appeared more salient if the child died in early childhood [20, 21].

For parents, several characteristics have been documented to explain different risks of psychological distress after losing a child [5, 18, 22]. Gender is one attribute of great relevance in identifying possible etiological processes underlying bereavement effects [15, 21, 23, 24]. With few exceptions [25], studies have generally indicated that women were more likely to experience suicidal ideation and attempts, anxiety, depression, lower quality of life, and reduced wellbeing after losing a child as compared with men [18, 19, 21]. Such gender differences in bereavement effects may result from biological vulnerability, social role expectation, and coping strategies [19, 25‐27]. Some evidence indicated that in response to life events or stressors, women appeared to use emotional coping styles more often and have their emotions expressive or confrontive. Men, on the other hand, were more likely to use detachment coping styles and have more emotional inhibition [25, 28].

Losing a child to death may exert detrimental effects on parents differentially in terms of observational interval and outcome indicators [14, 19, 22, 25, 29]. Evidence from qualitative interviews and quantitative analyses consistently showed that the bereaved parents mostly took 3–4 years or even longer to put the child’s death into perspective [22, 29]. A recent study analyzing 902 bereaved parents identified from the registry datasets in Finland found that the bereavement effects, as manifested in the use of psychotropic medications, reached the highest prevalence within the first year and fell sharply thereafter in the following three years. For children who died from diseases (not from external causes), the use of psychotropic medication was gradually elevated in the three years preceding the death of a child [22]. Meanwhile, a number of studies have noted that adverse health outcomes subsequent to a child loss may depend on subgroups defined by household characteristics, such as income status, household size, and social network [5, 8].

Although previous studies have highlighted some factors relevant to psychological distress of losing a child, the evidence was mostly collected in Western societies [25]. In traditional East Asian culture, son is more important and preferred for not only economic reasons but his ability to continue the family lineage [19]. In this context, some evidence has supported that physical and psychological distress in the bereaved parents appeared stronger after losing a son than a daughter [19, 30, 31]. Meanwhile, mothers were often the ones taking care of the demand posed by homemaking and childcare, and children take on great symbolic importance on married women’s identity. Finally, these studies often built upon a relatively small sample size and cannot systematically characterize subgroup variation in the manifestation of parental bereavement. In this study, we turned to the National Health Insurance data in Taiwan to identify a retrospective cohort of parents who lost their young children to death, following up throughout three years. Here, with a focus on “MDD” —the most common and disabling mental disorder in middle adulthood, the present study aimed to (i) evaluate father- and mother-specific risk of MDD treatment seeking after a child’s death, and (ii) to investigate whether such MDD risk may be manifested differentially by parents’ prior health condition.

Among those deceased children, nearly one third (n = 1861; 31.6%) had a catastrophic illness for more than 1 year compared to 1.4% (n = 320) in their birth-year- and gender-matched peers (see Table 1). Approximately 34% of deceased children’s health insurance was covered by their mothers (i.e., the mother had a job), significantly lower than the 44% in their counterparts. In general, the parents of deceased children were more likely to be older (e.g., mothers: 35 years old or older: 44% vs. 42%), to have prior history of MDD (e.g., fathers: 1.83% vs. 0.28%), and to have CPC (e.g., fathers: 39% vs. 12%). Nearly 5 % of bereaved mothers and two-point-4 % of bereaved fathers have sought treatment for MDD within three years after losing a child to death.

With stratification by parents’ gender, the estimated hazard of seeking MDD treatment after three years of a young child loss was depicted in Fig. 1. The hazards of MDD were generally greater in mothers, regardless of bereavement status. For the bereaved mothers, the hazard fell within the first four months (see the solid lines of the left panel) and reached the platform near the 14th month. In contrast, the hazard of seeking MDD treatment for the bereaved fathers abruptly fell in the first eight months after losing a young child, remaining moderately fluctuated till the end of 22nd month (see the solid lines of the right panel).

For the middle-aged mothers, univariate Cox regression analyses indicated that having lost a child to death (HR = 6.54), having the prior history of MDD (HR = 65.1), and having a CPC (HR = 8.13) were the three strongest predictors for increased hazard of visiting a doctor for MDD (see Table 2). In addition, having a child who received a diagnosis of a catastrophic disease (e.g., cancer) may increase maternal hazard of receiving the treatment for MDD by 172% (less than one year) and 386% (one year or more), and disadvantaged income may also slightly increase such risk (poverty: HR = 2.32). With statistical adjustment for child’s health and family characteristics, losing a child was found to increase the hazard of receiving the treatment of MDD by 506% (Model I: 95% CI = 4.47–8.21), and such risk was slightly attenuated while maternal sociodemographic and health characteristics were taken into account simultaneously (Model II: aHR = 4.71; 95% CI =3.35–6.64).

The profile of significant predictors for receiving the treatment for MDD in fathers was similar with those in mothers, although variation exists in the magnitude (Table 3). The crude hazard ratio of MDD associated with losing a child was 4.71 in fathers, and the hazard ratio linked with prior history of MDD was as high as 117.3. For the middle-aged men, having a child with a catastrophic illness for one year or more and disadvantaged incomes may accordingly increase the hazard of MDD treatment seeking by 215 and 86% (p < 0.01). With all listed variables adjusted simultaneously, losing a young child to death was found to moderately increase fathers’ risk of MDD by 93% (Model II: 95% CI = 1.27–2.95).

With a focus on risk variation in the subgroup defined by parents’ prior health condition, we found that the increased risk of MDD was especially prominent for those parents who received no medical treatment for chronic condition in the three years prior to the death of the index child (Fig. 2). A child’s death was associated with adjusted HR of 9.55 among those non-CPC affected mothers, whereas the estimate was only 2.38 for those who had at least one CPC. Indeed, for fathers, such increased hazard of bereavement effects was only significant for those who previously had no CPC (aHR = 5.54.06; 95% CI = 3.24–9.50) and had no MDD (aHR = 2.42; 95% CI = 1.55–3.78).

We utilized population-based data to identify the pairs of young children and their parents to investigate the extent to which losing a young child to death may affect parents’ subsequent bereavement effects, as indexed by treatment seeking for MDD, in the middle-aged population and explore how the risk operates differentially by prior health condition. Our analyses showed that the risk of MDD was especially prominent in the first year after the death of a child. The occurrence of MDD was elevated by 4.71-fold in the bereaved mothers, more than two times higher than 1.93-fold in fathers. Notably, the so-called “bereavement effects” were not homogenous across individuals: the significant increased risk of seeking MDD treatment was more salient in those who had no prior MDD or those who had no CPC (e.g., diabetes and hypertensive disease), especially in fathers.

On the basis of the ICD-9 CM codes, our findings indicated that parents’ increased risk of MDD subsequent to a child loss may manifest far beyond six months [7, 35‐38]. This observation may partly resonate with conceptualization of pathological bereavement (i.e., persistent complex bereavement disorder and prolonged grief disorder) in DSM-V and ICD-11 [39, 40]. Disturbed/complicated grief is distinct from depression and post-traumatic stress disorder, although the rates of comorbidity between these conditions are high and the duration can be longer than 6 to 12 months after the episode of loss [41‐44]. We also noted that the observed “bereavement effects” in the first year were slightly greater, yet less salient after that, than those reported previously in other Western societies (e.g., Canada). This discrepancy may be partly due to the variation in (i) cultural or societal differences in clinical manifestation of MDD, (ii) perceived need in treatments seeking, and (iii) the nature of outcome indicators (e.g., MDD vs. antidepressant; incident vs. prevalent cases) [45‐47].

Our results corroborate previous findings that mothers may suffer more than fathers while facing the loss of a child, in terms of bereavement responses [15, 20, 22, 48]. The mechanisms underlying the observed gender differences in seeking treatment for MDD may be attributed to a mixed result of vulnerability, stress coping, help seeking, and role expectation in childcare [22, 49]. For example, the loss of a child may pose a threat to the identity of bereaved parents, which is especially true for women. Since the majority of primary caregivers for children are women in this study context, this adjustment process can be challenging for mothers and even more difficult for those unemployed ones with limited social support or a poor social network [50‐52]. On the other hand, fathers often adopt emotional inhibition, detachment, or denial (i.e., ignoring their grief) as coping strategies, leading to unawareness of or even unresponsiveness to psychological distress. What makes the grief process complicated in the study context is that men may encounter more cultural or societal barrier in seeking medical help for mental health problems [45, 47]. In addition, given that the fathers may be the only breadwinner in the families of deceased children (see Table 1 for the relatively lower employment rate for mothers), the bereaved fathers likely accommodate the child loss with more work (or income) in order to regain the order of life, leaving no time for grief or selfcare [53‐55].

Although the relationship between child loss and parental adverse health condition is fairly well established, the magnitude of bereavement effects is potentially complex because of parents’ attributes [17, 22, 54]. Our study illuminates that the parental bereavement effect may vary by prior health condition; specifically, the increased risk of treatment seeking for MDD was notably pronounced in those without prior history of MDD or CPC. Three probable explanations may account for this phenomenon. First, although child loss is a traumatic event, child-rearing, especially for the kids with catastrophic illness, is often a long and even financially and emotionally demanding process [56, 57]. Some parents may be aware of psychological distress and seek treatment long before the child’s death, as indicated by the work reporting that parental utilization of antidepressants or anxiolytics was gradually elevated in the three years before they lost a child to diseases [22]. This possibility is further reinforced by our analyses showing that prior prevalence of MDD treatment among bereaved parents was 4–6-times higher than their comparison peers. Second, for parents with physical or mental health problems (e.g., MDD), a visit to a doctor can offer a great opportunity to call for medical attention to somatic problems or stress-related symptoms (e.g., adjusting medication or therapy); thus, subsequent risk of MDD treatment associated with child loss may be alleviated. This also implicated the efficacy of timely medical intervention for parental bereavement before a child’s death. Third, the observation may reflect the delay in treatment seeking and the rebound after loss of their child. Preoccupation with care work (especially for the severely ill children) may face time or financial constraints, increasing the barriers to accessing appropriate healthcare [53].

Children’s health condition has long been recognized as a source of stressful life events substantially shaping child-rearing adults’ mental health, which is particularly true when children were affected by chronic or catastrophic illness, such as cancer [56‐58]. However, in our analyses, diagnosis of catastrophic illness, a proxy for expectedness or disease chronicity, was not statistically associated with the hazard of MDD treatment seeking after we took into listed covariates into account. Even granting the existing confounding effects of family and parental characteristics, the observation can be the results of a wide heterogeneity in the subgroup of “not having catastrophic illness”, in terms of cause of death and the time from the index hospital visit to death. Future research linking other datasets (e.g., death certificates) can be of help in delineating how the nature of child’s disease or death may affect the process of parental grief and bereavement in our study population.

Some potential limitations of the present study should be acknowledged before further interpretation. First, the definition of exposure (i.e., child’s death) was primarily identified from health insurance records (e.g., insurance withdrawal), therefore restraining our ability to further characterize the heterogeneity of child’s death (e.g., cause of death) [22, 59]. In addition, we only used treatment seeking for MDD as a proxy for bereavement responses which may not be representative of a broad array of mental health sequelae (e.g., anxiety disorders, substance use disorder, and suicidal behaviors). It is possible that the observed association linking a child’s death and parental MDD treatment seeking may be partially attributed to other unadjusted variables which cannot be ascertained in the NHIRD, including family- (e.g., wealth), parents- (e.g., educational attainment), child- (e.g., low birthweight), and parent-child domains (e.g., quality of relationship). Another issue involves specification of parent-child pairs in relation to health insurance status. The insurance premium payer is usually determined by several factors (e.g., labor force participation and salary), some of which also have been found as correlates for bereavement [5]. The differential linkage rate in mother-child pairs (i.e., 51% in the deceased vs. 70% in the living, see Supplementary Fig. 1) may possibly result in the underestimated risk of MDD for the bereaved parents. Finally, the sole reliance on health insurance data can be problematic when only variables relevant to reimbursement can be gathered. Further exploration of variables at the child-, parents-, and household-levels was not feasible in this study; they should be evaluated in future research to better understand possible mechanisms.

Notwithstanding the limits just described, the present study has several strengths. To date, our study represents one of just a few population-based studies conducted in non-Western societies to investigate so-called parental bereavement effects in one society characterized by the very low (if not the lowest) birth rates. The large sample size and three-year follow-up period granted us the opportunity to delineate the effects of a child’s death on fathers and mothers with a rather thorough perspective in risk variation in individual health condition. Societies or cultures with similar parent-child bonding may benefit from this study in framing the concept of parental bereavement, which is particularly urgent in the context wherein mental health awareness is relatively limited and the need in mental healthcare is often overlooked.

Our findings provide compelling evidence for an association between losing a young child to death and increased MDD in the subsequent three years. Timely intervention to reduce psychological distress in parents who have a child died depends on recognition of reaction and adaptation (e.g., withdrawal from social interaction) by pediatric services providers in inpatient and emergency department or home visits [5, 38, 57, 60]. For many parents, particularly those in disadvantaged economic or social network groups, the intervention may also involve financial aid and social support (e.g., patient groups). Effective strategies addressing enduring bereavement may require family-based, integrated physical and mental healthcare, and even extended counseling service long before the death of a child (e.g., pediatric hospice or intensive care unit). Additional studies of follow-up service delivery of intervention for those bereaved parents or caregivers are warranted, especially those who have received little medical attention in the process of caring for a seriously ill child or limited mental health literacy.