Effects of Parenting Adults with Disabilities on Later-Life Health: The Role of Intergenerational Ambivalence

Abstract

According to family systems theory, strains from parenting an adult with disabilities may spill over to parents’ relationships with their other children and disrupt family dynamics and their well-being in later life. This study examined whether parental ambivalence toward their non-disabled children is greater in families of adults with disabilities [developmental disabilities (DD) or serious mental illnesses (SMI)] than families without any adult children with disabilities. The study also investigated whether ambivalence mediates the associations between having an adult child with DD or SMI and parents’ health. Data were from the 2011 Wisconsin Longitudinal Study in which aging parents (M_age = 71; n = 6,084) were asked about their relationship with each of their adult children. Multilevel regression models and multilevel structural equation models were estimated to analyze the data. Our findings showed that parents of an adult with SMI felt greater ambivalence toward their non-disabled adult children than comparison group parents of adult children without disabilities, whereas no significant differences were found between parents of an adult child with DD and comparison group parents. Parental ambivalence toward their non-disabled adult children played a significant indirect role in the negative association between having a child with SMI and parental physical and mental health. The findings have implications for clinical practice with aging families of adults with disabilities and suggest the need for additional research to better understand intergenerational parent-adult child dynamics in these families.

Introduction

Parenting a child with developmental disabilities (DD) or serious mental illnesses (SMI) is known to be stressful, taking a long-term toll on parents’ mental and physical health. Older parents who have an adult child with such conditions report elevated levels of depressive symptoms, lower levels of quality of life, and a greater number of physical symptoms than their age peers whose children do not have any disability (Aschbrenner, Greenberg, & Seltzer, 2009; Namkung, Greenberg, Mailick, & Floyd, 2018). According to family systems theory, the strains of providing care to a child with disabilities may spill over and affect relationships between caregiving parents and other family members, including their other children, which in turn may have long-term effects on parental well-being (Fingerman, Pitzer, Lefkowitz, Birditt, & Mroczek, 2008).

Gerontological researchers have generally focused on positive sentiments, such as feelings of solidarity or closeness to evaluate the quality of parent-adult child relationships. However, some researchers have found that ambivalence, or the co-occurrence of positive and negative sentiments, may better capture the complex emotional aspects of aging parent-adult child relationships than measuring closeness alone (Luescher & Pillemer, 1998). Ambivalence has been operationalized using two distinct approaches: a direct and an indirect approach (Suitor, Gilligan, & Pillemer, 2011). The direct approach asks parents to report their level of ambivalence through a series of direct questions such as whether they “have mixed feelings” and “feel torn or conflicted” in the relationship (Pillemer & Suitor, 2002, p. 607; Pillemer et al., 2007, p. 782). The indirect approach asks parents to report positive and negative feelings separately, and a standard algorithm is applied to represent the balance between these two sentiments (Suitor et al., 2011). Across both methods, greater parental ambivalence has been found to be associated with poorer parental health and well-being (Fingerman et al., 2008; Gilligan, Suitor, Fled, & Pillemer, 2015).

Feelings of ambivalence may be high in families of adults with disabilities because of the unique nature of the caregiving context. Aging parents and their non-disabled adult children share a common concern about the welfare of their family member with disabilities, possibly bringing the family closer together (Seltzer, Greenberg, Orsmond, & Lounds, 2005; Taylor, Greenberg, Seltzer, & Floyd, 2008). At the same time, the stressful behavior of the adult with disabilities may create tensions and conflicts in the relationships between aging parents and their non-disabled other children (Greenberg, Seltzer, Orsmond, & Krauss, 1999). Ultimately, these processes may take a toll on parental mental and physical health.

The goals of the current study were to examine (1) whether parental ambivalence is greater in aging families of adults with disabilities and (2) whether parental ambivalence is one underlying mechanism that helps shed light on why having a child with disabilities affects the well-being of the parents in later life. Specifically, using data from the 2011 wave of the Wisconsin Longitudinal Study (WLS) that tracked 6,084 aging parents who had 18,460 non-disabled adult children, this study examined whether parental ambivalence about the relationship with their non-disabled children was greater in families of adults with disabilities than families without any adult children with disabilities. This study also investigated whether ambivalence mediated the effects of having an adult child with DD or SMI on parents’ mental and physical health.

Intergenerational Ambivalence in Families of Adults with Disabilities

Socioemotional Selectivity Theory (SST) proposes that as individuals age, they become increasingly selective, investing greater time and resources in emotionally meaningful relationships, such as relationships with their adult children (Carstensen, 1992). SST also suggests that such greater emotional investments in relationships may increase the intensity and frequency of both positive and negative interactions. Consistent with this theory, research has shown that intergenerational ambivalence is prevalent in later life families (Gilligan et al., 2015).

It is well established that adult children serve as a ‘safety net’ as parents age and anticipate becoming physically less able to live independently. Parents of adult children with disabilities may have a greater investment in maintaining supportive relationships with their non-disabled adult children because these adult children represent not only a safety net for the parents but also are a safety net for the adult with disabilities who will continue to need support after their parents’ deaths (Seltzer et al., 2005). Research has found that siblings are more likely to feel close to and remain involved in the lives of their siblings with disabilities when they feel close to their aging parents (Greenberg et al., 1999).

Yet there are many unique challenges faced by families of children with disabilities that may give rise to relational conflict. There is a degree of ambiguity around how much care and support a person with a disability needs (Ingersoll-Dayton, Neal, Ha, & Hammer, 2003). This ambiguity may create disagreements between parents and their unaffected children about their parents’ levels of involvement in the life of their child with a disability. Ambiguity also arises around expectations of future caregiving involvement of the unaffected children for their siblings with disabilities. Aging parents of adults with disabilities may hope that their unaffected adult children will become the next generation of caregivers for their adult child with DD or SMI (Hatfield & Lefley, 2000); however, the willingness of the non-disabled children to take on a future caregiving role varies considerably (Smith, Greenberg, & Seltzer, 2007). In addition, long-term parenting for a child with DD or SMI may mean that parents have been unable to invest equally in their relationships with all of their adult children (Hodapp, Urbano, & Burke, 2010; Seltzer et al., 2005). This situation may lead to feelings of resentment among non-disabled children and produce conflict or tension in their relationships with their parents.

Differences between Parents and Siblings of Adults with Developmental Disabilities and Serious Mental Illness

Levels of ambivalence, as well as the effects of ambivalence on parental well-being, may differ for parents of adult children with DD versus parents of adult children with SMI because of differences in the nature of DD and SMI. These two disabilities differ most notably with respect to the age of onset and the nature and course of the symptoms (Parker et al., 2015). DD is typically identified at birth or in the first few years of life, and thus, for unaffected siblings, norms of caregiving and family solidarity may develop early in their lives and persist throughout the life course to have enduring effects on family relationships (Essex, 2002; Seltzer et al., 2005). Indeed, compared with siblings of non-disabled adults, adult siblings of individuals with DD have more contact with their family of origin and are more likely to live in the same state as their affected siblings (Taylor et al., 2008). In contrast, SMI is typically diagnosed during adolescence or young adulthood, a time in the family life course when youth begin to individuate from the family. Seltzer and colleagues (1997) found evidence suggesting that siblings of persons with SMI may cope by distancing themselves from their family of origin, which may have long-term consequences for relationships with their parents.

Additionally, the nature of the caregiving role and context may be quite different for families of persons with SMI versus families of persons with DD. Family caregivers of persons with DD face very different sets of demands than those whose family member has SMI, which in turn may differentially affect the well-being of the caregivers and family relationships (Parker et al., 2015). Some individuals with DD require substantial hands-on care by family members, and their care needs and symptoms remain relatively stable in adulthood (Woodman, Smith, Greenberg, & Mailick, 2015). On the other hand, caring for a child with SMI typically focuses on the provision of supervision (Parker et al., 2015). The need for supervision fluctuates considerably over time, given the cyclical nature of the mental illness, which introduces considerable uncertainty into family life. High levels of uncertainty are known to take a toll on health (Greco & Rogers, 2003).

Second, families of persons with SMI tend to be more stigmatized than families of persons with DD, and stigma is known to be related to lower levels of family functioning (Muralidharan, Lucksted, Medoff, Fang, & Dixon, 2016; van der Sanden, Bos, Stutterheim, Pryor, & Kok, 2013). Third, a unique challenge of caring for an adult with SMI is understanding whether a problematic behavior or symptom is an inherent aspect of the illness or under the individual’s control. Differences among family members in their attributions concerning the degree of control the adult with SMI has over his or her symptoms may contribute to greater relationship strain or conflict (Seltzer, Greenberg, Floyd, Pettee, & Hong, 2001), giving rise to greater ambivalence. Therefore, due to these differences in the caregiving context between families of persons with DD and families of persons with SMI, we examined parental ambivalence toward non-disabled adult children and its effects on parental well-being separately for parents of adults with DD versus parents of adults with SMI.

Ambivalent Relationships and Parental Health Outcomes

A growing body of research has found that greater parental ambivalence is associated with poorer parental health, particularly with respect to mental health outcomes. Specifically, research has shown that aging parents who experience greater ambivalence, assessed directly or indirectly, have more depressive symptoms and lower levels of life satisfaction, quality of life, and positive affect (Fingerman et al., 2008; Gilligan et al., 2015; Lowenstein, 2007). Fingerman and colleagues (2008) suggested that experiencing ambivalence may be more detrimental and difficult to cope with than negative emotions alone because ambivalent relationships tend to be less predictable and offer both more emotional rewards and more costs. Uchino and colleagues (2004) similarly found that ambivalent relationships were more highly associated with depressive symptoms than were solely negative relationships. Thus, these findings suggest that parental ambivalence might account in part for the poor mental and physical health outcomes experienced by parents of adults with disabilities.

Other Parental and Adult Child Characteristics Associated with Ambivalence

Research with non-disability samples has identified a number of parental and adult child characteristics that are associated with parental ambivalence, measured both directly and indirectly. Therefore, these characteristics are controlled for in the present study. Parent characteristics associated with ambivalence include the number of adult children, parental age, education, marital status, and functional limitations. Aging parents who have a greater number of adult children report greater ambivalence, assessed indirectly, in their relationships with their adult children than aging parents with fewer adult children (Birditt, Fingerman, & Zarit, 2010; Willson, Shuey, Elder Jr, & Wickrama, 2006). Aging parents who are better educated and who have more functional limitations perceive greater ambivalence in their relationships with their adult children (Cichy, Lefkowitz, Davis, & Fingerman, 2013; Sechrist, Suitor, Howard, & Pillemer, 2014; Silverstein, Gans, Lowenstein, Giarrusso, & Bengtson, 2010), irrespective of whether ambivalence is measured directly or indirectly. The findings regarding the association of parental age to levels of ambivalence have been inconclusive. Ha and Ingersoll-Dayton (2008) found that older parents reported greater ambivalence, measured indirectly, after adjusting for the parent’s and children’s perceived dependency on one another. Lendon and colleagues (2014), on the other hand, found that age was negatively associated with an indirect measure of ambivalence, but this relationship was not found for the direct measure of ambivalence. Other researchers found no relationship between parental age and indirect measures of ambivalence (Fingerman, Chen, Hay, Cichy, & Lefkowitz, 2006; Kiecolt, Blieszner, & Savla, 2011). Some studies employing an indirect measure of ambivalence reported that parents who are not currently married expressed greater ambivalence in relationship to their adult children than those who are married (Kiecolt et al., 2011; Willson et al., 2006), but other studies using a similar measure of ambivalence did not find such a relationship (Birditt et al., 2010; Ha & Ingersoll-Dayton, 2008).

Child characteristics including gender, age, and geographical proximity to parents influence parents’ reports of ambivalence about their adult children, regardless of the specific measure of ambivalence or whether it is measured directly or indirectly. Parents reported feeling greater ambivalence about the relationship with their children who are daughters, are younger, and live closer, as proximity fosters more frequent interaction but also greater emotional intensity (Birditt, Miller, Fingerman, & Lefkowitz, 2009; Birditt et al., 2010; Pillemer, Munsch, Fuller-Rowell, Riffin, & Suitor, 2012; Willson et al., 2006). Parents perceived greater ambivalence when the child was not married, had a lower level of education, and was unemployed (Fingerman et al., 2006; Pillemer et al., 2007; Willson et al., 2006). Finally, parents reported feeling greater ambivalence in their relationships with adult children who had become parents, perhaps because of tensions arising from different generational parenting styles (Kiecolt et al., 2011), although other studies did not find this effect (Birditt et al., 2010; Pillemer & Suitor, 2002).

The Current Study

This study examined whether parents of adults with DD or SMI report higher levels of ambivalence about their non-disabled other adult children than a comparison group of parents who did not have any children with disabilities. We also examined whether feelings of ambivalence mediated the association between having a child with disabilities and parental mental and physical health. Specifically, the following hypotheses were investigated:

H1. Parents of adults with disabilities, especially those of adults with SMI, will report higher levels of ambivalence about their non-disabled adult children relative to comparison group parents, controlling for parental and adult child characteristics.

H2. Parental ambivalence about their relationship with their non-disabled children will mediate the association between having a child with DD or SMI and parental mental and physical health, controlling for parental and adult child characteristics. We expected that this mediation effect would be more evident in the highest stress condition (i.e., in contrast between parents of adults with SMI and the comparison group parents than in contrast between parents of adults with DD and the comparison group parents).

Method

Data and Sample

This study used data from the Wisconsin Longitudinal Study (WLS), a population-based longitudinal study that followed 10,317 Wisconsin high school graduates of 1957 and their randomly selected siblings (Herd, Carr, & Roan, 2014). Over 99% of the WLS sample respondents were white, reflecting the racial composition of Wisconsin in 1957. About 74% of the original 1957 graduate sample and 56% of the sibling sample participated in the 2011 survey, the most recent data collection, after adjusting for mortality. The WLS identified respondents who had a child with DD or SMI using an extensive screening process. In data collection waves conducted in 2004/06 and 2011, screener questions were administered to examine if any of the parent respondents’ children had DD or SMI. DD conditions included specific diagnoses of developmental disabilities (e.g., Down syndrome, cerebral palsy, autism spectrum disorder), parental report of a child’s “developmental disability,” “mental retardation,” or “cognitive disability,” or a child’s low levels of education due to a developmental disability. Conditions of SMI were limited to schizophrenia, bipolar disorder, and major depression if the depression required hospitalization or interfered with the adult child’s ability to work or live independently.

This study used the 2011 wave of the WLS in which respondents were asked to identify by name each of their living children, and parent respondents were asked a series of questions about the nature and quality of their relationship with each child. In 2011, 7,285 respondents had at least one living child and completed the set of questions about the parent-child relationship. The sample was further restricted to (1) parents who had at least two biological or adopted children (n = 6,596), and then to (2) parents for whom at least one biological or adopted child had no known DD, SMI, or any other chronic condition (i.e., non-disabled children) (n = 6,574). The two-child criterion was applied so that all parents in the sample had at least one non-disabled adult child. Additionally, 490 parents were excluded because of ambiguous or missing information about one or more of their children’s disability status.

The final analytic sample includes 18,460 non-disabled adult children (M_age = 44) nested within 6,084 aging parents (M_age = 71). We omitted step-children given previous literature suggesting strained relationships between step-parents and adult children (Rossi & Rossi, 1990; Ward, Spitze, & Deane, 2009). In the group of families of adults with DD, 335 non-disabled adult children were nested within 134 parents. In the group of families of adults with SMI, 726 non-disabled adult children were nested within 308 parents. The remaining 17,399 adult children were nested within 5,642 parents who did not have any children with a disability, and they served as the comparison group. The primary conditions of individuals with DD were intellectual or developmental disabilities due to unspecified causes (n = 56), cerebral palsy (n = 19), autism spectrum disorder (n = 18), Down syndrome (n = 17), other specific neurodevelopmental disorders (n = 16), and brain injury before age 22 (n = 8). The primary conditions of adult children with SMI included bipolar disorder (n = 198), schizophrenia (n = 55), and major depressive symptoms that required hospitalization or interfered with their ability to work or live independently (n = 55).

Measures

Independent Variable.

The independent variable was the family type - whether a parent had an adult child with DD or SMI (1 = parents of an adult child with DD, 2 = parents of an adult child with SMI, 0 = parents without any adult child with known disabilities). This three-category variable was dummy coded in mediation analyses.

Ambivalence.

The WLS only included items to compute an indirect measure of ambivalence. Parents reported separately for each of their children about their positive and negative feelings toward the child. Parental ambivalence was operationalized as the extent to which older parents perceive their relationship with a given child as simultaneously both positive (i.e., affectual solidarity) and negative (i.e., affectual conflict). Affectual solidarity was assessed with the following three questions: “How much is the child willing to listen when you need to talk about worries or problems?”, “How much does the child make you feel loved and cared for?”, and “How close are you to the child?” Affectual conflict was assessed with the following two questions: “How much do you feel that the child makes too many demands on you?” and “How much is the child critical of what you do?” All five items were rated from 1 (not at all) to 5 (a great deal), and mean scores were calculated for solidarity and conflict. The Cronbach’s alphas were .86 for solidarity and .53 for conflict. A low Cronbach’s alpha for the conflict component of ambivalence has been reported in other studies (Ha & Ingersoll-Dayton, 2008; Suitor et al., 2011; Lendon et al., 2014) and is common for scales with few items.

The level of ambivalence was calculated using Griffin’s Similarity and Intensity of Components Formula (Thompson, Zanna, & Griffin, 1995): Empirically, this measure captures the intensity (i.e., the sum of the mean scores on solidarity and conflict scales should be high to get a high ambivalence score) and the similarity of the two sentiments (i.e., the difference between the ratings of solidarity and conflict should be small to get a high ambivalence score). Adding 1.5 points is included in the formula so that the values range between positive values (range from 0.5 to 6.5). This formula has been widely used in research examining intergenerational ambivalence (e.g., Gilligan et al., 2015; Lee & Szinovacz, 2016).

Parental Health Outcomes.

Six measures of parental mental and physical health were examined. Depressive symptoms were measured by the Center for Epidemiological Studies Depression Scale (CES-D; Radloff, 1977) consisting of 20 items, each of which assessed how many days in the past week respondents experienced a specific symptom. The data were recoded into four categories (0 = never, 1 = 1–2 days, 2 = 3–4 days, 3 = 5–7 days), consistent with the = 0.85), with higher scores indicating more depressive symptoms. For psychological well-being, a modified version of Ryff’s Psychological Well-Being measure (Ryff, 1989) was used. The measure encompassed six domains of current psychological well-being: Self-Acceptance, Positive Relations with Others, Autonomy, Environmental Mastery, Purpose in Life, and Personal Growth. This measure consists of 19 items, each of which was rated on a scale of 1 to 6 (1 = agree strongly, 6 = disagree strongly). The total score was the sum of the 19 items (α = 0.88), with higher scores indicating better overall psychological well-being.

Mental health and physical health were also assessed with the 12-Item Short-Form Health Survey (SF-12), a highly reliable and well-validated measure that produces two continuous summary subscale scores, one for physical health and the other for mental health during the past four weeks (Ware, Kosinski, & Keller, 1996). Sample items for the mental health subscale included how much of the time during the past four weeks the respondent felt calm and peaceful, had a lot of energy, and felt downhearted and blue (α = 0.73). The standard algorithm was applied (Ware et al., 1996) to derive the SF-12 mental health subscale score, which ranged from 0 to 100, with higher scores indicating better overall mental health. The SF-12 physical health subscale covers different health domains such as bodily pain, vitality, and physical health functioning (α= 0.82). Again, the standard algorithm was applied to compute the SF-12 physical health subscale score ranging from 0 to 100, with higher scores indicating better overall physical health (Ware et al., 1996).

Other measures of physical health included self-rated health and self-report physical symptoms. For self-rated health, respondents were asked to rate their health at the present time (range = 1–5; 1 = very poor, 5 = excellent). Number of self-reported physical symptoms was measured by a count of the number of symptoms the respondent experienced in the past six months from a list of 16 symptoms - a lack of energy, fatigue/exhaustion, headache, dizziness/faintness, numbness, ringing in the ears, upset stomach, constipation, diarrhea, aching muscles, stiff/swollen joints, back pain/strain, chest pain, shortness of breath, excessive sweating, and skin problems.

Parent-Level Covariates.

Covariates included parents’ sociodemographic characteristics, including gender (1 = mother, 0 = father), age (in years), education (in years), marital status (1 = currently married, 0 = currently unmarried), and the number of adult children alive in 2011. In order to control for possible familial associations for long-standing mental health problems in the affected families, we assessed parents’ mental health history using a question asking the respondents if they had ever been diagnosed with a mental illness (1 = has been diagnosed with a mental illness, 0 = never been diagnosed). Parents’ current functional ability was assessed by whether parents had limitations in activities of daily living (ADLs), which includes difficulty dressing; walking across a room; bathing or showering; eating; getting in or out of bed; and using the toilet, because of a health or memory problem, coded as 1 = at least one ADL limitation, 0 = no ADL limitation to adjust for left skewness and to be consistent with prior studies (e.g. Sekaran, Choi, Hayward, & Langa, 2013).

Adult Child-Level Covariates.

The demographic and socioeconomic characteristics of the non-disabled adult children were controlled for: gender (1= daughter, 0 = son), age (in years), biological relationship to the sampled parent (1 = biological, 0 = adopted), marital status (1 = married, 0 = unmarried), parental status (1 = have their own child, 0 = have no child), work status (1 = currently working part-time or full-time, 0 = currently not working), education (1 = ever attended college, 0 = never attended college), and geographical proximity between a parent and adult child. Parents were asked how many miles away each child lived, and this continuous variable was recorded as a categorical variable to adjust for left skewness (1 = same house, 2 = up to 5 miles, 3 = up to 10 miles, 4 = up to 20 miles, 5 = up to 50 miles, 6 = up to 300 miles, 7 = further than 300 miles). Parents’ and their non-disabled children’s characteristics by family type (i.e., whether a parent had an adult child with DD or SMI) are presented in Table 1.

Table 1.

Sample Characteristics by Family Type

	DD	SMI	Comparison	Group Differencea
	M (SD), %	M (SD), %	M (SD), %
	(n = 134)	(n = 308)	(n = 5,642)
Parent-level covariates
Age	71.1 (4.6)	70.5 (3.9)	70.8 (3.9)	13 (2)
Gender (1 = mother)	49%	63%	53%	12.2 (2)**	(b > c)
Education	13.9 (2.7)	14.2 (2.5)	13.7 (2.4)	6.4 (2)**	(b > c)
Marital status (1 = currently married)	76%	67%	77%	18.96 (2)***	(b < c)
Number of children alive	3.8 (1.9)	3.6 (1.4)	3.3 (1.4)	14.6 (2)***	(a, b > c)
Ever diagnosed as MI (1 = yes)	5%	11%	3%	68.0 (2)***	(b > c)
Has ADLs limitations (1 = yes)	15%	15%	9%	14.7 (2)**	(a, b > c)
Parental well-being
Depressive symptoms	8.3 (6.9)	9.5 (8.8)	7.9 (7.3)	6.43 (2)**	(b > c)
Psychological wellbeing	89.6 (13.4)	89.5 (13.4)	91.1 (12.4)	3.11 (2)*	(b < c)
SF-12 mental health	55.4 (7.6)	54.1 (8.2)	56.1 (6.6)	11.94 (2)***	(b < c)
Self-rated health	3.5 (1.0)	3.5 (1.0)	3.7 (0.9)	6.68 (2)**	(b < c)
Number of physical symptoms	5.5 (3.4)	5.8 (3.5)	5.0 (3.5)	10.1 (2)***	(b > c)
SF-12 physical health	44.3 (11.3)	44.8 (11.5)	46.7 (10.5)	6.93 (2)**	(a, b < c)
	(n = 335)	(n = 726)	(n = 17,399)
Child-level covariates
Age	43.5 (7.6)	43.8 (6.3)	43.8 (6.4)	0.4 (2)
Gender (1 = daughter)	50%	52%	49%	15 (2)
Relationship to parent (1 = biological)	91%	93%	97%	49.8 (2)***	(a, b < c)
Marital status (1 = married)	68%	67%	72%	7.5 (2)*	(b < c)
Parental status (1 = yes)	80%	75%	77%	3.2 (2)
Work status (1 = currently working)	90%	85%	90%	20.9 (2)***	(b < c)
Education (1 = ever attended college)	57%	66%	67%	12.7 (2)**	(a < c)
Geographical proximity	4.7 (2.1)	4.7 (2.1)	4.6 (2.0)	3.1 (2)*
Parent-child ambivalence
Ambivalence	1.7 (1.0)	1.7 (1.1)	1.6 (1.0)	8.85 (2)***	(b > c)

DD = Developmental Disabilities; SMI = Serious Mental Illness; MI = Mental Illness; ADL = Activities of Daily Living

Note.

^aDifferences of families of an adult child with DD or SMI from comparison group families are noted when the overall three-group differences were significant at p < .05; (a) represents DD, (b) represents SMI, and (c) represents comparison group.

^*p<.05,

^**p<.01,

^***p<.001

Analysis Plan

To address our first research question, whether parents of an adult with disabilities were likely to report greater ambivalence about their relationship with non-disabled other children than were comparison group parents, we conducted multilevel regression analyses. In these analyses, the 18,460 non-disabled adult children who were the units of analyses were nested within 6,084 families. Multilevel regression accounts for this lack of independence. Based on a stepwise approach, we examined the effect of having an adult with DD or SMI without any covariates in Model 1, added parent-level covariates (e.g., parental age, marital status, ADL limitations) in Model 2, and entered adult child-level covariates (e.g., age, education, employment) in Model 3. We also let the child-level residuals have different variances for different family groups (i.e., heteroscedastic child-level residuals or different within-family variances for DD, SMI, and comparison families) and relaxed the default homoscedastic assumption in multilevel models. This multilevel regression analysis was conducted using the xtmixed command in STATA/SE 14.2.

To address our second research question, whether ambivalence mediates the negative association between having an adult child with DD or SMI and parental health, we conducted multilevel structural equation model (MSEM) analysis (Preacher, Zyphur, & Zhang, 2010). Our research question examines how a Level-1 variable (i.e., parental ambivalence about their relationship with each child) predicts a Level-2 variable (i.e., parental health), which is a so-called “micro-macro situation” (Snijders & Bosker, 2012) or “upward influence” (Griffin, 1997) in multilevel modeling. Our model design, in particular, is termed a 2-1-2 design as the predictor (X_j) and outcome (Y_j) are assessed at Level-2 but the mediator (M_ij) is assessed at Level-1 (Preacher et al., 2010). The model for 2-1-2 data is a special case of the general MSEM framework with a random intercept for M_ij because the indirect effect is estimated only at the between-level without a within-level variation, as shown in Figure 1.

Figure 1.

Multilevel Mediation Model of Parental Ambivalence in the Association between Family Type and Parental Health

DD = Developmental Disabilities; SMI = Serious Mental Illness

Note. Path A (Parents of an adult child with DD/SMI → Parental ambivalence) was estimated controlling for child-level covariates (i.e., age, gender, biological relationship to parent, geographic proximity to parent, marital status, parental status, education, and working status) and parent-level covariates (i.e., age, gender, education, number of living children, marital status, Activities of Daily Living (ADLs) limitation, and history of mental health diagnosis); Path B (Parental ambivalence → Parental mental and physical health) was estimated controlling for child age and biological relationship to parent, as well as all parent-level covariates. ADLs limitation of parents was controlled only for depressive symptoms, psychological well-being, and the SF-12 mental health models, and parents’ history of mental health diagnosis was controlled only for self-rated health, physical symptoms, and the SF-12 physical health models.

In our MSEM analysis, we adjusted for both parent- and child-level covariates as we did for the previous multilevel regression models. Based on the same stepwise approach, we tested the mediation models without any covariates (Model 1), added parent-level covariates (Model 2), and then entered adult child-level covariates (Model 3). We estimated Path A (having an adult child with DD or SMI → parental ambivalence) controlling for the same parent and child covariates in the multilevel regression analyses to address our first research question. To estimate Path B (parental ambivalence → parental health outcomes), we controlled for parental age, gender, education, number of living children, marital status, and limitation on ADLs (only for mental health outcomes) or history of mental diagnoses (only for physical health outcomes) that are known to be associated with aging parents’ mental and physical health (e.g. Fingerman et al., 2008; Gilligan et al., 2015). We also controlled for child age and biological relationship to the sampled parents (e.g., Kiecolt et al., 2011). Child gender was not controlled for on Path B due to the lack of between-level variations of this variable.

To test the presence of significant mediation or indirect effects, Path A and B coefficients were multiplied (i.e., β_XM x β_MY) (Kenny, 2017; Preacher & Kelley, 2011). Monte Carlo simulations were conducted to obtain standard errors (Hayes, 2013; Selig & Preacher, 2008). The total effect (Path C) of having a child with DD/SMI on parental health is a combination of the direct effect (Path C’) and the indirect effect via the mediator, “ambivalence.” The mediation analysis was performed using MPlus, and Monte Carlo simulations were conducted using R code developed by Selig & Preacher (2008).

Results

Descriptive Statistics

As shown in Table 1, a higher percent of the parents of adults with SMI were mothers than in the comparison group. Parents of an adult child with SMI had more years of education than parents in the comparison group, and they were less likely to be currently married than comparison parents and parents of an adult child with DD. Parents of an adult child with DD and parents of an adult child with SMI had a greater number of children than comparison group parents. Parents of an adult child with SMI also were more likely to have been diagnosed with a mental illness in their lifetime than comparison group parents. Out of 186 parents who had a history of mental illness and reported a specific diagnosis, most were diagnosed with depression (67%), followed by anxiety (22%). Greater proportions of parents of a child with DD or SMI reported that they had one or more ADL limitations than comparison group parents.

Parents of adults with SMI showed poorer mental and physical health outcomes than comparison group parents, reporting more depressive symptoms, lower levels of psychological well-being and SF-12 mental health, poorer self-rated health, a greater number of physical symptoms, and lower levels of SF-12 physical health. Parents of adults with DD also reported lower levels of SF-12 physical health than the comparison group of parents, but the two groups of parents were otherwise comparable.

Characteristics of non-disabled adult children were also different between the three groups. Adult children whose siblings had DD or SMI were less likely to be biological children and more likely to be adopted children of their parents than those in the comparison group. Non-disabled siblings of adults with SMI were less likely to be married and currently working than those in the comparison group. Siblings of an adult with SMI were also less likely to be currently working than siblings of an adult with DD. Non-disabled siblings of adults with DD were less likely to have attended college than those in the comparison or SMI groups. Although geographical distances between parents and their non-disabled children were significantly different across the three family groups overall, none of the post hoc comparisons was significant. Compared to those of adults without disabilities, aging parents of adults with SMI reported greater ambivalence in their relationships with their non-disabled children.

Multilevel Analyses of the Associations between Family Type and Parental Ambivalence

Results from multilevel regression analyses also showed that aging parents of adults with SMI reported greater ambivalence about their relationship with their non-disabled adult children than comparison group parents (B = 0.14, p < .01; see Model 1 in Table 2). After controlling for parent characteristics, the difference between parents of adults with SMI and comparison group parents remained statistically significant (B = 0.12, p < .05 in Model 2), and continued to be significant when child characteristics were entered in Model 3 (B = 0.11, p < .05).

Table 2.

Multilevel Models on Parental Ambivalence

	Model 1		Model 2		Model 3
	B	SE	B	SE	B	SE
Fixed part
Intercept	1.60***	0.01	3.07***	0.22	3.38***	0.22
Family with an adult with DD a	0.09	0.08	0.08	0.08	0.07	0.08
Family with an adult with SMI a	0.14**	0.05	0.12*	0.05	0.11*	0.05
Parental covariates
Mother b			−0.20***	0.02	−0 17***	0.02
Parental age			−0.01***	0.00	−0.01*	0.00
Parental education			−0.01**	0.00	−0.01	0.00
Number of children			−0.03**	0.01	−0.03***	0.01
Currently married c			−0.13***	0.03	−0.09***	0.03
Ever been diagnosed with MI d			0.29***	0.06	0.27***	0.06
Has ADLs limitations e			0.30***	0.04	0.28***	0.04
Child covariates
Child age					−0.00**	0.00
Daughter f					−0.03*	0.01
Biological child g					−0 24***	0.04
Geographic proximity					−0 04***	0.00
Child is currently married h					−0.21***	0.02
Child has own children i					0.05**	0.02
Child currently works j					−0.15***	0.02
Child has attended college k					−0.09***	0.02
Random part
Intercept (ζ1j)	0.70		0.69		0.67
Residual (ϵij)
Family with an adult with DD	0.80		0.79		0.77
Family with an adult with SMI	0.77		0.77		0.78
Comparison family	0.69		0.69		0.67
Log likelihood	−22644.95		−22246.50		−20330.12
Likelihood-ratio test (vs. homoscedasticity of ϵij)2	18.87***		18.68***		24 41***

DD = Developmental Disabilities; SMI = Serious Mental Illness; MI = Mental Illness; ADLs = Activities of Daily Living

Note.

^1.Reference groups are (a) comparison group family, (b) father, (c) unmarried, (d) not been diagnosed as MI, (e) has no functional limitation, (f) daughter, (g) adopted child, (h) currently unmarried, (i) does not have own child, (j) not working, and (k) has not attended college.

^2.Heteroscedasticity at child-level residual (ϵ_ij) means within-family variance differs by group

^*p<.05,

^**p<.01,

^***p<.001

As a follow-up test, we estimated predicted means of ambivalence scores for each parent controlling for parent- and child-level covariates to calculate the effect size of the difference on parental ambivalence across three family groups (i.e., DD, SMI, and comparison). Aging parents of an adult child with SMI (M = 1.73, SD = 0.57) perceived greater ambivalence about their relationships with their non-disabled children than comparison parents (M = 1.57, SD = 0.61) with Cohen’s d = 0.27, indicating a small to medium effect size. In contrast, levels of parental ambivalence about their relationship with their non-disabled children did not differ between aging parents of adults with DD and comparison group parents before and after adjusting for all model covariates. When aging parents of adults with DD and those of adults with SMI were compared, the difference in their levels of ambivalence was not significantly different before or after controlling for covariates (not shown in Table 2).

As shown in Model 3 in Table 2, aging parents perceived greater ambivalence about their relationship with their non-disabled children when they were fathers, were younger, had fewer children, were not currently married, had been diagnosed with a mental illness, and reported any ADL limitation. Aging parents reported feeling greater ambivalence toward sons relative to daughters, younger children relative to older children, and adopted children relative to biological children. Aging parents reported less ambivalence toward their non-disabled adult children who were married, did not have their own children, were working, had attended college, and lived farther from them.

The results from a series of likelihood ratio tests confirmed the heteroscedastic child-level residuals by family type. Each family type had a different variance of child-level residuals (Chi² = 24.41, p < .001 in Model 3). As shown in Random Part, aging parents of adults with DD and SMI showed greater within-family variability than the comparison group.

Mediation Analysis

We estimated separate mediation models for the contrast of parents of an adult with DD and the comparison group and the contrast of parents of an adult with SMI and the comparison group with respect to each of six parental health outcomes. In each model, Path A represents the effect of having a child with DD or SMI on the mediator “parental ambivalence.” Path B estimates the mediator’s effect on the mental and physical health outcome, and Path C’ represents the direct effect of having a child with DD or SMI on the outcome. Path A x Path B estimates the indirect effect, and Path C is the total effect of having a child with DD or SMI on the parental outcome.

Parents of an Adult with DD.

Having an adult child with DD was not significantly associated with any of the three mental health outcomes or with two of the physical health outcomes before and after controlling for the model covariates. The only significant association was found between having a child with DD and lower levels of SF-12 physical health after controlling for child and parental covariates (total effect B = −2.05, p < .05; direct effect B = −2.15, p < .05; not shown in tables). Path A, the association between parenting a child with DD and parental ambivalence, was not significant. Although greater ambivalence predicted negative health outcomes (Path B), the products of the coefficients on Path A and Path B (i.e., indirect effects) were not statistically significant for any of the parental outcomes.

Parents of an Adult with SMI.

As shown in Table 3, the three mediation models on physical health outcomes controlling only for parental level covariates (i.e., Model 2) showed a poor model fit indicated by the Root Mean Square Error of Approximation (RMSEA) > .10, even though all of our models showed acceptable levels of Standardized Root Mean Square Residual (SRMR) both at the between- and within-levels (SRMR < .08; Hu & Bentier, 1999; Preacher et al., 2010; Sideridis, Tsaousis, & Al-Sadaawi, 2018). The overall fit of the model, as measured by the RMSEA, improved substantially in Model 3, indicating the importance of controlling for child-level characteristics. Between-level specific SRMR (ranges from 0.018 to 0.035 for three physical health outcomes) indicates the improved model fit at the between-level as well. In a follow-up analysis, we examined the improvement in fit for each child characteristic. Although several individual child characteristics substantially improved the fit including the child’s education, marital and employment status, and the child’s biological relationship to the parent, no one child characteristic consistently improved the overall model fit across all of the outcome measures based on the difference in the Akaike Information Criterion (Raftery, 1995), chi-square difference tests, and RMSEA. Given the model fit indices, we report the findings based on our final model controlling for parent and child covariates (i.e., Model 3).

Table 3.

Mediation Analysis: Parental Ambivalence as a Mediator between Having an Adult Child with Serious Mental Illness (SMI) and Parental Health

	Model 1 (unadjusted for parent and child covariates)			Model 2 (adjusted for parent-level covariates)			Model 3 (adjusted for parent- and child-level covariates)
	B	SE	95% CI	B	SE	95% CI	B	SE	95% CI
Depressive symptoms
Path A	0.14**	0.05		0.12*	0.05		0.14*	0.06
Path B	3.38***	0.19		3 23***	0.19		3 17***	0.19
Path C: total effect	1.61**	0.52		1.06*	0.48		1.03*	0.49
Path C’: direct effect	1.23*	0.52		0.69	0.48		0.58	0.48
Path A × B: indirect effect	0.48**	0.19	[0.13, 0.85]	0.37*	0.17	[0.03, 0.70]	0.45*	0.20	[0.06, 0.84]
Model fit indices	RMSEA = .000;			RMSEA = .053;			RMSEA = .033;
	SRMRw= .000; SRMRb= .000			SRMRw= .000; SRMRb= .017			SRMRw= .001; SRMRb= .014
Psychological well-being
Path A	0.14*	0.06		0.12*	0.05		0.14*	0.06
Path B	−5 71***	0.28		−5.42***	0.29		−5.34***	0.30
Path C: total effect	−1.62*	0.79		−1.56*	0.75		−1.74*	0.78
Path C’: direct effect	−0.82	0.81		−0.93	0.78		−0.98	0.79
Path A × B: indirect effect	−0.81*	0.31	[−1.42, −0.20]	−0.63*	0.29	[−1.21, −0.06]	−0.77*	0.33	[−1.43, −0.13]
Model fit indices	RMSEA = .000;			RMSEA = .034;			RMSEA = .021;
	SRMRw= .000; SRMRb= .000			SRMRw= .000; SRMRb= .009			SRMRw= .001; SRMRb= .009
SF-12 Mental
Path A	0.15**	0.06		0.12*	0.05		0.15*	0.06
Path B	−1.96***	0.17		−1.98***	0.18		−1.87***	0.18
Path C: total effect	−1.98***	0.50		−1.69***	0.48		−1.56**	0.48
Path C’: direct effect	−1.69**	0.59		−1.46**	0.48		−1.29	0.48
Path A × B: indirect effect	−0.29**	0.11	[−0.51, −0.08]	−0.23*	0.11	[−0.45, −0.03]	−0.27*	0.12	[−0.51, −0.04]
Model fit indices	RMSEA = .000;			RMSEA = .061;			RMSEA = .035;
	SRMRw= .000; SRMRb= .000			SRMRw= .000; SRMRb= .022			SRMRw= .000; SRMRb= .019
Self-rated health
Path A	0.14**	0.06		0.12*	0.05		0.15*	0.06
Path B	−0.25***	0.02		−0.26***	0.02		−0.26***	0.02
Path C: total effect	−0.18**	0.06		−0.19**	0.06		−0.21***	0.06
Path C’: direct effect	−0.14*	0.06		−0.16*	0.06		−0.17**	0.06
Path A × B: indirect effect	−0.04*	0.01	[−0.06, −0.01]	−0.03*	0.01	[−0.06, −0.00]	−0.04*	0.02	[−0.07, −0.01]
Model fit indices	RMSEA = .000;			RMSEA = .184;			RMSEA = .076;
	SRMRw= .000; SRMRb= .000			SRMRw= .000; SRMRb= .038			SRMRw= .000; SRMRb= .024
Physical symptoms
Path A	0.15**	0.06		0.12*	0.05		0.16*	0.06
Path B	1.06***	0.08		1 14***	0.09		1 17***	0.09
Path C: total effect	0.88***	0.21		0.70**	0.21		0.80***	0.22
Path C’: direct effect	0.73**	0.22		0.57**	0.22		0.62**	0.22
Path A × B: indirect effect	0.15**	0.06	[0.04, 0.27]	0.13*	0.06	[0.01, 0.26]	0.18*	0.07	[0.04, 0.33]
Model fit indices	RMSEA = .000;			RMSEA = .130;			RMSEA = .053;
	SRMRw= .000; SRMRb= .000			SRMRw= .000; SRMRb= .029			SRMRw= .000; SRMRb= .018
SF−12 Physical
Path A	0.14**	0.06		0.12*	0.05		0.15*	0.06
Path B	−2 41***	0.25		−2.94***	0.29		−3.04***	0.30
Path C: total effect	−1.89**	0.71		−1.72*	0.69		−2.04**	0.71
Path C’: direct effect	−1.54*	0.71		−1.38*	0.60		−1.57*	0.71
Path A × B: indirect effect	−0.34*	0.14	[−0.62, −0.09]	−0.34*	0.69	[−0.67, −0.03]	−0.47*	0.19	[−0.87, −0.10]
Model fit indices	RMSEA = .000;			RMSEA = .253;			RMSEA = .105;
	SRMRw= .000; SRMRb= .000			SRMRw= .000; SRMRb= .056			SRMRw= .000; SRMRb= .035

SE = Standard errors; CI = Confidence interval based on the Monte Carlo method (available at http://www.quantpsy.org); RMSEA = Root Mean Square Error of Approximation; SRMR_w and SRMR_b = Standardized Root Mean Square Residual, within-and between-level, respectively

Note. Multilevel structural equation modeling was estimated for each health outcome measure. Path A (Parents of an adult child with SMI → Parental ambivalence) was estimated controlling for child-level covariates (i.e., age, gender, biological relationship to parent, geographic proximity to parent, marital status, parental status, education, and working status) and parent-level covariates (i.e., age, gender, education, number of living children, marital status, Activities of Daily Living (ADLs) limitation, and history of mental health diagnosis); Path B (Parental ambivalence → Parental mental and physical health) was estimated controlling for child age and biological relationship to parent, as well as all parent-level covariates. ADLs limitation was controlled only for depressive symptoms, psychological well-being, and the SF-12 mental health models, and history of mental health diagnosis was controlled only for self-rated health, physical symptoms, and the SF-12 physical health models.

^*p<.05,

^**p<.01,

^***p<.001

The total effects of having an adult child with SMI were significant for all six health measures before and after adjusting for covariates. Parenting a child with SMI was associated with greater ambivalence about their relationship with their non-disabled adult children (i.e., Path A) after controlling for the child- and parent-level covariates (B ranges from 0.14 to 0.16 for each of six outcome models in Model 3, p < .05). Greater parental ambivalence was also found to be a significant predictor of higher levels of depressive symptoms, lower levels of psychological well-being and SF-12 mental health, poorer self-rated health, a greater number of physical symptoms, and lower levels of SF-12 physical health (i.e., significant Path B).

The indirect effects of having an adult with SMI via parental ambivalence were significant across three mental health outcomes. Specifically, the indirect effect on depressive symptoms was statistically significant after adjusting for all covariates in Model 3 (B = 0.45, 95% Confidence Intervals [CI] = 0.06, 0.84), indicating that the effect of having a child with SMI on depressive symptoms was mediated through ambivalence. In total, 43.7% of the total effect of having a child with SMI on depressive symptoms was mediated through parental ambivalence (not shown in Table 3). Similarly, the indirect effects of having a child with SMI on psychological well-being (B = −0.77, 95% CI = −1.43, −0.13) and on SF-12 mental health (B = −0.27, 95% CI = −0.51, −0.04) were significant, indicating that the negative association between having a child with SMI and parents’ psychological well-being and SF-12 mental health were mediated by parental ambivalence, explaining approximately 44.3% and 17.3% of the respective association.

The mediation analysis on the three physical health outcomes also indicated that the negative effect of having a child with SMI on self-reported health, physical symptoms, and SF-12 physical health were mediated via parental ambivalence, controlling for parent- and child-covariates. Specifically, the indirect effects of having a child with SMI on self-rated health (B = −0.04, 95% CI = −0.07, −0.01), physical symptoms (B = 0.18, 95% CI = 0.04, 0.33), and SF-12 physical health (B = −0.47, 95% CI = −0.87, −0.10) via ambivalence were significant. Parental ambivalence explained between 19% (self-rated health) to 23% (physical symptoms and SF-12 physical health) of the associations between having a child with SMI and the three physical health outcomes.

Discussion

This study examined whether parental ambivalence was greater among parents of adults with disabilities and investigated parental ambivalence as an underlying mechanism that can explain how having a child with disabilities affects the well-being of the parents in later life. Consistent with our hypothesis, parents of adults with SMI reported higher levels of ambivalence toward their non-disabled other children than parents in the comparison group. The differences between parents of adults with SMI and comparison parents were evident even after controlling for parental and adult child characteristics. This greater parental ambivalence toward non-disabled adult children was not evident in families of adults with DD.

Even though the level of ambivalence among parents of adults with SMI was significantly higher relative to the parents in the comparison group, the effect size was in the small to medium range. This suggests that most parents of adults with SMI are able to maintain harmonious relationships with their other adult children, especially when parents are married and do not have functional limitations, and whose non-disabled adult children have achieved a level of independence as indicated by being currently working and married. It is possible that earlier in the life course, relationships between parents and their non-disabled adult children were characterized by higher levels of ambivalence, but by observing the relationship only later in the family life course we may be capturing families after they have made successful efforts to negotiate differences and conflicts between aging parents and their non-disabled adult children. Future research on the effects of a child’s disability on the parental relationship with their other children would benefit from studies investigating the trajectory of ambivalence over several points in the life course.

Consistent with our hypothesis, parental ambivalence mediated the negative associations of parenting a child with SMI on all six outcomes of physical and psychological well-being, but this process was not evident in the sample who had a child with DD. As we noted, the relatively stronger effects for the parents of children with SMI may follow from several unique features of their circumstances, including the uncertainty introduced into family life due to the cyclical nature of the symptoms of SMI, the fluctuating needs of care and supervision for the affected adult children, higher levels of stigma surrounding SMI, and differences among family members in their attributions of control individuals with SMI have over their symptoms (Greco & Roger, 2003; Muralidharan et al., 2016; Seltzer et al., 2001; Parker et al., 2015). These explanations could not be tested because the WLS does not include measures that capture these processes. Future research is warranted to understand how unique characteristics associated with different types of disabilities contribute to intergenerational ambivalence and parental well-being.

Finally, it should be noted that in the entire sample, parental ambivalence had a significant direct effect on all six outcomes of psychological and physical health. Previous studies have established the link between parental ambivalence and mental health outcomes, especially lower levels of psychological well-being and more depressive symptoms (e.g., Fingerman et al., 2008; Suitor et al., 2011). However, this study is among the first to demonstrate the association between ambivalence in the parent-adult child relationship and parental physical health outcomes. Ambivalence was one of the strongest predictors across all three physical health outcomes, showing comparable standardized coefficients as parental education. Thus, our findings provide further evidence for the critical importance of including measures of parental ambivalence in studying the health and well-being of aging parents. Additionally, the study points to the importance of child characteristics in predicting levels of parental ambivalence and well-being, as evident by the fact that the fit of the path models improved significantly when child characteristics were adjusted. The findings from the follow-up exploratory analysis examining each child characteristic individually further suggested that it is the set of interrelated child characteristics that affects the parents’ level of ambivalence toward an adult child. Thus, in developing programs and services to promote the well-being of aging parents, it is important for such parents to support their non-disabled adult children in their pursuit of employment, education, and marriage as they were all predictors of lower levels of ambivalence among aging parents, which in turn was associated with more positive physical and mental health outcomes. Ambivalent relationships also potentially have a negative effect on the well-being of the non-disabled adult children, and thus reducing ambivalence would benefit them as well as their aging parents.

The present study had several limitations that must be acknowledged. First, the sample was drawn from Wisconsin, and there were very few members of racial/ethnic minority groups, reflective of Wisconsin’s population in the mid-20th century when the sample was selected. It should also be noted that fathers were less represented than mothers in families of adults with SMI, although mothers and fathers of adults with SMI did not differently perceive their relationships toward non-disabled children. Second, we had unequal sample sizes for different family groups: the number of families with an adult with DD was considerably smaller than that of families of adults with SMI, and thus, it is possible that we were underpowered to find small effects of having an adult with DD on intergenerational ambivalence and parental health. Third, the measure of ambivalence was indirect, calculated based on parental report of their positive and negative feelings toward their children; thus, it is unknown whether our findings would generalize to direct measures of ambivalence. Forth, the low reliability of the conflict measure may attenuate the coefficients, although such reliability is not uncommon in the literature, particularly for scales with few items (e.g., Barber & Buehler, 1996; Ryff & Keyes, 1995). Finally, because the 2011 wave is the only wave of the WLS with intergenerational multilevel data, our findings rely on cross-sectional mediation analyses. Thus, this study’s findings should be viewed as preliminary. Future research using longitudinal data is needed to establish the temporal order of the relationships among having a child with a disability, intergenerational ambivalence, and parental well-being.

Despite these limitations, there were several manifest strengths of the present study, including the use of a random community-based sample, which is uncommon in the studies of later life families of an adult child with DD or SMI. The multilevel structure of the data allowed us to examine parental relationships toward each non-disabled child rather than using a global item reflecting the parent’s overall rating of the quality of their relationship with their children as a unit, which has been a limitation of prior family research. This multilevel structure also allowed us to take into account child characteristics, which are known to be highly predictive of parental ambivalence and well-being (Birditt et al., 2010; Luescher & Pillemer, 1998). Together these strengths and limitations point the way toward future research, including the study of culturally diverse families of an adult child with DD or SMI and the use of longitudinal analysis to assess temporal ordering between ambivalence and parental well-being.

In conclusion, the findings highlight an issue that has received virtually no attention in research on families of adults with DD or SMI – the quality of relationships between parents and their non-disabled adult children and the effects of these relationships on parental health and well-being. The study findings provide further evidence that aging parents of adults with SMI are a particularly vulnerable group. They experienced higher levels of ambivalence in their relationships with their non-disabled adult children, even after controlling for many factors found in past research associated with ambivalence. Further, high levels of parental ambivalence were associated with lower levels of psychological and physical health of the parents. Clinicians working with aging parents of an adult child with SMI should routinely assess parental relations with their non-disabled other children, particularly as it relates to ambivalence in the relationship, which plays a pivotal role in the well-being of the aging parents. Understanding and giving attention to family dynamics may also benefit the individuals with disabilities themselves, as well as their non-disabled siblings, by making their family support systems stronger.