Abstract
Although maternal socioeconomic status and health predict in part children’s future health and socioeconomic prospects, it is possible that the intergenerational association flows in the other direction such that child health affects maternal outcomes. Previous research demonstrates that poor child health increases the risk of adverse maternal physical and mental health outcomes. The authors hypothesize that poor child health may also increase the risk of poor maternal health outcomes through an interaction between child health and factors associated with health outcomes, such as marital status, marital quality, and socioeconomic status. Using data on women in the National Longitudinal Study of Youth 1979 cohort (N = 2,279), the authors find evidence that the effects of certain maternal marital quality and socioeconomic factors on maternal physical and mental health depend on child health status and vice versa.
Keywords
The prevalence of childhood chronic conditions and activity limitations has substantially increased over the past several decades, due to treatment advances, improved survival, and the emergence of new childhood conditions (Hogan, Msall, Rogers, & Avery, 1997; Perrin, Bloom, & Gortmaker, 2007; Van Cleave, Gortmaker, & Perrin, 2010). With the rise in chronic conditions among children, there is a critical need to understand the impact of child health on families.
Systems theory predicts that having a child in poor health in the family is a stressor that is associated with the well-being of members of the family system across multiple life domains, such as parenting, health, work, and marriage (Bronfenbrenner, 1977; Diez Roux, 2007). In addition to the balance of family needs experienced by all parents, parents of children with health conditions or activity limitations experience greater care burden compared with parents of children with no limitations: They are responsible for more of the physical care of their children; helping children cope physically and emotionally with their condition; negotiating medical, educational, recreational, and other services; and experiencing concerns about their child’s well-being and future prospects (Raina et al., 2004).
Many studies focus on the impact of child health on maternal physical and mental health outcomes, because mothers are most often the primary caregivers of children and as a result are more likely to be affected by child health or functional status and the member of the family on whom parent–child data are collected. Clinical and population-based studies have demonstrated a risk for adverse maternal physical and mental health outcomes for mothers of children with developmental disabilities, chronic health conditions, and activity limitations, as well as a higher risk for adverse physical and mental health outcomes compared with mothers of children with no disabilities, health conditions, or activity limitations (Abbeduto et al., 2004; Early, Gregoire, & McDonald, 2002; Glidden & Schoolcraft, 2003; Hastings, Daley, Burns, & Beck, 2006; Kuhlthau, Kahn, Hill, Gnanasekaran, & Ettner, 2010; Miller, Cohen, & Ritchey, 2002; Raina et al., 2005; Seltzer, Greenberg, Floyd, Pettee, & Hong, 2001; Singer, 2006; Witt et al., 2010; Witt, Gottlieb, Hampton, & Litzelman, 2009). The effect of child health on maternal physical and mental health outcomes can be explained in terms of a stress process perspective, where the stressors associated with having a child with a health condition or activity limitation are conceptualized as chronic strains which produce physiological and psychological distress (Pearlin, 1989; Pearlin, Schieman, Fazio, & Meersman, 2005). When an individual perceives a chronic environmental challenge as stressful, these challenges lead to adverse physiological and psychological outcomes through chronic exposure to fluctuating or heightened neural, neuroendocrine, and immune system responses (Adler et al., 1994; McEwen & Seeman, 1999; Taylor & Seeman, 1999).
Previous research has shown that child activity limitations are negatively associated with maternal labor force participation (Lukemeyer, Meyers, & Smeeding, 2000; Porterfield, 2002). Although the causal direction is not always clear, previous research has shown that child health and socioeconomic status are clustered together such that children with health conditions or activity limitations are more likely to live in economically disadvantaged families than are children with no health conditions or activity limitations (Kuhlthau & Perrin, 2001; Porterfield, 2002; Witt et al., 2009; Witt, Riley, & Coiro, 2003). Children with activity limitations are also more likely to have divorced, separated, or never married parents and less likely to have married parents, and poor marital quality is higher among parents of children with activity limitations (Risdal & Singer, 2004; Witt et al., 2003).
The idea that child health and functioning is associated with the well-being of other family members across multiple life domains corresponds to the concept of stress proliferation. Stress proliferation is the process by which an initial stressor leads to (a) the expansion of the initial stressor, (b) additional stressors in other life domains, (c) additional stressors across the life course, and (d) additional stressors across generations (Grzywacz, Almeida, & McDonald, 2002; Pearlin, Aneshensel, & LeBlanc, 1997, Pearlin et al., 2005; Thoits, 2010). Stress proliferation is a useful conceptual tool for noting that stressors tend to cluster and that researchers interested in particular stressors should account in some way for other stressors.
Research in both the stress process and cumulative advantage literatures supports the notion that there may be a multiplicative effect of multiple stressors on individual health and well-being. Research on the social patterning of stressors finds some evidence that the effects of chronic stressors on various measures of health are exacerbated in the face of episodic stressors (Lepore, Miles, & Levy, 1997; Marin, Chen, Munch, & Miller, 2009; Marin, Martin, Blackwell, Stetler, & Miller, 2007; Pike et al., 1997; Serido, Almeida, & Wethington, 2004). Researchers posit that the presence of one stressor reduces the resources available to the individual for coping with other stressors, such as resolving issues before they become larger problems and reducing negative and promoting positive appraisals of issues (Lazarus, 1999; Serido et al., 2004). Furthermore, research on mechanisms for altering stress effects demonstrates that certain positive factors—such as sense of control over life, high self-esteem, and social support—may weaken the effect of stressors on health (Lazarus & Folkman, 1984; Taylor & Stanton, 2007; Thoits, 2010). Thus, it is plausible that the presence of one stressor may exacerbate the effect of another stressor on health. Finally, stratification research on cumulative advantage in the relationship between socioeconomic status and health demonstrates interactions between socioeconomic status and factors such as race, ethnicity, and health behaviors in predicting health outcomes (Pampel & Rogers, 2004; Shuey & Willson, 2008).
Ample evidence exists demonstrating that marital status (Carr & Springer, 2010; Waite & Gallagher, 2000), marital quality (Gallo et al., 2003; Umberson, Williams, Powers, Liu, & Needham, 2006; Wickrama et al., 2001; Wickrama, Lorenz, Conger, & Elder, 1997), and socioeconomic status (Adler et al., 1994; Kitagawa & Hauser, 1973; Link & Phelan, 1995) are independent predictors of physical and mental health. However, it is unclear how maternal marital and socioeconomic factors are associated with child health in terms of deleterious maternal physical and mental health effects. We hypothesize that there is a multiplicative effect of multiple stressors on maternal physical and mental health outcomes, such that the effects of maternal marital and socioeconomic factors on maternal physical and mental health depend on child health status and vice versa.
Method
Data
Data come from the National Longitudinal Study of Youth 1979 (NLSY), a nationally representative sample of 12,686 men and women who were between the ages of 14 and 22 years when first surveyed in 1979 (http://www.bls.gov/nls/nlsy79.htm). Data were collected yearly from 1979 to 1994, and biennially from 1996 to the present, initially to chronicle the labor force experience of this cohort. However, other sets of questions have been added over the years and broadened the scope of the survey. Starting in 1986, data on the children of NLSY 1979 female respondents have been collected from both the mothers and children (http://www.bls.gov/nls/nlsy79ch.htm).
When the women from the original cohort turned 40, they were asked a range of health questions in the subsequent wave of the NLSY (between 1998 and 2006). To maximize use of the NLSY sample, data were pooled across waves so that they included all the women who participated in this health module in one of the five waves and were mothers of a child aged 18 years and younger for whom data on child activity limitations existed (N = 2,279, see the “Independent Variables” section). Among households where only one child had an activity limitation, we selected that child to represent the household. Using a random number generator, we randomly selected one child with a limitation in households where more than one child had a limitation, and we randomly selected one child to represent households where no children had activity limitations.
Measures
Dependent variables
Maternal physical health was measured as the physical component summary of the 12-item Short Form (SF-12), a brief inventory of 12 physical and mental health questions that captures functional health and well-being from the respondent’s point of view (Ware, Kosinski, & Keller, 1996). The inventory includes questions on self-reported health, limitations and pain hindering various activities, and how often the respondent experiences various emotional states. The physical component summary score was constructed by the NLSY from the 12 items using the physical health regression weights available in Ware, Kosinski, and Keller (1995). Scores were calibrated so that respondents with a score greater than 50 have better health than the typical person in the general U.S. population (age is not held constant), and each one-point difference above or below 50 corresponds to one 10th of a standard deviation; the NLSY sample has a slightly higher mean score and slightly lower standard deviation compared with the general U.S. population. The scores range from 0 to 100 in the population and from 13.66 to 66.63 in the analytic sample. Because the raw score was negatively skewed, we reversed the distribution to make it positively skewed by subtracting each SF-12 physical component score from one plus the maximum SF-12 physical component score in the sample, and used the natural log of this reversed score to correct for nonnormality. Thus, higher scores indicate worse maternal physical health in this analysis. (See the “Additional Analyses” section for results using the mental component summary score from the SF-12 as a dependent variable.)
The maternal depressive symptoms index was constructed as the sum of seven items from the Center for Epidemiologic Studies Depression scale (Radloff, 1977). This index was constructed by summing the seven items about how often the respondent felt each depressive symptom, with answers ranging from not at all (1) to most days (4). Scores ranged from 7 to 28 with a higher score indicating more depressive symptoms. The natural log of the index was used for analysis because the raw score was positively skewed.
Independent variables
Researchers have begun to move away from the “categorical” or “disease-specific” approach to defining child health based on evidence of commonalities across childhood illnesses in terms of child functional status and the experiences of the children’s families (Kohen et al., 2007; Stein, Bauman, Westbrook, Coupey, & Ireys, 1993). Defining child health using survey questions about activity limitations captures how children are physically, emotionally, or cognitively limited in a way that affects their age-appropriate activities, and improves on the disease-specific approach because a diagnosis from a medical professional is not necessary, so that measures of child health are not limited to those who have medical care.
Child health was defined in terms of child activity limitation status in this study. Mothers of children aged 15 years and younger reported whether the child currently had a condition that affected regular school attendance, regular school work, and usual childhood physical activities, and whether the child had a condition that required frequent attention or treatment from a health professional, the use of medicine or drugs, or the use of any special equipment. 1 Children aged 15 years and older reported on whether they currently had a condition that limited their ability to attend school regularly or do regular schoolwork; whether they had a condition that required frequent attention or treatment from a health professional, the use of medicine or drugs, or the use of any special equipment; and whether they were or would be limited in the kind or amount of work they could do for pay because of their health. 2 If any one of these items had a “yes” response, the child was coded as having an activity limitation. If all the items had a “no” response, the child was coded as having no activity limitation. The assumption was made that there was no difference in maternal and child reports of activity limitations for the purposes of this study, as we wanted to describe the relationships for all children and adolescents with data available on their activity limitation status.
To assess how changes in child activity limitation status modified the effect of marital and socioeconomic factors on maternal physical and mental health and vice versa, child activity limitation persistence dummy variables were constructed using data from two waves of the study (Witt et al., 2009). The child activity limitation persistence variables were operationalized as no limitation (no limitation reported in either wave), newly reported limitation (no limitation reported in the first wave and a limitation reported in the subsequent wave), resolved limitation (limitation reported in the first wave and no limitation reported in the subsequent wave), and ongoing limitation (limitation reported in both waves).
Among all mothers in the sample (N = 2,279), we were interested in examining whether child activity limitations moderated the effect of maternal marital status (being divorced, separated, widowed, or never married compared with being married) on maternal physical and mental health. Among married mothers (N = 1,542), we were interested in whether child activity limitations altered the effect of maternal marital quality factors on maternal physical and mental health. Based on prior research (see, e.g., Amato & Rogers, 1999; Rogers & Amato, 1997; Tach & Halpern-Meekin, 2009), a series of marital quality items were compiled into three indices: marital unhappiness, lack of quality time, and marital arguing. Marital unhappiness was constructed from one item that asked, “Would you say that your marriage is very happy, fairly happy, or not too happy?” Lack of quality time index was constructed as the sum of three items about the frequency with which the mothers did the following with their spouse: have calm discussions, laugh together, and talk about their day, with answers ranging from once a week (0) to less than once a month (3). A higher score (ranging from 0 to 9) indicates that quality time happened infrequently; Cronbach’s alpha for the scale was .66. The marital arguing index was constructed as the sum of 10 items about the frequency with which mothers argued with their spouses about different topics, with answers ranging from never (0) to often (3). A higher score on the arguing index (ranging from 0 to 30) indicates that the mother argued often with her spouse about several different topics; Cronbach’s alpha for the scale was .71. The continuous variables—lack of quality time and marital arguing—were mean-centered to remedy the multicollinearity between the interaction terms and their first-order effects.
Among all mothers in the sample (N = 2,279), we were also interested in examining whether child activity limitations moderated the effect of maternal socioeconomic factors on maternal physical and mental health. Family poverty status was defined as total family income being below the poverty level for the past year, as issued by the U.S. Department of Health and Human Services and based on Census Bureau poverty guidelines. Two measures of employment status were used in the current analysis: any time unemployed and any time out of the labor force. Any time unemployed was defined as not employed but available for work at least 1 week in the past year. Any time out of the labor force was defined as not employed and unavailable for work at least 1 week in the past year, where “unavailable” means that the person was not looking for work, for example, because they were engaged in housework, in school, unable to work because of long-term physical or mental illness, retired, or voluntarily idle. Because the statuses were defined as occurring for at least 1 week in the past year, a respondent could be both unemployed and out of the labor force in the same year.
Sociodemographic variables
Certain sociodemographic variables were introduced as control variables in the analysis if they had previously been shown to be potential confounders in the relationship between child health and maternal outcomes (see, e.g., Witt et al., 2009). The control variables include child age and gender, and maternal race, age, education, smoking, body mass index (BMI), marital status, poverty status, any time unemployed, any time out of the labor force, any health condition that limits working for pay, depressive symptoms, and number of children in the household.
Analytic Approach
The analyses were conducted using Stata 10.1 (StataCorp LP, College Station, Texas). The values for missing data were derived from the multiple imputation technique of multivariate imputation by chained equations, and the results reported were generated using Stata’s “micombine” command, which averaged coefficients across the five imputed data sets and adjusted their standard errors using Rubin’s (1987) formula that combines the estimated variability within and across replications with a small correction factor to the variance. The descriptive statistics (presented in a table later) were weighted using custom weights from the NLSY to adjust for the multiple survey years from which the data were drawn as well as the oversampling of certain populations in the NLSY 1979.
The empirical models examined focus on child activity limitations, maternal marital and socioeconomic factors, and the interactions between the two in their effects on maternal physical and mental health: log (MH) = β0 + β1AL + β2 MF + β3AL * MF + βκXκ + ε, where the natural log of each maternal outcome of interest (maternal physical and mental health; MH) is regressed on child activity limitation status (AL), a maternal marital or socioeconomic factor (MF), the interaction between child activity limitations and the maternal factor of interest, and a vector of sociodemographic control variables (X). When captured at the same point in time, the interaction between child activity limitations and a maternal marital or socioeconomic factor (β3) is statistically equivalent, such that child health moderates the effect of the maternal factor on maternal health and vice versa. Because child activity limitation persistence is an ordinal variable, partial F tests were conducted to ascertain whether the additional parameters used to estimate the interaction terms significantly improved the estimation of poor maternal physical health and depressive symptoms compared with the first-order effects models.
Since maternal health outcomes were measured in the wave after mothers turned 40 (between 1998 and 2006), data were pooled across the five waves so that they included all the women who participated in the over-40 health module and were mothers of a child aged 18 years and younger for whom data on child activity limitations existed (N = 2,279). To minimize reverse causality, child activity limitation status and maternal marital factors were measured in the wave before the maternal health outcomes were reported. (The socioeconomic factors were measured in the same wave as the maternal health outcomes because they refer to income and employment patterns in the previous year.) Child activity limitation persistence was measured using data from one and two waves before the maternal health outcomes were reported. Sociodemographic control variables were measured two waves prior to the maternal outcomes of interest.
Models including maternal socioeconomic factors (poverty, unemployed, out of the labor force) as independent variables of interest do not control for the baseline measure of that socioeconomic factor of interest (when the baseline measure is included, the results are substantively and statistically the same as those reported). Models including marital quality factors (marital unhappiness, lack of quality time, marital arguing) as independent variables of interest are restricted to mothers who were married at the time marital quality was measured (N = 1,542).
Results
Table 1 presents the descriptive statistics for the analytic sample, weighted to adjust for the multiple survey years from which the data were drawn as well as the oversampling of certain populations in the NLSY 1979. Overall, 19.5% of mothers had a child aged 18 years or younger with an activity limitation reported one wave before maternal health status was reported. In terms of child activity limitation persistence, about 72% of mothers had a child with no limitation in either wave, 8.5% of mothers had a child with a resolved limitation, 10.4% of mothers had a child with a newly reported limitation, and 9.1% of mothers had a child with persistent activity limitations.
Descriptive Statistics for Analytic Sample, NLSY 1979 a .
Note: NLSY = National Longitudinal Study of Youth 1979; SD = standard deviation.
Analytic sample averaged across the five imputed data sets.
N = 1,542 for marital unhappiness, lack of quality time, and marital arguing.
Imputation restricted to married mothers for marital quality characteristics.
SF-12 is the physical component score of the 12-item Short Form (Ware, Kosinski, & Keller, 1996).
CES-D is the Centers for Epidemiologic Studies–Depression scale (Radloff, 1977).
Main Effects on Maternal Physical and Mental Health
Our study accords with results from previous studies which found that having a child with activity limitations led to worse physical and mental health outcomes for mothers (see Table 2). Child activity limitations are associated with worse physical health and increased depressive symptoms compared with mothers of children with no activity limitations (Model 1). Having a child with persistent activity limitations across two waves of data is associated with worse maternal physical and mental health compared with mothers of children with no limitations in either wave, and having a child with newly reported activity limitations is significantly associated with worse physical health compared with having a child with no activity limitations in either wave (Model 2). Additional analyses (not shown) reveal that having a child with persistent activity limitations is also significantly associated with worse physical health and increased depressive symptoms compared with having a child with resolved and newly reported activity limitations.
Effects of Child Activity Limitations on Maternal Physical and Mental Health.
Note: SE = standard error. Coefficients averaged across the five multiply imputed data sets; standard errors adjusted according to Rubin (1987). All models control for child age; maternal age, race, ethnicity; and various maternal baseline characteristics: education, body mass index, smoking status, health conditions limiting work, depressive symptoms, marital status, poverty status, unemployment status, time out of the labor force, and children in the household.
SF-12 is the physical component score of the 12-item Short Form (Ware, Kosinski, & Keller, 1996).
CES-D is the Centers for Epidemiologic Studies–Depression scale (Radloff, 1977).
p < .05. **p < .01. ***p < .001.
Interaction Effects
Tables 3 and 4 report the regressions of maternal poor physical health and depressive symptoms, respectively, on child activity limitations and each maternal marital and socioeconomic factor of interest. Model 1 regresses the maternal outcome on child activity limitation status, the maternal marital or socioeconomic factor of interest, and sociodemographic control variables. Model 2 includes the interaction between child activity limitation status and the maternal factor of interest. Models 3 and 4 replicate Models 1 and 2 with the measure of child activity limitation persistence, comparing persistent activity limitations with no limitations. (New and resolved child activity limitations do not have significant interactions with maternal marital and socioeconomic factors in their effects on maternal physical and mental health compared with mothers of children with no limitations; results available from the authors.)
Main and Interaction Effects of Child Activity Limitations and Maternal Marital and Socioeconomic Factors on Maternal Physical Health (SF-12).
Note: SE = standard error. Coefficients averaged across the five multiply imputed data sets; standard errors adjusted according to Rubin (1987). Model 1 regresses maternal poor physical health on child activity limitation status, the maternal marital or socioeconomic factor of interest, and controls. Model 2 includes the interaction between child activity limitation status and the maternal factor of interest. Models 3 and 4 replicate Models 1 and 2 with the measure of child activity limitation persistence (comparing persistent child limitations with no child activity limitations). With the exceptions noted below, models control for child age; maternal age, race, ethnicity; and various maternal baseline characteristics: education, body mass index, smoking status, health conditions limiting work, depressive symptoms, marital status, poverty status, unemployment status, time out of the labor force, and children in the household. Models including maternal socioeconomic factors (poverty, unemployed, out of the labor force) as independent variables do not control for the baseline measure of that socioeconomic factor of interest. Models including marital quality factors (marital unhappiness, lack of quality time, marital arguing) as independent variables are restricted to mothers who were married at the time marital quality was measured. SF-12 is the physical component score of the 12-item Short Form (Ware, Kosinski, & Keller, 1996).
p < .05. **p < .01. ***p < .001.
Main and Interaction Effects of Child Activity Limitations and Maternal Marital and Socioeconomic Factors on Maternal Depressive Symptoms.
Note: SE = standard error. Coefficients averaged across the five multiply imputed data sets; standard errors adjusted according to Rubin (1987). Model 1 regresses maternal depressive symptoms on child activity limitation status, the maternal marital or socioeconomic factor of interest, and controls. Model 2 includes the interaction between child activity limitation status and the maternal factor of interest. Models 3 and 4 replicate Models 1 and 2 with the measure of child activity limitation persistence (comparing persistent child limitations with no child activity limitations). With the exceptions noted below, models control for child age; maternal age, race, ethnicity; and various maternal baseline characteristics: education, body mass index, smoking status, health conditions limiting work, depressive symptoms, marital status, poverty status, unemployment status, time out of the labor force, and children in the household. Models including maternal socioeconomic factors (poverty, unemployed, out of the labor force) as independent variables do not control for the baseline measure of that socioeconomic factor of interest. Models including marital quality factors (marital unhappiness, lack of quality time, marital arguing) as independent variables are restricted to mothers who were married at the time marital quality was measured. CES-D is the Centers for Epidemiologic Studies–Depression scale (Radloff, 1977).
p < .05. **p < .01. ***p < .001.
There is a significant marital unhappiness by child activity limitation persistence interaction effect for poor maternal physical health (Table 3, Model 4). Because poor physical health is log transformed, the coefficients can be interpreted in terms of percentage change, such that poor physical health changes by 100 * [exp(β) − 1] percent for a one-unit increase in the independent variable, holding all other variables in the model constant (Wooldridge, 2009). Compared with mothers of children with no limitations and who report they are “very happy” in their marriage, there is a 13% increase in poor physical health for mothers of children with persistent activity limitations who are “very happy” in their marriage 100 * [exp(0.12) − 1], and no change in poor physical health for mothers of children with no limitations who report that they are “not too happy” in their marriage (the coefficient for “not too happy” is not statistically significant). However, there is a 46% increase in poor physical health for mothers who both have a child with a persistent activity limitation and are “not too happy” in their marriages compared with mothers who both have a child with no limitations and who are “very happy” in their marriages 100 * [exp(0.12 − 0.05 + 0.31) − 1]. The significant interaction term indicates that the effect of marital unhappiness on poor physical health is greater for mothers of children with persistent activity limitations; conversely, the interaction may also be interpreted as the effect of persistent child activity limitations on poor physical health being greater for mothers who are “not too happy” in their marriages.
The effect of marital arguing on poor maternal physical health is exacerbated by child activity limitations and vice versa (see Table 3). There is no change in maternal physical health for a one-unit increase in marital arguing when the child has no limitations (100 * [exp(0) − 1]), but each one-unit increase on the marital arguing scale is associated with a 1% increase in poor physical health among mothers of children with activity limitations (in Model 2; 100 * [exp(0.00 + 0.01) − 1]) and a 2% increase in poor physical health among mothers of children with persistent activity limitations (in Model 4; 100 * [exp(0.00 + 0.02) − 1]).
There is a significant child activity limitation persistence by poverty status interaction effect for poor maternal physical health (Table 3, Model 4). Compared with mothers of children with no limitations and not in poverty, there is a 12% increase in poor physical health for mothers of children with persistent activity limitations and not in poverty (100 * [exp(0.11) − 1]), a 7% increase in poor physical health for mothers of children with no limitations and in poverty (100 * [exp(0.07) − 1]), and a 46% increase in poor physical health for mothers of children with persistent activity limitations and in poverty (100 * [exp(0.11 + 0.07 + 0.20) − 1]). Thus, the effect of poverty on poor physical health is greater for mothers of children with persistent activity limitations; conversely, the effect of persistent child activity limitations on poor physical health is greater for mothers in households below the poverty line.
There is also a significant child activity limitation persistence by labor force status interaction effect for poor maternal physical health (Table 3, Model 4), such that the effect of being out of the labor force on poor physical health is exacerbated by having a child with persistent activity limitations and vice versa. Compared with mothers of children with no limitations and who spent no time out of the labor force, there is a 12% increase in poor physical health for mothers of children with persistent activity limitations who spent no time out of the labor force (100 * [exp(0.11) − 1]), a 13% increase in poor physical health for mothers of children with no limitations who spent any time out of the labor force (100 * [exp(0.12) − 1]), and 51% increase in poor physical health for mothers of children with persistent activity limitations and who spent any time out of the labor force (100 * [exp(0.11 + 0.12 + 0.18) − 1]).
The effect of marital unhappiness on maternal depressive symptoms is exacerbated by child activity limitations and vice versa. Compared with mothers of children with no limitations and who report they are “very happy” in their marriage, there is a 43% increase in depressive symptoms for mothers who both have a child with an activity limitation and are “not too happy” in their marriages (100 * [exp(0.02 + 0.11 + 0.23) − 1]; Table 4, Model 2). Compared with mothers of children with no limitations and who are “very happy” in their marriages, there is a 27% increase in depressive symptoms for mothers who both have a child with persistent activity limitations and who are “fairly happy” in their marriages (100 * [exp(0.01 + 0.08 + 0.15) − 1]) and a 60% increase in depressive symptoms for mothers who both have a child with persistent activity limitations and who are “not too happy” in their marriages (100 * [exp(0.01 + 0.11 + 0.35) − 1]; Table 4, Model 4).
There are significant lack of quality time by child activity limitation status (Table 4, Model 2) and persistence (Table 4, Model 4) interaction effects for maternal depressive symptoms. These interaction effects reveal a pattern where each one-unit increase on the lack of quality time scale is not associated with changes in depressive symptoms among mothers of children with no limitations (the coefficient is not statistically significant), but is associated with a 5% increase in depressive symptoms among mothers of children with activity limitations (100 * [exp(0.01 + 0.04) − 1]; Table 4, Model 2), and a 6% increase in depressive symptoms among mothers of children with persistent activity limitations (100 * [exp(0.01 + 0.05) − 1]; Table 4, Model 4). Finally, there is a significant marital arguing by child activity limitation persistence interaction effect for maternal depressive symptoms (Table 4, Model 4), where each one-unit increase on the marital arguing scale is associated with a 2% increase in depressive symptoms among mothers of children with persistent activity limitations (100 * [exp(0.01 + 0.01) − 1]).
Partial F tests (not shown) reveal that a more parsimonious coding of child activity limitation persistence improves the fit of the models predicting maternal poor physical health and depressive symptoms. (For simplicity, the results of the less parsimonious models are reported, and the results are substantively the same as the more parsimonious models.) In the models predicting poor maternal physical health, combining no limitations with resolved limitations resulted in significantly (p < .05) improved model fit when poverty status is the exposure of interest, and combining no, resolved, and new limitations results in significantly (p < .05) improved model fit when marital unhappiness, marital arguing, and time out of the labor force are the exposures of interest (“very happy” and “fairly happy” were also combined in the marital unhappiness model). In the models predicting maternal depressive symptoms, combining no limitations with resolved limitations results in significantly (p < .05) improved model fit when lack of quality time is the exposure of interest, and combining no, resolved, and new limitations results in significantly (p < .05) improved model fit when marital unhappiness and marital arguing are the exposures of interest (“very happy” and “fairly happy” were also combined in the marital unhappiness model).
Additional Analyses
The analyses discussed above were repeated with other measures of maternal physical and mental health and showed no significant interaction effects: SF-12 mental component score, whether the mother had any limitations that would or could inhibit her ability to work; maternal self-reported health in a multinomial logistic regression; and maternal self-reported health dichotomized as fair or poor health versus excellent, very good, or good health in a logistic regression. An ordinal logistic regression of maternal self-reported health showed significant interactions of persistent child activity limitations with marital arguing and maternal time out of the labor force. However, this model violated the proportional odds assumption underlying an ordinal logistic regression model.
It is also plausible to examine whether maternal marital and socioeconomic factors mediate the effect of child activity limitations on maternal physical and mental health. We used two tests of mediation available in Stata (“sgmediation” and “binary_mediation,” both with bootstrapped standard errors and confidence intervals) to examine whether the effects of child activity limitations on maternal physical and mental health were mediated by the marital and socioeconomic factors of interest, controlling for additional covariates. The effect of child activity limitations on poor maternal physical health was significantly mediated by maternal poverty status, and the effect of poverty status on poor maternal physical health was significantly mediated by child activity limitations. The effect of child activity limitations on maternal depressive symptoms was significantly mediated by lack of quality time and poverty status, and the effects of lack of quality time and poverty status on maternal depressive symptoms were significantly mediated by child activity limitations. However, the indirect effects in these models were substantively small and at most made up 9% of the total effect of child activity limitations on maternal physical or mental health.
Limitations
The relationships between child activity limitation status and maternal physical and mental health may be overestimated to the extent that the maternal reports of child health were influenced by maternal physical or mental health or unobserved underlying variables. This study mitigated these potential biases by using longitudinal data to control for measures of baseline maternal physical and mental health status as well as baseline measures of other covariates that could lead to a spurious relationship. However, to the extent that these controls were not exhaustive or were imperfect measures of the constructs they proxy, the results may be biased.
Discussion
This study confirms that having a child with activity limitations in the family is a stressor that directly influences the physical and mental health of the mother, and expands on prior research to account for the possible moderating role that child health may play in determining maternal physical and mental health outcomes. The effects of marital unhappiness, marital arguing, household poverty status, and maternal time out of the labor force on poor maternal physical health are greater when the mother has a child with persistent activity limitations compared with having a child with no limitations (Model 4); the effect of marital arguing on poor maternal physical health is greater when the mother has a child with an activity limitation compared with mothers of children with no limitation (Model 2). The effects of marital unhappiness, lack of quality time with one’s spouse, and marital arguing on maternal depressive symptoms are greater when the mother has a child with persistent activity limitations compared with having a child with no limitations (Model 4); the effects of marital unhappiness and lack of quality time on maternal depressive symptoms are also greater when the mother has a child with an activity limitation compared with mothers of children with no limitations (Model 2). Overall, having a child with activity limitations exacerbates the effects of some maternal marital quality and socioeconomic factors on maternal poor physical health and depressive symptoms and vice versa, particularly when child activity limitations are persistent.
Our findings speak to both role theory and stress process models and their implications for the well-being of members of the family. Stress may be produced when there are strains within particular roles such as parenthood, marriage, or work particularly because these roles are important for individuals and the society at large (Avison & Turner, 1988; Pearlin, 1989; Wheaton, 1999). Our findings point to the importance of strains within and across parental, marital, and socioeconomic roles for maternal well-being, in that multiple role strains appear to have a multiplicative adverse effect on maternal physical and mental health. Future research should investigate the implications of poor child health for other members of the family in terms of the multiplicative effect of multiple role strains, as well as whether coping mechanisms and social support (Lazarus & Folkman, 1984; Taylor and Stanton, 2007; Thoits, 2010) can buffer against the effects of these multiple role strains on the physical and mental health of family members. Given the importance of race and ethnicity in predicting differences in socioeconomic status and health outcomes in the cumulative advantage literature (see, e.g., DiPrete & Eirich, 2006; Kelley-Moore & Ferraro, 2004; Shuey & Willson, 2008), future research should also examine the role of maternal race and ethnicity in the multiplicative effect of multiple role strains. Finally, future research should examine how maternal and child reports of child activity limitations differ and the implications of these differences for the multiplicative effect of multiple role strains.
Any time spent out of the labor force—out of work and not looking for work—is a potentially heterogeneous category, in that some women choose to be out of the labor force during childbearing and child-rearing years and are able to do so without financial impediments. We found that any time out of the labor force was significantly associated with poor maternal physical and mental health outcomes in this sample, and that the negative effect of maternal time out of the labor force on maternal physical health was exacerbated by having a child with persistent activity limitations and vice versa. These findings complement research on respite effects, which demonstrates that working outside the home provides mothers of children with activity limitations, disabilities, or chronic health conditions a buffer against the stressful effects of caregiving (see, e.g., Morris, 2011; Thyen, Kuhlthau, & Perrin, 1999; Warfield, 2001, 2005). Future research should continue to examine how the heterogeneous components of maternal labor force participation may have both positive and negative consequences for maternal well-being in the presence of child activity limitations.
The findings of this study have particular policy implications. In general, a measure of activity limitations is likely more relevant for policy decisions where functionality, rather than diagnosis, is the target. This study supports prior studies which have found that child activity limitations, and especially ongoing child activity limitations, are associated with adverse maternal physical and mental health outcomes. Thus, policies that promote health and health care for family members are necessary to prevent the adverse physical and mental health outcomes associated with caring for a child with activity limitations, especially persistent activity limitations. We envision policies such as targeting family members for additional health screenings and health care once one member of the family is identified as having a health condition, disability, or activity limitation. Particular to this study, it appears that additional socioeconomic and marital or relationship support for mothers of children with activity limitations is also necessary to help buffer against the synergistic effects of child activity limitations with marital quality and socioeconomic factors on maternal physical and mental health. Such policy initiatives could target family members once one member of the family is identified as having a health condition, disability, or activity limitation, and provide marital and relationship counseling through health care services. Additionally, policies that promote flexible work schedules and adequate and affordable child care may reduce the impact of a child’s limitation on a mother’s ability to work (Parish & Cloud, 2006; Smith et al., 2002), thereby reducing the synergistic effects of child health and socioeconomic factors on maternal physical health outcomes.
Footnotes
Acknowledgements
We thank the editor and three anonymous reviewers for their helpful comments.
Authors’ Note
All opinions and errors are the responsibility of the authors and do not necessarily reflect those of the funding agencies or helpful commentators.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article:
This research was supported by funding from the Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD; T32 HD007014, A. Palloni, Principal Investigator and HD049533, W. P. Witt, Principal Investigator), and from core funding to the Center for Demography and Ecology (R24 HD047873) and Center for Demography of Health and Aging (P30 AG017266) at the University of Wisconsin–Madison.