Article Text
PDF
Abstract
Background Despite rises in reconstituted and lone-parent families, relatively little is known about how the health of children in different family types varies, and the extent to which any differences might be explained by poverty. The authors examined this using cross-sectional data on 13 681 seven-year-olds from the Millennium Cohort Study.
Methods The authors estimated RRs and 95% CIs for having poor physical (general health, long-standing illness, injury, overweight, asthma, fits) and mental health (using strengths and difficulties scores) according to family structure using Poisson regression. The authors adjusted for confounders (aRR) and then investigated the role of poverty as a mediator by entering a poverty score (based on income, receipt of benefits, subjective poverty and material deprivation) into the main model.
Results Children living in reconstituted and lone-parent families were at a slight increased risk of poor health compared with those living with two natural parents. Adjusting for poverty tended to remove the elevated risk of poor physical health in children living in lone-parent and reconstituted families. However, for the mental health outcomes, poverty tended to remove the elevated risk for lone parents but not for reconstituted families. For example, the aRR for borderline–abnormal total difficulties fell from 1.45 (1.22 to 1.72) to 1.34 (1.13 to 1.59) in children living in reconstituted families and from 1.29 (1.14 to 1.45) to 1.05 (0.92 to 1.19) in those living with lone parents.
Conclusions Poor physical and mental health was slightly more prevalent in children living in lone-parent or reconstituted families. Poverty reduction may help to reduce these differences, especially for children living with lone parents; however, alternative mechanisms should be also explored, particularly for children living in reconstituted families.
- Child health
- inequalities
- poverty
- policy
Statistics from Altmetric.com
Introduction
Family structure in the UK has changed significantly in recent decades, for example, the proportion of lone-parent families increased from 13% in 1961 to 27% in 2009.1 It is now estimated that the UK has the highest prevalence of children living in lone-parent families in the European Union.1 Reconstituted families are also on the increase; however, national statistics have yet to capture more complex family structures.1 These changes in family structures have led to a body of research to determine if living in different family types (and experiencing changes in family structure) affects the children and adolescents living in them. Children from lone-parent and reconstituted families are more likely to be born with low birth weight and to have mothers who suffer from depression or smoke.2 These children are also at increased risk of behavioural problems,3–5 poor health6 and less favourable outcomes in adulthood (such as poor educational achievement, lower occupational status and premature mortality).7 ,8 However, most research has been carried out in children pre-1980s when some family structures were less common7 ,9 ,10 or has focused on longer term outcomes in adolescence or adulthood.7–9
In the recent Child Poverty Strategy, the UK coalition government announced a shift in approach from that adopted by the previous administration, focusing on tackling the root causes of disadvantage (which was stated to include family structure) in addition to lifting families out of poverty through boosts in income.11 According to data from the Family Resources Survey, in 2009/2010, 28% of children residing with a lone parent were living in poverty compared with 17% in couple families. After accounting for housing costs, these figures were 46% and 24%, respectively.12 However, there is little research exploring the extent to which higher poverty rates experienced by children living in these family types explains any differences observed in their physical and mental health.
Using contemporary and nationally representative data, this study provides a description of the physical and mental health of 7-year-old children (in 2008) according to family structure, to inform hypotheses about the role of poverty as a mediator between family structure and health.
Methods
Subjects and design
We examined data from the Millennium Cohort Study (MCS), a longitudinal study of children born in the UK between September 2000 and January 2002. A disproportionality stratified clustered sampling design was used to over-represent children living in Wales, Scotland and Northern Ireland, disadvantaged areas and areas with high proportions of ethnic minority groups.13 The first contact with the cohort child was at age 9 months, when information was collected on 72% of those approached, giving 18 296 singleton infants. Survey interviews were carried out by trained interviewers in the home with the main respondent (usually the mother) and their partners. Since then, data have been collected at three further sweeps when the children were aged 3, 5 and 7 years old. This paper uses data on 13 634 (75%) singleton children, who took part in the most recent sweep (age 7), which was conducted in 2008. Children from ethnic minority groups and from less advantaged backgrounds were less likely to take part in the survey at age 7 than those from Caucasian and more advantaged groups. Response weights are used in all analyses to take into account this attrition and also the sample design.14 Data were obtained from the UK Data Archive, University of Essex in May 2010. Ethical approval was received from the Northern and Yorkshire Multi-Centre Research Ethics Committee for the fourth sweep; the present analysis did not require additional ethics approval.15 Further information on the MCS can be found at website (http://www.cls.ioe.ac.uk/MCS).
Family structure
The child's current family structure was reported by the main respondent and was classified as: ‘natural couple families’ consisting of two ‘natural’ parents (wording used in the survey), ‘reconstituted families’ consisting of one ‘natural’ parent and one adoptive/foster or step parent and ‘lone parent families’ where there was one ‘natural’ father or mother. Atypical family structures were excluded due to low numbers: adoptive/foster parent(s) where there was no natural parent (n=13), lone grandparent (n=19), two grandparents (n=22), lone ‘other’ parent (n=7), two ‘other’ parents (n=2), leaving a sample of 13 618.
Physical and mental health
A number of health measures were explored, based on those available from the MCS data set (table 1).
Description of outcome measures
Potential confounders
Family size (at age 7), child's ethnicity, the age of the main carer, the child's sex, gestational age and maternal malaise (at age 9 months) were examined as potential confounders because they were associated with health and family structure.
We also explored a number of measures representing socioeconomic circumstances (SECs) at age 9 months (mother's social class (NS-SEC), tenure and highest educational qualification). We considered SECs to be potential confounders rather than mediators because they are likely to be less sensitive than poverty to changes in family structure.
Only measures that contributed to the explanatory power of the model (using likelihood ratio tests) were included in the adjusted analyses. Because measures of SECs are likely to capture aspects of poverty, we adjusted for confounders in two stages: demographic and health factors (age, sex, malaise and ethnicity) followed by the SECs measures (NS-SEC and education). We tested for multicollinearity between poverty, NS-SEC and education, using the ‘Variance Inflation Factor’. All Variance Inflation Factor values were under 5 indicating that multicollinearity would not pose a problem in the model.16
Poverty as a mediator
Poverty was modelled as a potential mediator because family structure can lead to changes in wealth, which can in turn influence health via mechanisms such as ability to afford healthy foods, activities and home environments. A poverty score was created to measure ‘multiple poverty’,17 ranging from 0 to 4 and based on how many of the following four components of poverty families were rated as ‘poor’: income poverty (<60% national median income), material poverty (not able to afford one or more of birthday celebrations, annual holidays, money to spend on self, two pairs of shoes, a weather-proof coat), subjective poverty (whether the main respondent felt that they were just about getting by financially or worse) and receipt of benefits (income support, working tax credit, housing benefit or council tax benefit).
Statistical analysis
All analyses were conducted in STATA/SE V.11.2 (Stata Corporation). Poisson regression was used to calculate RRs and 95% CIs. First, we described patterns of poor physical and mental health according to family structure, estimating percentages using survey weights. Using the conditions for mediation defined by Baron and Kenny,18 we then explored the role of poverty as a potential mediator between family structure and health as demonstrated in figure 1 and described below:
Diagram to demonstrate the four stages of the analysis to test poverty as a mediator of the association between family structure and child health.
-
A. Using Poisson regression, we estimated unadjusted RRs (uRR) for having poor outcomes for each aspect of health according to family structure (with children living in natural couple families as the baseline). We then estimated adjusted RRs (aRR) controlling first for demographic and health confounders (aRR[1]) and then for demographic and health confounders and also SECs (aRR[2]).
-
B. We then explored the association between family structure and poverty, adjusting for demographic and health confounders (aRR[1]).
-
C. Following this, we investigated the association between the poverty score and the health outcomes, adjusted for demographic and health confounders (aRR[1]).
-
D. Lastly, we entered the poverty score into the fully adjusted (aRR[2]) model shown in A below. Mediation was taken to be a reduction in, or elimination of, statistically significant RRs.
All analyses in A–D were carried out using complete samples so that RRs could be directly compared before and after adjustment. The level of missing data for physical health and confounding variables ranged from 0 (for sex and family size) to 915 (for maternal malaise), as listed under table 2. The Strengths and Difficulties Questionnaire (SDQ) data had a higher level of missingness, reaching 2720 for the total difficulties score. Each component of the SDQ is made up of five items. As recommended,19 where individuals were missing one to two items on one component, we took the average value from the remaining three to four items to make up a complete component, which reduced the level of missing data substantially (eg, from 2720 to 448 for the total difficulties score).
Weighted percentage (N) of children according to physical health, mental health (Strengths and Difficulties Questionnaire (SDQ)) and poverty, by family type
A number of sensitivity analyses were also carried out. We examined whether health differences between children in reconstituted and lone-parent families were statistically significant by repeating the analyses with lone parents as the baseline. Family structures can be transient and so we constructed a variable indicating whether there had been any change in family structure since age 5. We used likelihood ratio tests to examine whether adding this variable to the main model containing the cross-sectional family structure variable explained any additional variance in the outcome. We also repeated the main analyses investigating the association between family structure and health, limited to children who had been in the same family type at age 7 and 5.
Results
Seventy per cent (n=9859) of children were living with both natural parents, 8% (n=889) in reconstituted families and 23% (n=2870) with a lone parent. As shown in table 2, all aspects of physical and mental health varied significantly by family type, with children living in natural couple families consistently having the best health. However, absolute differences were small for physical and mental health. For example, the proportion of overweight children was just 4% higher in lone-parent families than in natural couple families. Compared with children living in couple families, the prevalence of abnormal total difficulties scores was 14% higher for children in reconstituted families and 10% for those in lone-parent families. Only small proportions reported having fair or poor general health in all groups (ranging from 2.4% to 4.7%).
Poverty levels varied significantly by family structure (table 2). Lone parents had the highest levels of all four components of poverty, followed by reconstituted families. There were large differences in the poverty score by family structure; 82% of lone parents were experiencing two or more aspects of poverty compared with 58% reconstituted families and 33% couple families. The association between family structure and the poverty score remained significant after adjusting for health and demographic confounders (aRR[1]) (table 3). The poverty score was also positively associated with all measures of physical and mental health after adjustment (table 4).
Adjusted RRs* (95% CI) for living in poverty according to family structure
RRs (95% CI) for demonstrating poorer physical health and borderline–abnormal mental health according to the poverty score
Table 5 presents RRs and 95% CIs for having poor physical and mental health, before (uRR, column A) and after adjustment for confounders (aRR[1], column B), SECs (aRR[2], column C) and poverty (aRR[3], column D). For almost all aspects of health, children living in lone-parent families and reconstituted families had poorer outcomes than those living in natural couple families (except for overweight, which did not vary between reconstituted and natural couple families). Adjusting for health and demographic confounders (aRR[1]) attenuated the association for most outcomes in both reconstituted and lone-parent families (table 5, columns C and D) and in some cases removed the elevated risk of poor health in reconstituted families altogether (for general health, injury, fits). Adjusting for SECs (aRR[2]) removed the elevated risks of injury in children living in lone-parent families and asthma and emotional problems in reconstituted families.
RRs (95% CIs) for poor physical and mental health according to family structure and before and after adjustment for confounders and poverty
As shown in column D, poverty removed the elevated risk of poor health in lone-parent families in almost all cases where adjusting for confounders and SECs had not already done so (for general health, long-standing illness, overweight, asthma, fits, and borderline–abnormal total difficulties, and emotional, conduct and hyperactivity problems). The elevated risk of peer problems was halved but remained significant. Poverty did not appear to be as important in reconstituted families in explaining the elevated health risks. Where RRs remained significant after adjusting for confounders and SECs (in the case of total difficulties, conduct problems, peer problems and hyperactivity scores), poverty attenuated the elevated risks but did not remove them.
Direct comparison of children living in reconstituted families and those living in lone-parent families suggested that there were few significant differences in the risk of poor health between these two groups. Children living in a reconstituted family were less likely to be overweight or obese (uRR 0.78 (0.62 to 0.92)) and more likely to have a borderline–abnormal hyperactivity score (uRR 1.35 (1.16 to 1.56)) than those living with a lone parent. After adjustment for all confounders and poverty, these differences remained significant, and children from reconstituted families also became significantly more likely to have a borderline–abnormal total difficulties score (aRR[3] 1.28 (1.05 to 1.56)). A number of MCS children (9%) experienced a change in family structure between age 5 and 7, and the most common changes were from natural couple family to lone parent (48%) and from lone parent to reconstituted family (27%). Adjusting for change in family structure since the previous sweep did not add to the explanatory power of the original model, and the RRs by current family structure were unchanged (data not shown). Finally, we repeated the main analyses exploring the association between family structure and health limited to children who experienced the same family structure at both time points (table S1 supplement). The RRs and patterns of mediation remained the same, with one exception: the elevated risk of conduct disorders in reconstituted families was no longer significant after adjustment for confounders.
Discussion
Summary of findings
Children living in reconstituted and lone-parent families experienced slightly higher rates of poor physical and mental health than those in natural couple families. While all differences were statistically significant, absolute and relative differences were relatively small, and only a small proportion of mothers perceived their child to have fair or poor general health. Children living in poverty were more likely to experience poor health, and, as expected, children living with lone parents experienced the highest rates of poverty, followed by reconstituted families. Entering the poverty score into the main regression model between family structure and health (after adjusting for all confounders) tended to remove all the elevated risks of poor health in children living in lone-parent families. While it explained some of the elevated risk in children from reconstituted families, a sizeable proportion of the risk remained for SDQ outcomes, indicating that other factors are also at work. Additional analyses indicated that, compared with those living in lone-parent families, children from reconstituted families were at a higher risk of hyperactivity and total difficulties scores. When limiting the associations to families which had not experienced change in family structure since the last sweep, findings remained similar with one exception: after adjusting for confounders, there was no longer an elevated risk of conduct disorders in reconstituted families. This could imply that the elevated risk of conduct disorders in children living in reconstituted families is short lived or can be easily modified but further research in this area is required.
Strengths and limitations
To our knowledge, this is the first study to explore the role of poverty as a mediator between family structure and child mental and physical health in a contemporary sample. We were able to do this in a large data set (n=13 618) of children aged 7 years in 2008. It is possible that our analyses lacked statistical power in some cases, particularly where health outcomes were rare (eg, fair–poor general health) and for reconstituted families which was a relatively small group. The wide range of information available in the MCS allowed us to adjust for a range of potential confounding factors and to explore a number of poverty measures including material hardship. Adjusting for SECs attenuated the association between family structure and health. It is likely that SECs capture elements of poverty as well as social position, and therefore, the effect of poverty as a mediator has most likely been underestimated. Our a priori hypothesis was that poverty would mediate the association between family structure and health due to the higher levels of poverty in lone-parent and reconstituted families. However, it is also possible that poverty modified the association between family structure and health. We therefore tested for interactions and results indicated that in this instance mediation rather than modification was occurring.
Previous research in the MCS indicates that changes to family structure are followed by changes in income.2 However, it is also possible that SECs and poverty can lead to changes in family structure. We attempted to address this limitation, through adjusting for SEC measures captured in infancy. The poverty score included the receipt of benefits, and a higher proportion of lone parents were receiving these than other family types (due to income support eligibility). We therefore repeated our analyses with a reduced poverty score based only on income, material deprivation and subjective poverty to examine whether the extent of mediation between family structure and health changed. The RRs after adjustment for the reduced poverty remained very similar to those adjusted for the full poverty score.
We were able to explore a range of health outcomes. Overweight was captured using objective measures of height and weight, and mental health was explored using a validated questionnaire completed by the main respondent.20 The remaining measures were captured using questions that are frequently used in cohorts and health surveys but, to our knowledge, are not validated. General health and long-standing illness of the child were reported by the main respondent; there is evidence that adult self-reported general health is an independent predictor of mortality21; however, parents have been found to over-estimate the perceived general health of their child when compared with the child's self-report.22 Parental-reported long-standing illness in children has been found to reflect children's reporting of their own long-standing illness.22 Unintentional injury was based on maternal report of whether the child had been taken to see a medical professional for an accidental injury. Studies have shown a reasonable to high level of agreement between maternal recall of injury and medical records, with no differences by socioeconomic characteristics.23 ,24 It is possible that the accuracy of maternal report of child health varies by family structure, for example, due to heightened awareness of health problems, social desirability and/or normalisation, and selective recall. However, there is a lack of evidence exploring this so it is not possible to make hypothesises about the potential impacts on the associations observed in this paper. General health and long-standing illness were classified as physical health in this paper, although both are likely to also capture aspects of mental health.
Sampling and response weights were used in all analyses to account for the sampling design and attrition. However, the response weights may not have entirely accounted for differences observed in health and family structure or in cases where children had missing data. Compared with children living in couple families, those from reconstituted families at age 5 were 6% (95% CI 4% to 9%) and lone-parent families were 7% (95% CI 5% to 8%) less likely to take part at age 7. We carried out analyses using complete samples, which ranged from 12 501 for general health and fits to 12 221 for the SDQ total difficulties score. While it is unlikely that the impact of family structure on health would have been substantially different for those who were and were not included in our analyses, we were unable to assess this further.
We explored a simple measure of family structure, which cannot capture the complexities and differences that may exist within these broad groups (eg, level of parenting support from partners in natural couple or reconstituted families or from grandparents in lone-parent families). In addition, we had to exclude a number of uncommon family types, such as lone adoptive/foster parents, due to small numbers. Although the measures used to derive family structure referred to natural and non-natural parents, we cannot be entirely sure that ‘natural’ fathers were always a birth parent. Family structures are transient and 9% of the MCS children had experienced a change in family structure between age 5 and 7. Sensitivity analyses were carried out to explore the potential impact of change in family structure on the associations observed, and the relationships observed in the main models did not appear to be dependent on change in family structure in most cases. However, the elevated risk of conduct disorders seen in children living in reconstituted families did not persist when only explored in children who had experienced no change in family structure since the last sweep. This may indicate that the potential impacts of family change on socio-emotional health may be short lived. Due to the cross-sectional nature of these analyses, we are not able to make inferences about causality. However, the purpose of this paper was to describe the social patterning of health in a large, contemporary and nationally representative sample of children in order to inform future hypotheses about the potential role of poverty for further research.
Comparison with other findings
In 2009, 24% of dependent children in the UK were living with a lone parent[1], rates were comparable for children aged 7 years in the MCS (23%). Data from the Family Resources Survey in 2008/2009 suggest that 18% of couple families and 34% of lone-parent families were living in households with <60% average income.12 While rates for natural couple families were similar in the MCS (18%), rates in lone-parent families were a lot higher in the MCS (63%). This is likely to be due to differences in sampling frames of the two studies.
Our findings are in agreement with other research exploring the effects of growing up in different family structures. Findings from the first sweep of the MCS showed that birth weight, maternal depression and smoking varied by family structure,2 and findings from the 1997 Health Survey for England found that children (aged 4–15 years) living in lone mother or reconstituted families were at increased risk of behavioural and emotional problems (using SDQ).5 Similarly findings from the Avon Longitudinal Study of Parents and Children indicate that children from reconstituted and lone-parent families were at an elevated risk of emotional and behavioural problems at age 4.3 Studies focusing only on the health of children living in lone-parent families (and not those in reconstituted families) have also found these children are at risk of poorer health compared with those living with two parents. For example, a cross-sectional analysis of 0–18-year-olds in the 2001 Families and Children Study (FACS) indicated that lone parenthood was associated with increased risk of poor general health and long-standing illness, educational problems and antisocial behaviour.6 An analysis of almost 1 million records from Swedish national registers found that children living in lone-parent families had increased risks of psychotic disease, suicide attempts, alcohol- and narcotics-related disease, in adolescence and early adulthood.8
We found that after adjustment for confounders, poverty attenuated the association between family structure and physical health. However, the elevated risk of poor socio-emotional health remained in children living in reconstituted families. Some of the previously mentioned studies also sought to explore financial situation as a mediator and also found it important for those in lone-parent families and less so for those in reconstituted families.3 ,5 One of the studies focusing on only lone-parent and natural couple families found that the elevated risks of poor outcomes (in the Families and Children Study) was removed after adjusting for household hardship and tenure,6 whereas the other (using Swedish registry data) found that, while the risk attenuated after adjusting for SECs and housing situation, it remained statistically significant.8
Implications for further research, policy and practice
Investment in early years is an important component in the current Coalition government's policy agenda.25 However, there has been a shift in focus away from lifting families out of poverty through financial support to increasing social mobility, addressing the root causes of poverty and providing targeted support to at risk groups.11 ,26 We found that children living in poverty were more likely to experience poor health, and after adjustment for poverty levels, the health of children living in lone-parent families was no different from those living with both natural parents. This implies that short-term steps to reduce poverty will benefit all children who live in poorer circumstances and could also put the health of today's lone-parent children on an equal footing with those living with two natural parents. Our findings also indicate that additional support may be required for reconstituted families whose children are more likely to suffer from behavioural problems, when compared with those living in natural couple families and lone-parent families, and after adjustment for poverty. The Child Poverty Strategy also aims to reduce the impacts of family breakdown.11 Future research should attempt to gain a better understanding of why children living in reconstituted families experience poor mental health in order to help develop or inform targeted interventions, such as the Family Nurse Partnership, Sure Start Children's Centres and the recently announced trial of parenting classes for parents with children aged 0–5 years.27 Further investigation using measures of poverty (such as material hardship) rather than traditional measures of SECs is also required.
Although the elevated risk of children living in lone-parent families was not large in magnitude, many children live in these families in the UK and therefore efforts to improve child health in these families could still have a substantial influence on population health. While children living in reconstituted families are a relatively smaller group, the number of children living in these families is still large and likely to increase in the future. Our analyses indicate that family structure (and its consequences, such as poverty) is associated with child health and that these relationships are, on the whole, independent of family structure change. However, the literature indicates that family transitions do become influential as children enter adolescence and so research involving later sweeps of the MCS should fully exploit the longitudinal nature of the data to explore this. The Child Poverty Strategy highlights the importance of differentiating between persistent poverty, transient poverty (experiencing poverty for only for short periods) and cyclical poverty (movement in and out of poverty). Future research should investigate how these different classifications of poverty relate to child health.
What is already known on this subject
-
A small body of research indicates that, compared with children living with two natural parents, children living in reconstituted and lone-parent families experience higher rates of poor physical and mental health.
-
However, it is not yet clear what role poverty plays in explaining the variation in child health by family structure, particularly in more contemporary samples where rates of lone-parent and reconstituted family types and poverty are high.
What this study adds
-
Using data from the UK MCS on 7-year-old children (in 2008), we found that those living in reconstituted and lone-parent families suffered slightly higher rates of poor physical and mental health compared with those living with two natural parents.
-
However, after adjusting for poverty, the health of children living with lone parents no longer differed from those with two natural parents.
-
For children living with reconstituted families, the elevated risks of poor health often remained after poverty was taken into account (especially for mental health).
-
Poverty reduction may help to improve the health of all children and especially those living with lone parents; however, alternative mechanisms should be also explored, particularly for children living in reconstituted families and with regard to mental health.
Acknowledgments
We would like to thank all the Millennium Cohort families for their participation, and the director of the Millennium Cohort Study (MCS) and colleagues in the management team at the Centre for Longitudinal Studies, Institute of Education, University of London. We would also like to thank the MCS user group at the MRC Centre of Epidemiology for Child Health: Carol Dezateux, Lucy Griffiths, Tim Cole, Helen Bedford, Carly Rich, Phillippa Cumberland, Jamie Fagg, Steven Hope, Francesco Sera, Mario Cortina-Borja, Marco Geraci. We would also like to thank members of the Policy Research Unit for the health of children, young people and families: Terence Stephenson, Ruth Gilbert, Russell Viner, Miranda Wolpert, Amanda Edwards, Steve Morris, Helen Roberts, Catherine Shaw and finally, Mark Petticrew and Margaret Whitehead.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Data supplement 1 - Online Table S1
Footnotes
-
Funding This work was supported by the Policy Research Unit in the Health of Children, Young People and Families (funding reference 10090001), which is funded by the Department of Health Policy Research Programme. This is an independent report commissioned and funded by the Department of Health. The views expressed are not necessarily those of the Department. The Centre for Paediatric Epidemiology and Biostatistics was supported in part by the Medical Research Council in its capacity as the MRC Centre of Epidemiology for Child Health. Research at the UCL Institute of Child Health and Great Ormond Street Hospital for Children receives a proportion of the funding from the Department of Health's National Institute for Health Research Biomedical Research Centres funding scheme. The Millennium Cohort Study is funded by grants to former and current directors of the study from the Economic and Social Research Council (Professor Health Joshi and Professor Lucinda Platt) and a consortium of government funders. The study sponsors played no part in the design, data analysis and interpretation of this study; the writing of the manuscript or the decision to submit the paper for publication, and the authors' work was independent of their funders.
-
Competing interests None.
-
Provenance and peer review Not commissioned; externally peer reviewed.