Background Children from disadvantaged backgrounds are more likely to experience unintentional injuries and poor home environments. The aim of this study was to explore the home environment as a potential mediator between socioeconomic circumstances and unintentional injuries, in the UK Millennium Cohort Study (n=14 378).
Methods RRs and 95% CIs for being injured in the home between age 9 months and 3 years were estimated according to four measures of socioeconomic circumstances: social class, maternal education, lone parenthood status and tenure. Proxy indicators of housing quality (build type, storey, garden access, rooms per capita, central heating and presence of damp) and safety equipment use (use of fireguards, safety gates, electric socket covers and smoke alarms) were then controlled for in order to observe potential mediation.
Results Children from routine and manual backgrounds were more likely to be injured than those from managerial and professional backgrounds (RR=1.33, 95% CI 1.21 to 1.47), as were children of lone parents (compared with couple families) (RR=1.23, 95% CI 1.12 to 1.36), those whose mothers had no educational qualifications (compared with a degree) (RR=1.42, 95% CI 1.24 to 1.63) and those living in socially rented accommodation (compared with owned/mortgaged homes) (RR=1.35, 95% CI 1.24 to 1.46). However, controlling for the indicators of housing quality and safety equipment use did not alter the elevated risk of injury experienced by children from less advantaged backgrounds.
Conclusions In this contemporary UK cohort, proxy indicators of the home environment did not appear to explain socioeconomic inequalities in injuries. Research exploring alternative explanations for inequalities in injuries could help contribute to the design or adaptation of policies to reduce them.
- Child accidents
- housing and health
- social inequalities
Statistics from Altmetric.com
Although rates have decreased in recent years, injury remains the main cause of death and morbidity in childhood in the Western world.1 Mortality2 and morbidity3–5 from childhood injuries are highly socially distributed and have widened as injury rates have fallen.6 Furthermore, data from hospital admission data in the Trent region, UK, found that the socioeconomic gradient was more marked in preschool than in older children.4 Most injuries in preschool children occur in the home,7 where they spend most time. Exploring their surroundings is crucial to children's development; however, the development of their physical ability precedes their awareness of risks.7 Young children are therefore particularly at risk of hazards associated with poor quality housing.6
In the UK, indicators of poor quality housing (such as the presence of overcrowding, damp or condensation, living in high storey flats, not having access to a garden and not having central heating) are strongly associated with low income.8 Minimising the impact of poor housing was an important component in the previous UK government's strategy to improve welfare and reduce health inequalities, including inequalities in unintentional injury. Policies included improving the quality of social housing, for example, by providing access to a garden,9 the reduction of overcrowding10 11 and the launch of a national home safety equipment scheme to provide low-income families with free or low-cost safety equipment and consultations.11 More recently, the Marmot et al12 review emphasised the role of housing for achieving a fairer distribution of health.
Given that both the quality of the home environment and childhood injuries are socially distributed, it is possible that the home environment lies on the causal pathway between socioeconomic circumstances (SECs) and unintentional injury. An ecological study in the USA found that housing characteristics (tenure and build type) mediated the association between poverty and injury rates and highlighted the need for individual-level studies exploring this.5 However, to our knowledge, no individual-level studies have set out to explore this. We use data from a contemporary UK cohort to explore indicators of the home environment at age 9 months as potential mediators between SECs at age 9 months and injuries occurring in the home between the ages of 9 months and 3 years.
We examined data from the Millennium Cohort Study (MCS), a longitudinal survey of 18 296 singleton children born in the UK between September 2000 and January 2002. Data were downloaded from the UK Data Archive, University of Essex, in April 2008. The MCS was collected using stratified clustered sampling design to over-represent children living in Wales, Scotland and Northern Ireland, disadvantaged areas, and those with high proportions of ethnic minority groups.13 Ethical approval was received for the MCS from the South West and London Multi-Centre Research Ethics Committees.14 Information was collected by trained interviewers in the children's homes. Further information on the cohort and sampling design can be found in online (http://www.cls.ioe.ac.uk/mcs).
Our analysis utilised data that were collected when the children aged 9 months and 3 years old (n=14 630), excluding children where the main respondent was not the natural mother (n=196). Of these 14 434 children, 14 378 (99.6%) had information on injuries.
Measures of injury
When the children were aged 3 years, mothers were asked if the child had visited a health centre, hospital or general practitioner due to an unintentional injury since the last sweep (when the children were approximately aged 9 months). Mothers were asked to report where the injury had occurred (responses included at home, in a playground, on the road) if there had only been one injury; for those injured more than once, the location of the most severe injury was recorded. We categorised children according to whether they had been injured in the home or not, based on these responses.
Measures of SECs
Measures of SECs were chosen to represent both the mother and household and were all reported at age 9 months. These were social class of the mother (using the National Statistics Socio-economic Classification), highest maternal educational qualification, lone parenthood and housing tenure. Data for mothers who were unemployed or who had ‘other’ educational qualifications (eg, qualifications from overseas) are not presented (because it is hard to understand their socioeconomic status without additional information). Tenure was explored because it is likely to capture aspects of SECs which education, social class and lone parenthood cannot, such as level of control over the home environment.
Indicators of the home environment
The home environment was taken to be represented by housing quality and safety equipment use, using proxy indicators collected when the children were aged 9 months.
The following indicators of housing were reported by the mother and were used to create a score for housing quality. A value of 2 was assigned to a ‘negative feature’, 0 for a ‘positive’ feature and 1 for ‘intermediate’ features (where relevant).
Rooms per capita: <1 room=2; 1 to <2 rooms=1; 2+rooms=0.
Build type: bedsit/other=2; flat=1; house/bungalow=0.
Storey of main living accommodation: third floor or above=2; first/second floor/basement=1; ground floor=0.
Garden: no access=2; shared access=1; sole access=0.
Central heating: no=2; yes=0.
Presence of damp or condensation on walls: yes=2; no=0.
The total score ranged from 0 (households with none of the negative features) to 12 (households reporting all of the negative features). Two thirds of households scored 0 or 1 and very few (5.9%) scored 6 or more (see table 2). Therefore, when considering quality housing as a binary variable (table 1), a score of 4 or more was considered to be ‘poor’. When estimating risk of being injured by housing quality (table 2) and when adjusting for housing quality as a mediator (table 3), the score was treated as an ordered categorical variable and collapsed as follows: 0, 1, 2–3, 4–5, 6+.
Safety equipment use
Mothers were asked if they used any of the following items of safety equipment (replying yes or no):
Electric socket covers
No indication was made of whether the equipment was used correctly or was in working order. Low safety equipment use (table 1) was considered to be using none or just one of the reported items since not all pieces of equipment would be relevant to all households (eg, a score of 2 in households without an open fire and stairs could still indicate reasonable levels of safety equipment use). When estimating risk of being injured by safety equipment use (table 2) and when adjusting for safety equipment use as a mediator (table 3), the score was treated as an ordered categorical variable ranging from 0 to 4.
In the adjusted analyses, the number of children in the family, maternal age at first live birth, ethnicity and main childcare between age 9 months and 3 years were controlled for, because they were identified as potential confounders, or a risk factor for injury, in other analyses.15
The analysis was carried out in four stages, as demonstrated in figure 1 and described below:
Using Poisson regression, we estimated unadjusted RRs and 95% CIs for living in homes with poor quality housing (score of 4 plus) and low safety equipment use (using 0–1 items), according to SECs.
Following this, RRs for being injured were estimated according to the individual measures of housing quality and safety equipment and also the scores.
We then estimated RRs for being injured according to each of the SECs measures.
Finally, adjusted RRs for being injured were estimated according to each of the SECs measures, controlling for housing quality and safety equipment use (entered as ordered categorical variables). Any observed change in RRs (≥10%) was taken to indicate potential mediation. Following this potential, confounders were adjusted for.
A number of sensitivity analyses were also carried out to address some of the limitations in the data. All analyses were conducted in STATA/SE V.10.0 (Stata Corporation), using survey commands to take account of the sampling design and attrition at the second sweep.
Twenty-two per cent of children (n=3151) had been injured at home between age 9 months and 3 years and 14% (n=1957) had been injured somewhere outside the home.
Unadjusted association between SECs and the home environment
Compared with those from the most advantaged groups, children from less advantaged SECs were consistently more likely to live in homes with a poor housing score (≥4) (table 1) and also in homes using one or none of the reported items of safety equipment.
Unadjusted association between indicators of the home environment and injuries occurring in the home
Most (60%) children lived in households with one or no negative housing indicators (table 2). In general, children living in households with poor housing scores were more likely to have been injured in the home than those with no negative housing indicators. When exploring the different indicators of housing quality individually compared with children living in homes with one to two rooms per capita, children living in homes with more than two rooms per capita were slightly less likely to have been injured. While all of the other indicators of poor housing were also associated with increased risks of injury, the associations were non-significant.
Households with between one and three items of equipment were at a slight elevated (but non-significant) risk of injury compared with those with all four (table 2). However, children living in households with no safety equipment were around 20% less likely to have been injured than those with all four. Exploring the association between individual items of equipment and injury indicated that not owning safety equipment was associated with a slightly reduced risk of injury for all equipment types. However, none of the RRs were statistically significant.
Unadjusted association between SECs and injuries occurring in the home
As shown in table 3 (column B), children from lower SECs were significantly more likely to be injured in the home than those from the most affluent backgrounds, according to all measures of SECs.
Home environment as a mediator
The associations between SECs and injuries changed little when controlling for the housing score and safety equipment use (table 3, columns C and D). Controlling for confounders (column E) reduced the size of many of the RRs, and in some cases, they were no longer statistically significant (eg, children living with lone parents were no longer at an elevated risk of injury). Closer inspection revealed that in most cases, this attenuation was largely due to children of younger mothers being more likely to have been injured and also to come from less advantaged backgrounds (data not shown).
In the main adjusted analysis (table 3), we controlled for the housing quality score as an ordered categorical variable (0, 1, 2–3, 4–5, 6+ negative features). We conducted several sensitivity analyses adjusting for the score as a linear variable (ranging from 0 to 12), as a binary variable (0–3, 4–12) and also entering each of the housing measures separately, and each time the findings were unchanged. Comparable sensitivity analyses were also carried out for the safety equipment score, and the findings were unchanged in all instances (data not shown). We also conducted sensitivity analyses to investigate the association between specific items of safety equipment and relevant injury types: the use of fireguards and the risk of burns and scalds in children living in homes with fires (eg, open or gas fires) and safety gate use and the risk of injuries potentially associated with falls (such as head injuries). There was still no association between safety equipment use and injuries in either of these analyses (table 4). Finally, we repeated the analyses excluding children who had been injured more than once and whose most severe injury had occurred outside the home from the baseline; the results remained very similar to those reported here (data not shown).
Summary of findings
Children from disadvantaged SECs were more likely to have experienced unintentional injuries in the home than those from the most advantaged SECs. Children from less advantaged SECs were also more likely to live in poorer home environments (in terms of proxy measures for housing quality and safety equipment use). Few indicators of the home environment were significantly associated with injuries occurring in the home. Most of the associations between SECs and home injuries were not affected when controlling for the indicators of the home environment. This implies that the indicators of housing quality and safety equipment use explored in this analysis and in this context did not mediate the association between SECs and unintentional injuries occurring to young children in the home.
Strengths and limitations
This study was carried out using secondary data from a large and contemporary UK cohort. There was no significant difference in injury rates at age 9 months between children who did not respond to the second sweep and those who did, with an unweighted risk difference of 0.48% (95% CI −0.52 to 1.47%), although those who were not included in the second sweep were significantly more likely to live in a flat or bedsit, in a basement or on a higher floor; have fewer rooms per capita and were less likely to have access to a garden (p=<0.05). However, we were able to use survey weights to take into account the sampling design and differential response. A further strength was the wide range of information collected in the MCS, which allowed us to explore several measures of SECs and various indicators for the home environment.
A limitation of this study was that injury was based upon maternal report. Studies have shown a reasonable to high level of agreement between maternal recall of injury and medical records, although accuracy declines as the period of recall increases16 17 and particularly if it exceeds 2 years.17 The maximum period of recall for the measure of injury used in this paper is approximately 2.5 years but in many cases would be less than this. Studies have also indicated that there is little variation in recall by socioeconomic background.16–18 Mothers were only asked to report injuries for which the child attended a hospital, health centre or GP; therefore, injuries for which no professional advice was sought have not been explored. Seeking professional advice does not necessarily give an indication of the severity of the injury. Evidence suggests that parents from more advantaged backgrounds are more likely to take their child to Accident and Emergency department for minor injuries than those from less advantaged backgrounds.19 20 If such biases exist in the MCS, it is possible that the inequalities in injury we have found have been underestimated.
We were only able to determine whether the most severe (or only) injury occurred in the home or elsewhere, and approximately 3% of children had been injured twice or more and had experienced their most severe injury outside the home. These children were included in the baseline, although it is likely that some had also been injured at home. However, a sensitivity analysis excluding these children returned similar associations to those reported here. While we were able to control for main childcare type, it was not possible to take into account hours spent in childcare (and therefore time spent in the home), and further research might explore this.
A strength of this study was the availability of objective indicators of the home environment meaning that any response bias should be limited. However, these indicators can only be considered as a proxy for housing quality and safety equipment use. One might argue that households which have fewer rooms per capita or no access to a garden or that are situated on higher storeys may pose a greater injury risk for children than homes with outdoor space and more room inside. High rise flats are more likely to have balconies, communal stairs and unsecured windows21 than homes which are based on the ground floor. Households which do not have central heating or that have damp may be in a greater state of disrepair or be more likely to have unsafe electrical equipment than those that have central heating and no damp.
Arbitrary values (of 0, 1 and 2) were assigned to each of the housing features to create the housing score, thus giving equal weighting to each aspect of housing. This score ranged from 0 (no negative housing features) to 12 (households reporting negative features for all six housing characteristics), and an arbitrary cut-off of 4 or more was used to define ‘poor housing’ in the descriptive analysis. In the regression analyses, the score was collapsed into 0, 1, 2–3, 4–5, 6+ in order to maximise power. Sensitivity analyses were conducted exploring the score as a continuous variable (from 0 to 12), as a binary variable (poor vs not poor), and this made no difference to the findings. There is a lack of research exploring these individual aspects of housing in relation to injury risk or whether there is a cumulative level of exposure at which a threshold may exist. The analyses were repeated using the individual variables rather than scores, and the results were not altered (data not shown). There was limited variation in the housing quality indicators, and this may explain why the housing score had a limited impact on the social distribution of injury. However, if this is the case (and assuming the MCS is representative of the general population), then the implication might be that changes to housing quality are unlikely to have a significant impact on injury rates at the population level because most households have reached a threshold of reasonable housing quality. For example, it has been pointed out that smoke alarm ownership in the UK is so high that the potential for reducing inequalities through increasing ownership further is limited.22
Studies investigating the validity of parental report of home safety practices have found it to be generally reliable,23–25 particularly if interviews were carried out in the home23 or if exploring relative differences between groups rather than identifying individual need or risk.26 Mothers were asked about a limited number of safety equipment items (not, eg, window and cupboard locks which would also be relevant); additionally, it was not known whether the equipment was being used correctly.
A further limitation is that it was not always possible to ascertain from the data whether certain items of equipment were relevant or necessary to the home, and this may have diluted any effects. For this reason, a safety equipment score was compiled to act as a proxy for parental safety behaviours. All types of injuries were explored together in order to maximise power, and only certain types of injury can be prevented by different items of safety equipment. A recent randomised controlled trial in the USA estimated that just 28.4% of medically attended injuries in very young children were modifiable by safety devices.27 The lack of specificity even in the sensitivity analyses conducted to address these limitations (table 4) might explain the persistence of the null effects (eg, burns in children are commonly caused by exposure to hot water or hair straighteners, which cannot be prevented with fire guards). Unexpectedly, children living in households with none of the items of safety equipment reported in the MCS were less likely to be injured than those with all four. A number of potential factors were explored which might explain this association (eg, SECs), although none seemed to attenuate the association (data not shown). Households that owned no safety equipment may be different from those with all four items of the reported equipment in ways that we were not able to capture using the data available. A final limitation of this study is the potential for multicolinearity. It is possible that the measures of the home environment will be, to some extent, capturing the same characteristics of individuals or households as the measures of SECs.
Comparison with the literature
The social gradients in childhood injuries demonstrated here have been widely observed elsewhere, and it has been hypothesised that the home environment may explain these inequalities.5 28 However, to our knowledge, this is the first individual-level study to have investigated whether the home environment mediates the association between socioeconomic background and injuries occurring to preschool children in the home. While two studies exploring predictors of injuries in preschool children (occurring anywhere) investigated measures of socioeconomic background and the home environment,3 4 neither specifically sought to investigate the home environment as a mediator between SECs and injury.
It has been shown that indicators of housing quality are related to injury risk. For example, a study in adults living in Wales found that build type was related to injury rates, with those living in flats and terraces being at higher risk and those living in detached houses being at lower risk (compared with semidetached).29 Findings from the Avon Longitudinal Study of Parents and Children3 indicated that preschool children who lived in a flat or bedsit, who did not have access to a garden, who had moved since the last sweep or who lived in houses with a higher number of adults were at an increased risk of injury before adjustment for SECs and other child and mother characteristics. Findings from the univariate analysis presented here also demonstrate that children who were living in households with negative housing indicators (when combined as a score) were at an increased risk of injury.
A number of studies and reviews have explored the association between safety equipment use and unintentional injuries. The MCS analysis reported in this paper did not detect an effect, although findings from other observational studies in preschool children are mixed. One study set in Avon, England, reported no association between safety equipment use and injuries after controlling for other risk factors,3 while two others in Nottingham and Scotland demonstrated a beneficial impact of safety equipment use.4 30 A recent review of trials and before-and-after studies assessed the impact of safety equipment distribution and accompanying educational programmes and risk assessments on childhood injury.31 The authors found that there is only weak evidence to suggest a positive impact of these interventions on childhood injury, and this is in agreement with findings from an earlier systematic review.32 However, a recent randomised controlled trial, carried out in Ohio, investigated the differences in exposure to hazards and modifiable medically attended injuries in early childhood (mean age 6.3 months) according to an intervention comprising a number of items of safety equipment. Children living in homes that received the intervention experienced a significant reduction in exposure to hazards and injuries.27
Implications for further research, policy and practice
The measures of housing quality used in this analysis are the best available for exploring potential mediation at the individual level between SECs and injuries in young children. However, future research is required in order to construct and validate scores using these common types of measures. Data linkage between routine data containing more detailed housing information, health records and/or survey data would enable more rigorous analyses in the future. Findings reported in this paper imply that the use of the four items of safety equipment asked about in the MCS does not significantly reduce the risk of household injuries in young children, or their social gradient, at a population level. However, this is not to say that, at the individual level, specific items of safety equipment (if used correctly) will not have benefits for certain types of injury in certain households. A good quality home environment should help to prevent injuries occurring in the home. However, findings from this study imply that improvements to housing quality (based on a proxy score) are unlikely to be sufficient in reducing inequalities in injuries at the population level. Nevertheless, there are many other potential gains from further improvements to housing quality, for example, in other aspects of health, such as asthma and mental health,12 33 educational attainment,34 and crime reduction,34 which could benefit young children and other household members.
Parental factors such as supervision,35 risk taking behaviours,36 37 ability to match children's capabilities to tasks, wanting to foster independence in their children, ability to recognise hazards,38 maternal depression and social support39 are associated with childhood injuries and may also be socially distributed. Further research into these areas (and how they might interact with the home environment) could help establish why inequalities in injuries persist and contribute to the design or adaptation of policies to reduce them.
What is already known on this subject
Unintentional injuries are socially distributed, and most injuries to very young children occur in the home.
Less affluent families are more likely to live in poorer housing conditions.
The home environment may therefore lie on the causal pathway between socioeconomic status and injuries occurring in the home, and there is some ecological evidence to support this.
What this study adds
Using longitudinal data from a contemporary and representative cohort of UK children, this individual-level study found that proxy indicators for the home environment do not appear to mediate the association between socioeconomic circumstances and unintentional injuries occurring in the home.
Further research is required to establish whether this is the case in other populations and also to investigate alternative explanations for inequalities in childhood injury.
We would like to thank all the Millennium Cohort families for their participation and the director of the Millennium Cohort Study 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 late Richard Jenkins and members of the Millennium Cohort Study Child Health Group: Carol Dezateux, Lucy Griffiths, Tim J Cole, Helen Bedford, Carly Rich, Phillippa Cumberland and Jane Ahn, all of the Centre for Paediatric Epidemiology and Biostatistics, UCL Institute of Child Health, London, UK.
Funding This work was undertaken as part of the Public Health Research Consortium. The Public Health Research Consortium is funded by the Department of Health Policy Research Programme. The views expressed in the publication are those of the authors and not necessarily those of the Department of Health. Information about the wider programme of the PHRC is available from http://www.york.ac.uk/phrc. LL is funded by a Medical Research Council Career Development Award in Biostatistics. 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 Professor Health Joshi, Director of the study from the Economic and Social Research Council 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 declared.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.