The effects of family type on child mortality

Abstract

Background: A growing number of children live in single-parent families, which seems to be associated with negative effects on a child’s health. Little is known about the health of children in cohabiting two-parent families that are also increasingly common, and more susceptible to family break-up than married two-parent families. This study seeks to determine if family type is associated with child mortality and whether any association remains after controlling for socio-economic factors. Methods: We used longitudinal nationally representative register data from Statistics Finland to study deaths between ages 1–14 years (1780 deaths, N = 201 211) during 1990–2004. The relative effects of family characteristics on mortality were estimated using Cox regression models. Results: Compared with children of married parents, children of single parents carried an excess risk in mortality in ages 1–4 years [Hazard Ratio (HR) 2.02, 95% CI 1.63–2.51] and in ages 5–9 years (HR 1.44, 95% CI 1.15–1.80). The relationship between single parenthood and mortality was largely, but not entirely, explained by associated low parental education and lower household income. Mortality among children in cohabiting-parent families showed no difference from children of married parents. Conclusion: Mainly due to accidental and violent causes of death, the largest excess mortality risks concentrated among children of single, less-educated and less-earning parents. The most vulnerable age period in this respect was early childhood (ages 1–4 years), whereas no association between mortality and family type was found among children aged 10–14 years.

Introduction

During the recent decades, unmarried parenthood has become increasingly common due to rising rates of cohabitation, union dissolution and non-marital childbearing. These ongoing changes in family life have raised concerns about the effects of family type on the health of children. Potentially harmful consequences of living in other than an intact two-parent family include (i) less parent-time allocated to childcare, (ii) fewer material and social resources and (iii) less stable family life, which may entail experiencing even several changes in family structure.^1,2

Previous research has documented higher mortality among children of single parents. During 1986–90 in Finland, 5- to 14-year-old boys living in single-parent families carried a slightly elevated risk of mortality that was due to both diseases and accidents.³ In Sweden, Östberg found no association between family type and disease mortality among children aged 0–18 years in 1980–86. In accidental deaths, however, gender and social class produced the largest differentials but the mortality risk was also higher for children of single parents and immigrant parents and for those living in the least densely populated areas. These differentials remained also after mutual adjustments.⁴ In contrast, Blakely et al.⁵ concluded in a 1991–94 follow-up of deaths under the age of 15 years in New Zealand that the association between one-parent families and child mortality was totally due to associated low socio-economic position.

Weitoft et al.⁶ showed in a population-based cohort follow-up (1991–99) that children of single parents in Sweden had increased risks of both morbidity and mortality even if controlling for a wide range of demographic and socio-economic factors. In a 5-year follow-up (1991–95) on injury deaths of all children under the age of 15 years in Sweden in 1985, Hjern and Bremberg⁷ found parental socio-economic position to be the most important parental determinant. Nonetheless, family situation (death of mother or single parenthood) still explained 11% of suicides and 18% of homicides.⁷ In the USA, Hussey⁸ instead concluded that household structure (unmarried female head) had no independent effect on risks of homicide, suicide or other injuries. The results of these three studies are not, however, directly applicable to the child population as a major proportion of the followed children had reached adolescence or even adulthood at the time of death. Many authors have indeed concentrated on the long-term effects of single parenthood, thus evidence on the association between single parenthood and mortality during childhood remains scarce.

To our knowledge, no studies have yet been conducted on the mortality of children living with cohabiting parents, although the excess mortality risk among cohabiting adults as opposed to married couples has been shown to apply to cohabiters with children as well.⁹ Out of all families with children under the age of 18 years in Finland, the proportion of cohabiting-parent families has grown from 9% in 1990 to 17% in 2004.¹⁰ Based on simultaneous rises in cohabitation, union dissolution and non-marital births, the number of cohabiting-parent families has grown across Europe and the USA, notwithstanding a wide diversity of rates between countries.^11,12 From a child’s point of view the marital status of parents may be irrelevant, but at least up to date children born to cohabiting parents have been typically twice as likely to experience a family break-up than children born to married parents.¹³ Not only more often socio-economically disadvantaged than married parents, cohabiting parents may also receive less support from their social networks than married parents do.^14,15

This study examines if family type (married, cohabiting or single-parent family) is associated with child mortality, and whether such an association can be explained by socio-economic factors. Possibly confounding factors such as number of children in the family, parental age and degree of urbanization of the living area will be adjusted for. As our focus lies on the effects of family type during childhood when children still live at their parental homes and are largely dependent on their parents, the mortality follow-up from 1990 to 2004 will concern deaths of children between the ages 1–14 years.

Methods

Study population

The data come from the annually updated individual-level registers maintained by Statistics Finland. The data set was formed by combining information from longitudinal population statistics and causes of death statistics with the help of personal identification numbers. It includes a representative 11% sample of the Finnish population during 1987–2003. To maintain statistical power in the mortality analyses, a random over-sample of the population that died during the period was added to the data set so that altogether 80% of all deaths in the period were covered. Appropriate weights were used in the analyses to account for the unequal sampling probability.

Excluding infant deaths, we restricted the sample in this study to 1- to 14-year-old children who had been born in Finland. Immigrants, who settled in Finland during the follow-up period, were not included in the analysis. The baseline consisted of all 1- to 14-year-old individuals at the end of 1989. Children born in 1989 or later during the following 15 years entered the analysis on the month of their first birthday. Subjects became censored when reaching the age of 15 years or at the end of the follow-up in 2004. In addition, a small number of emigrants became censored at the end of the calendar year preceding their move abroad. The unweighted data included altogether 1 454 643 person-years, 201 211 subjects and 1780 deaths. The median follow-up time was 7.3 years per subject.

Measurement of mortality

The death statistics cover the deaths in Finland or abroad of persons who were at the time of death residents in Finland. The determination of the cause of death is based on the medical or forensic evidence providing the grounds for the issuing of a death certificate by a medical doctor.¹⁶ Total mortality was broken down to mortality from diseases and accidental and violent causes of death according to the ICD-10 classification. For consistency, deaths due to the Asian Tsunami of December 2004 were excluded from analysis.

Measurement of family characteristics

Children were linked to parents on a household basis, i.e. children were matched with the adult or adults they lived with, regardless of biological relationship. The linkage failed in ∼1% of cases. These individuals that mostly consisted of children in institutional and foster care were excluded from further analyses. Data on the family characteristics were registered at the turn of each year. To allow changes in the measures during follow-up, all the variables were included in the analyses as time-varying covariates. As the measures were all household based, any changes in the household composition following, for example, a divorce or forming a reconstituted family, showed also in the respective covariates. The used measures and their classifications are listed below.

• ‘Family type’ classified into married two-parent families, cohabiting two-parent families and single-parent families. Two-parent families include both intact and reconstituted families. A non-married couple with no common child was defined as a cohabiting two-parent family only if the two adult residents were of different sex and not siblings, over the age of 18 years and their age difference was <16 years.

• ‘Degree of urbanization’ based on grouping of municipalities into urban, semi-urban and rural according to the proportion of population living in urban settlements, or the size of the largest urban settlement in the municipality.

• ‘Parental age at child’s birth’ classified into <25 years, 25–29, 30–34 or ≥35 years. The mother’s or stepmother’s age was used in the first place, and father’s age for children living with a single-parent father.

• ‘Number of children’ classified into 1, 2, 3 or ≥4 children of fewer than 18 years of age in the family.

• ‘Parental education’ based on the highest completed degree or certificate: those with basic level qualifications have completed up to 9 years of education, those with secondary education 11–12 years and those with higher education at least 13 years. The level of parental education was determined on the basis of the highest parental education in the household according to the dominance approach.

• ‘Equivalized household income’ (later on referred to as household income) is based on the data concerning income subject to state taxation. This consists of wages and salaries, entrepreneurial income and other income such as other earned income, pension income and unemployment benefits. To allow comparison between households of different size and structure, the household income was divided by the size of the consumption unit determined as the sum of the weights of its members following the OECD recommendation (first adult aged ≥18 years contributes 1.0, subsequent over 13-year-old persons 0.5 and children aged 0–13 years 0.3). In the regression models, income was divided into annual quintiles.

Statistical methods

All-cause mortality rates and HRs of mortality were calculated by each covariate. The relative effect of family characteristics on the risk of death was estimated using Cox proportional hazards models¹⁷ with attained age as the time scale. Due to a clear sex difference in the level of mortality and a significant mortality decline during the 15-year follow-up, the analyses were stratified on sex as well as calendar time period (1990–2004 divided into three 5-year periods). The results are presented for males and females together as the patterns of mortality differentials by family type were highly similar between the sexes. We also present the results for the whole 15-year study period although there was some evidence that among boys aged 1–4 years mortality differentials by family type varied in time. The degree of urbanization of the living area was included in all models as a control variable as in Finland rural areas have higher mortality rates and single-parent families more often live in urban municipalities. First, we calculated hazard ratios of mortality for each covariate separately, controlling only for the degree of urbanization (Model 1). Next, in Model 2, other family characteristics (parental age and number of children) were added to the model simultaneously, and finally factors reflecting family’s socio-economic position (parental education in Model 3 and household income in Model 4). All analyses were conducted with Stata 10.¹⁸

Results

Despite the growing proportion of families with cohabiting or single parents, over 70% of the person-years in ages 1–14 years were still spent in married two-parent families during 1990–2004. The all-cause mortality rates (figure 1) showed no significant difference between children in married and cohabiting two-parent families whereas children in single-parent families had clear excess mortality. The strength of the association was, however, tied to the child’s age at death. In the youngest age group, aged 1–4 years, the mortality for children of single parents was >80% higher than among children in married parent families. In contrast, a less salient difference existed among those aged 5–9 years and no significant differentials could be found among those aged 10–14 years. Due to the absence of significant differences in both absolute and relative mortality in the oldest age group, further analyses are presented only for the age groups 1–4 and 5–9 years.

The relative mortality differentials (tables 1 and 2) showed a similar pattern with no difference between married and cohabiting-parent families, but over double (aged 1–4 years) and >40% higher (aged 5–9 years) mortality risks for children in single-parent families. In the younger age group (table 1), adjusting for parental age at birth and number of children in the family had no major effect on the association between single-parent families and mortality, but the adjustment did accentuate the excess mortality risk associated with families of three or more children. Both of these associations weakened when level of parental education and household income were taken into account. In the full model, the risk of mortality was still elevated for children of single parents (1.38, 95% CI 1.07–1.79), as well as for older parents. The strongest differentials concerned nevertheless children of the less-educated and less-earning parents.

Among those aged 5–9 years, cohabiting-parent families were much less common and single-parent families more common than among the younger children (table 2). In this age, adjusting for parental age and number of children slightly attenuated the association between single-parent families and mortality, and further adjustment for parental education rendered the association insignificant. Compared with the ages 1–4 years, the socio-economic gradients were likewise weaker to begin with and after all adjustments only basic parental education was significantly associated with excess mortality. The effect of the number of children in family differed also from the younger age group. For children, aged 5–9 years, the mortality risk was clearly lower among families with two children whereas being the only child or one of three or more carried equal risks.

The relative excess mortality among children of single parents was more accentuated in accidental and violent causes of death (table 3). In deaths due to diseases, higher risk for children of single parents existed only among those aged 1–4 years. Children in cohabiting-parent families had the lowest mortality risk due to diseases among those aged 5–9 years but as the number of deaths in this group was quite low, the result might be coincidental. Consistently, adjusting for socio-economic factors resulted in disappearance or diminution of the difference between single-parent and two-parent families. Among those aged 1–4 years, however, excess mortality due to accidental and violent causes remained high (1.78, 95% CI 1.20–2.65) even after all adjustments.

We further evaluated whether different associations by the family type of the child’s year of birth as opposed to the current family type could be observed (results not shown here). Among those aged 1–4 years, the results were highly similar, although the association between mortality and current family type was slightly stronger and more resistant to adjustments. Among those aged 5–9 years, the difference was clearer: the hazard ratio in all-cause mortality for children born into single-parent families was 1.32 (95% CI 0.88–1.99) whereas for the time-varying measure it was 1.61 (95% CI 1.20–2.15). In fact, among children born into a single-parent family, a quarter of the person-years in ages 5–9 years were actually spent in married and 19% in cohabiting-parent families.

Discussion

We studied the effects of family type on mortality at ages 1−14 years in Finland in the 1990 s and early 2000 s. As in many other high-income countries of Europe, the level of child mortality in Finland is low.¹⁹ Our results are likely to be relevant for countries with relatively similar family structures and may also be interesting for countries where family forms are converging towards those observed in the Nordic countries. We showed that living with a single parent increased the risk of mortality among children aged 1–9 years whereas mortality of children in cohabiting-parent families showed no difference from children of married parents. The differentials were mainly due to accidental and violent causes of death, but among those aged 1–4 years the risk for children of single parents was clearly elevated also in disease mortality.

The excess mortality among children in single-parent families was largely explained by socio-economic factors, although among those aged 1–4 years family type remained an independent risk factor even after all adjustments. On an average, single parents were younger than other parents and the number of children in single-parent families was fewer but these family characteristics showed at least partly independent and less straightforward associations with child mortality. Both older parental age and larger number of children in the family were significant risk factors only among children aged 1–4 years and only in accidental and violent mortality. On the other hand, the beneficial effect of having two children in the family among those aged 5–9 years was confined to deaths due to diseases, which might point to selection effects among families with only one child.

Lower level of parental education and lower household income were both strongly associated with single parenthood. For example, over half of the children in single-parent families fell into the lowest income quintile. In international comparison, poverty among single parents is still low in Finland.^20,21 Nevertheless, considering the clear association between lower household income and mortality in this study, deterioration of the financial situation of single-parent families might prove to be a cause of concern. During 1990–2003 in Finland, relative poverty in the child population grew faster than in the whole population and especially families with many children, young children and single-parent families had not been able to keep up with others.²² Similar development has occurred in many other developed countries: during the 1990s, child poverty increased in 17 out of 24 OECD countries.²⁰

Lack of parental time to care for children could be argued to be even more detrimental than lack of financial resources. The stronger differentials in accidental and violent deaths lend some support to the hypothesis that single parents are able to impose a lesser amount of control over their children.²³ Single parents cannot rely on the support of a partner to the same extent as married or cohabiting parents. In addition, the kin and other social networks of a single parent are also likely to be more limited, at least in numbers. One limitation of register data in this respect is the inability to identify the living arrangements of children in joint custody. In Finland, children can be registered in only one address, but children in joint custody may live alternately with both of their parents and hence benefit from both parents’ resources.

In Finland, parental separation is nowadays the major path into single parenthood. Based on survey data, at early 1990s rates, 22% of a birth cohort is expected to experience living with a single parent by the age of 15.²⁴ In contrast, living with a single mother in the first year of life is still relatively rare (8% in 2005).¹⁰ Greater role of selection into single parenthood from the child’s birth rather than through parental separation might be hypothesized to entail a more disadvantageous position and thus a larger impact on mortality. Our results comparing the effects of family type at birth and family type as a time-varying covariate gave, however, no support to this idea, but instead highlighted the importance of the current family type on child mortality.

In this study, mortality of children of cohabiting parents did not differ from children of married parents. On the one hand, the role of cohabiting as a temporary phase on a path to marriage makes it likely that many cohabiting parents are qualitatively similar to married parents. On the other hand, cohabiting-parent families are also a very heterogeneous group as people may choose cohabitation over legal marriage for several reasons.¹¹ The bonds between members of cohabiting-parent families may be looser or tighter, for example, reconstituted families with no common children are unsurprisingly much more fragile than cohabiting couples with only common children.²⁵ As the household-based measure made no distinction between intact and reconstituted families, our data enables no comment on the potential negative effects of living in reconstituted families. Anyhow, such analysis regarding mortality during childhood might prove less fruitful as reconstituted families are still relatively rare (during 1990−2004 in Finland, 7−9% of all families with children under the age of 18 years).¹⁰

To conclude, based on extensive register data with annual updates from 15 years, children living with single parents had higher mortality than children in two-parent families, and this difference was strongly associated with low parental socio-economic position. Counter to expectation, children living with cohabiting parents did not experience any excess mortality. During the study period, child mortality has followed a constant decline, and the absolute number of deaths among children is fortunately rather low. Nonetheless, if mortality is regarded as only a tip of the iceberg of numerous other acute and chronic health conditions, mortality differentials offer us valuable knowledge on inequalities in health in childhood.

Funding

The Ministry of Social Affairs and Health in Finland and the Academy of Finland.

Conflicts of interest: None declared.

Acknowledgements

We thank Statistics Finland for granting access to the data (permission TK 53-601-06).

References