Introduction

Research on the health of mothers has tended to focus only on the early postpartum period. This combined with the prominence of the biomedical perspective has, until more recently, precluded analysis of maternal health through a structural lens that illuminates the social, material, and economic pathways through which mothers of young children are made vulnerable to health risks (Onarheim et al. 2016), risks that extend well beyond the first year postpartum.

Stress, excessive workloads, and economic strain (Parkes et al. 2015; Sperlich and Geyer 2015; van Veldhoven and Beijer 2012) are among the potential precursors (or intensifiers) of poor health experienced by mothers of young children on almost a daily basis. Not all mothers, however, are equally vulnerable to such insalubrious effects of caregiving (Parkes et al. 2015; Sperlich and Geyer 2015). As it has been well recorded in Canada and elsewhere (Van de Velde et al. 2014; Witvliet et al. 2014), lone mothers have poorer health than their partnered peers (Colton et al. 2015; Gucciardi et al. 2004). While this relationship is often explained through the mechanism of poverty (Van de Velde et al. 2014; Gucciardi et al. 2004; Curtis and Phipps 2004), other factors, such as time stress (Colton et al. 2015) and welfare state generosity (Van de Velde et al. 2014; Curtis and Phipps 2004; Burstrom et al. 2010) may be equally relevant. Socio-economic position (SEP), measured by education and income levels, has been positively associated with health among mothers of young children in the USA (Link et al. 2017; Shippee et al. 2015) and among European countries (Brennenstuhl et al. 2015). While socio-economic inequalities in birth outcomes and early postpartum health indicators have been identified in Canada (Daoud et al. 2015), how these inequalities play out for women’s health in the years following giving birth appears to be largely unknown. Also unclear is the relationship between race and health of mothers of young children in Canada, despite evidence from the USA citing strong racial inequalities in health among these women, which are not entirely explained by SEP (Link et al. 2017; Shippee et al. 2015).

As early motherhood is a particularly vulnerable lifestage, most high-income countries have established some degree of welfare state protection for women, in the form of maternity leaves, parental leaves and/or subsidized forms of childcare (Gornick and Meyers 2003). There is some evidence that more generous social policy contexts are associated with better health among women overall (Borrell et al. 2014) and among lone mothers in particular (Van de Velde et al. 2014; Burstrom et al. 2010). Canada is considered among “Liberal” welfare states, which are known for promoting mostly market-based (private) care options and providing minimal caregiving leave supports (Gornick and Meyers 2003). However, since 1997, the province of Québec has diverged from the rest of Canada by implementing universal, government-subsidized daycare, with costs ranging from $7 to $20 per day (compared to almost $90/day in some regions) (Macdonald and Friendly 2016). Also, since 2006, Québec has offered a stand-alone, parental leave program that provides more comprehensive and generous benefits than other Canadian provinces (Labour Standards in Québec 2016). Due to the unique nature of the social policies relating to being a mother of a young child in Québec, its comparison with other regions of Canada may be particularly informative.

The magnitude and type of health inequalities that exist among mothers of young children in Canada is largely unknown, with research being piecemeal and tending to focus on the immediate postpartum period. The purpose of this study, therefore, is to identify dimensions of inequalities in health among mothers of young children in Canada based on SEP, race, partner status, and region and determine whether they are independent of one another. It was hypothesized that being single, identifying as a person of colour, and/or having a low education level will be risk factors for poor health among mothers of young children, while higher income and living in Québec, a proxy for welfare state generosity, will protect against poor health. As health inequalities are known to be reproduced across generations (Link et al. 2017; Shippee et al. 2015), a better understanding of how mothers of young children fare in Canada may help prevent this transmission and improve public health and, in particular, the well-being of women in the prime of their lives.

Methods

Sample

Data were derived from the public use version of the 2014 Canadian Community Health Survey (CCHS). The CCHS is a cross-sectional survey designed to address health status, health care utilization, and health determinants of the Canadian population aged 12 years and older and can provide reliable estimates at the health region level (Statistics Canada 2015). A multistage stratified cluster design was used to sample from the population, with the dwelling as the final sampling unit. Less than 3% of the population was excluded from the sampling frame. Data were collected using Computer-Assisted Interviewing either over the telephone (60%) or in person (40%). Of the selected units, 97,467 were in-scope for the survey, of which 73,190 households agreed to participate in the survey, resulting in an overall household-level response rate of 75.1%. One individual was selected from each responding household, and a response was obtained for 63,964 individuals, resulting in an overall person-level response rate of 87.4%. At the national level, these figures yield a combined response rate of 65.6%. For more details of the sample design, data collection, and data quality, see Statistics Canada documentation (Statistics Canada 2015).

For the current study, the sample was restricted to women living in a household with a child aged 5 or under (n = 3611). Those who did not identify as a parent living with a child(ren) were excluded (n = 677), as were those under the age of 20 or over the age of 50 (n = 17), resulting in a final sample of 2917. Further, those with missing data on one or more of the variables of interest were removed (n = 261; 8.9% of sample), resulting in a final sample size of n = 2656. All data were weighted using person-level weights as directed by Statistics Canada (Statistics Canada 2015). Sample size is reported in its unweighted form.

The primary outcome was poor health, defined as presence of two or more current chronic conditions. Multimorbidity is linked with increased health care use, including hospitalization, and when experienced at a younger age, may indicate increased vulnerability to poorer outcomes (Gruneir et al. 2016), making it a potentially useful indicator of poor health in younger samples. Chronic conditions included asthma, arthritis, back problems, high blood pressure, migraines, lung disease (e.g., COPD, chronic bronchitis), diabetes, heart disease, cancer, ulcers, stroke, bowel disorders (e.g., Crohn’s Disease, irritable bowel syndrome), depression, or anxiety. Respondents were asked to say yes to any “long-term conditions” they had which are “expected to last or have already lasted 6 months or more and that have been diagnosed by a health professional”.

Potential dimensions of inequalities included education level (coded as ≤high school or >high school), race (white or non-white), income level (in deciles, missing data were imputed by Statistics Canada), and partnership status (married/common law or single/divorced/widowed) and the proxy for welfare state generosity: region (Québec or rest of Canada). Control variables included age (in 5-year age categories), and employment status in the last 12 months (employed or not employed).

Descriptive statistics were used to summarize the demographic characteristics of the sample, and included frequency counts and percentages and means and standard deviations for categorical and continuous variables, respectively. Multiple logistic regression was used to estimate the odds ratio of poor health according to each type of inequality being investigated. Two models were tested for each type of inequality: in the first model, the odds of poor health were calculated adjusted for age and employment status only. Increasing age is associated with poorer health status, as well as changes in the magnitude of health inequalities (Sacker et al. 2005). Non-employment is more common among mothers of young children, and varies by region, partner status, and race (Moser 2017; Chui and Maheux 2011), and, therefore, may help account for the health inequalities being tested. In the second model, the odds were additionally adjusted for SEP, as it has been shown to help explain the association between each of race (Link et al. 2017; Shippee et al. 2015) and partner status (Van de Velde et al. 2014; Gucciardi et al. 2004; Curtis and Phipps 2004) and health among mothers of young children. In the final model, the odds were adjusted for age, employment status, SEP, and all other types of inequality to determine whether each form of inequality was independent of each another. In total, ten models were tested separately for each type of inequality, plus one fully adjusted model including all inequalities simultaneously. Statistical significance was established at p < 0.05. The analysis was undertaken using SPSS (version 24).

To validate the findings, a number of sensitivity analyses were undertaken. First, the robustness of the definition of poor health as the presence of two or more chronic conditions was tested by respecifying the outcome as (a) one or more chronic conditions and (b) total number of chronic conditions, and re-running the analysis to ensure a similar pattern of findings was observed. Second, to further investigate the proxy for welfare state generosity (comparing the region of Québec vs. the rest of Canada), region was broken down into smaller units, including each of the three big provinces of Ontario, Alberta and British Columbia, the Eastern provinces together and the Prairie provinces together (combined due to smaller population sizes). This was done to determine if the hypothesized protective effect of living in Québec was also found when comparing it to each region separately. It is possible that variation across regions would wash out an overall difference with Québec or, due to a strong difference with a large province, exaggerate the magnitude of the relationship. Last, to validate the selection of mothers, the analysis was rerun including only women who had given birth in the last 5 years and comparing the results with those of the primary analysis. The question on giving birth, however, was only asked in eight provinces/territories (including Québec), resulting in a less representative and smaller sample size (n = 1923, see Statistics Canada documentation for details).

Results

A total of 2656 mothers of young children were included in the sample. The largest proportion of mothers was age 30 to 34 years (31.0%) and just under a third identified as being part of a racial group other than “white” (28.5%). More than three quarters of mothers were living as married or common law (87.7%) and had more than a high school diploma (82.1%). The mean income decile was 5.29 (sd = 2.8). Less than a third of the sample were not working in the year prior to the survey (30.7%). A full description of the sample is found in Table 1. In total, 484 mothers were coded as having poor health due to the fact that they reported having two or more chronic conditions, resulting in a weighted prevalence of 17.8% (95% CI = 16.4–19.3).

Table 1 Characteristics of mothers of young children in Canada (n = 2656)
Full size table

Table 2 provides the results of the logistic regression models of the presence of two or more chronic conditions. The first row presents the estimates from the models adjusted for age and employment status only. As shown, every dimension of inequality is significantly related to the odds of poorer health among mothers of young children. Among the binary risk factors, the magnitude of increased odds is greatest for being single (OR = 2.82; 95% CI = 2.17–3.65). As for protective factors, for every increase in income decile, the odds of poorer health decreases by about 18% (OR = 0.85; 95% CI = 0.82–0.89); living in Québec is associated with two times lower odds of poorer health (OR = 0.50; 95% CI = 0.38–0.67).

Table 2 Adjusted odds ratios of presence of two or more chronic conditions among mothers of young children in Canada (n = 2656) for each dimension of inequality separately
Full size table

The second row of Table 2 presents the estimates from the models adjusting for age, employment status, and SEP. Including SEP in the model attenuates the odds of poor health for all dimensions of inequality excluding race, for which the odds of poor health increase to 1.77 (95% CI = 1.37–2.25). The largest attenuation is observed for single status, whereby the increased odds of poorer health decline by about 26% (OR = 2.07; 95% CI = 1.37–2.25). Notable attenuation is also noted for education level, which is reduced to a non-significant OR of 1.29 (95% CI = 0.99–1.68). Very little or no attenuation is noted for the protective factors of higher income and living in Québec.

The third row of Table 2 presents the fully adjusted estimates from the models of presence of two or more chronic conditions. Full adjustment includes age, employment status, SEP, and each other dimension of inequality. Four of the five dimensions of inequality remain significant when all other variables are considered: single status (OR = 1.80; 95% CI = 1.35–2.39); non-white race (OR = 1.72; 95% CI = 1.34–1.21), income (OR [each decile] = 0.86; 95% CI = 0.82–0.90), and living in Québec (OR = 0.50; 95% CI = 0.38–0.67).

Several sensitivity analyses were undertaken to validate the findings. First, the outcome measure was re-specified using the threshold of one or more chronic conditions. The results of this analysis are highly consistent with the primary analysis; however, the odds ratios are slightly smaller on average. The outcome was also re-specified as number of chronic conditions. Once again, the results are consistent: four out of the five dimensions of inequality are significant in the fully adjusted model. These two sensitivity analyses are provided in a supplementary file.

Second, smaller regional units were investigated to determine if the protective effect of living in Québec was maintained when compared to all other regions of Canada tested separately. As shown in Table 3, the odds of poor health among mothers of young children are significantly higher in all regions tested compared to Québec. The ORs from the fully adjusted model range from about two times as large in the Eastern provinces (OR = 2.09 95% CI = 1.33–3.29) and British Columbia (OR = 2.20; 95% CI = 1.47–3.29) to over 80% higher in Alberta (OR = 1.86; 95% CI = 1.28–2.69).

Table 3 Adjusted odds ratios of presence of two or more chronic conditions for region among mothers of young children in Canada (n = 2656)
Full size table

Finally, the analysis was rerun using only women who had given birth in the last 5 years with complete data on the all the variables of inquiry (n = 1923). As shown in Table 4, the results are highly consistent with those presented from the primary analysis: four out of the five dimensions of inequality are associated with poor health in all models tested.

Table 4 Adjusted odds ratios of presence of two or more chronic conditions among women who gave birth in the last 5 years in eight provinces (n = 1923) for each dimension of inequality separately
Full size table

Discussion

There is a lack of research on the structural factors that influence the health of women in Canada during the period they are mothers of young children. This current research attempts to provide some illumination on this topic by revealing that, on one hand, women who are mothers of young children are not immune to health inequalities and, indeed, seem to be clearly differentiated by race, SEP and partner status. On the other hand, mothers living in regions providing more generous social policy contexts (i.e., Québec) appear to be protected from poorer health, providing preliminary evidence that social policies, such as universal access to affordable daycare and generous parental leaves, may help to mitigate the health risks precipitated by maternal caregiving.

This is the first study that I am aware of to demonstrate differences in the health of women who are mothers of young children by regions of Canada. There is a solid evidence base suggesting that welfare state generosity is related to better health status (Brennenstuhl et al. 2012), but despite the socio-economic vulnerabilities associated with motherhood, only a few of such studies have focused on the health of mothers in particular (Van de Velde et al. 2014; Burstrom et al. 2010; Brennenstuhl et al. 2015). The finding that mothers in Québec have lower odds of poor health than those in the rest of Canada, while preliminary, is consistent with this small literature in showing health variation by welfare regime. Interestingly, all of the dimensions of inequality (i.e., SEP, race, and partnership status) remained significant when region was included in the model, suggesting that while region (the proxy for welfare state generosity) may relate to overall health status, it does not necessarily account for health inequalities. This finding is consistent with the extant literature, which shows that the magnitude of health inequalities does not always correlate with welfare state generosity (Brennenstuhl et al. 2012). This oft-named “paradox” may be explained by social selection, whereby inequalities persist in more equal societies because, while fewer people are worst off, those who are have become increasingly homogeneous with respect to characteristics that predict poor health (Mackenbach 2012). At the same time, non-material factors, such as personality traits, which intervention by the welfare state has not addressed, may have become more strongly predictive of health, further adding to selection effects (Mackenbach 2012). Thus, living in a certain region can be protective of health even if health inequalities persist. Given the novel nature of our findings of regional differences in health among mothers of young children, ongoing research is needed, which can replicate these results using other datasets and different health indicators.

The findings regarding dimensions of inequalities based on race and SEP are consistent with hypotheses and aligned with prior research in the USA (Link et al. 2017; Shippee et al. 2015), although they are novel in the Canadian context. The results about partnership status further contribute to the body of evidence demonstrating the health risks associated with being a single mother in Canada (Colton et al. 2015) and internationally (Van de Velde et al. 2014; Witvliet et al. 2014). What is interesting is the finding that covariation among the dimensions of inequality does not appear to account for the relationships. Thus, while pathways to inequality may be somewhat interrelated, their effects on health among mothers of young children appear to be independent from one another. This suggests that each pathway should be considered separately when developing public health interventions. For instance, the race-health relationship may be explained by the experience of discrimination, which is thought to contribute to the poor health of mothers of colour (Shippee et al. 2015). Discrimination may be experienced through a variety of systems, including healthcare, childcare, and education. With women’s increased reliance on these systems when their children are young, interventions that address systemic discrimination may help mitigate poor health. Along similar lines, time stress may help explain the relationship between partner status and health, which has been proposed as an equally relevant mechanism of poor health as income (Colton et al. 2015). Thus, provision of flexible childcare options for lone mothers (e.g., longer/weekend hours) may be an effective public health intervention. To better understand pathways of inequalities among mothers of young children and to inform public health interventions to address them, future research should include more proximal causes of health inequality, such as discrimination and time stress, along with their structural roots, including race, gender, and SEP.

This research is vulnerable to a number of limitations that should be considered when interpreting its results. First and foremost, this research was not designed to address questions of causation. Health status prior to motherhood was not controlled for, so it is entirely possible that those in poor health were already so prior to becoming a mother. Regardless of how women achieved their health status, this research remains informative because it sheds light on the health inequalities that exist during a particularly vulnerable time in the lives of women and their young children. Future research using longitudinal data following women from prior to giving birth to at least 5 years postpartum would be able to shed light on questions of causation and provide insights into health dynamics over time.

Another important limitation is that the CCHS is not designed to provide a representative sample of mothers. Consequently, some mothers, especially those with fewer resources, may have been less likely to respond to the survey, reinforcing the need to replicate the findings of this research. Further to sampling issues, mothers themselves comprise a selected population. Previous research has suggested that social policies help shape whether, when and how women go on to become mothers, with women living in some less-generous welfare states being less likely to become mothers at all (Brennenstuhl et al. 2015; Rovny 2011). It is possible, therefore, that the circumstances surrounding decisions (or lack thereof) to become a mother vary across Canada, and thus creating heterogeneity in the samples of mothers from each region. Future research could use methods such as propensity score matching to ensure the samples of mothers from Québec and elsewhere are truly comparative.

With the current study design, the effect of social policy arrangement on health outcomes cannot be isolated. A time series design that can look at trends in maternal health before and after implementation of universal daycare in Québec, for example, may be one method for attaining stronger evidence in that regard. Finally, a very broad measure of race including only two categories was used, which does a poor job of representing the multicultural makeup of Canada. The findings of this research, as well as a previous work suggesting that the relationship between maternal and child health might be stronger among certain races (Shippee et al. 2015), underline the need for a more nuanced population-based study of the relationship between race and health among mothers within the Canadian context.

To conclude, this research provides preliminary evidence that health inequalities exist among mothers of young children in Canada by income, race, and partnership status, and that living in Québec appears to protect against poor health. The identification of these structurally-based inequalities has important policy implications. For example, it suggests that public health programs that address the root causes of inequalities may be more effective in mitigating poor maternal health than the individual approach most commonly used in postpartum care (Storey-Kuyl et al. 2015). An example of such programs is universal access to childcare, which has been endorsed as a method for reducing health inequalities by the World Health Organization (World Health Organization 2008), but has yet to be implemented nationally in Canada. If we seriously want to help mitigate the reproduction of health inequalities across generations and improve the well-being of women today, investments need to be made now.