Household and neighbourhood risks for injury to 5–14 year old children

Abstract

Injuries in childhood are strongly related to poverty at the household level and to living in a deprived neighbourhood, but it is not clear whether these effects are independent. In this prospective population study, all injuries to 5–14 year old children living in the city of Norwich, UK, and presented at the hospital Accident and Emergency Department over a 13 month period were recorded (N=3526). Information on the population of resident children and household composition was assembled from the health authority population register. Neighbourhood information was extracted from the census and local surveys. Unadjusted risks were calculated for individual and neighbourhood factors, followed by multilevel modelling in which predictors were included at three levels: individual, enumeration district and social area (neighbourhood). The overall injury rate was 16.44 per 100 children per year. Injury rates between neighbourhoods varied two-fold and were highest in more deprived areas. In the final multilevel model injury risk was related to gender (boys vs. girls OR=1.35), age of child (OR=1.07 per year), number of adults in the household (OR=0.91 per adult), and age gap between child and eldest female (15–24 years vs. 25–34 years, OR=1.15). Injury rates were also related to social area deprivation, although variations in injury rates between neighbourhoods were not wholly explained by deprivation. The adjusted odds ratio between the most and least deprived social areas was 1.35. Excluding less serious injuries did not substantially change the results. The risks were very similar to those found in a previous study of pre-school children, with the same neighbourhoods identified as high and low risk as before. This evidence that neighbourhood factors independently influence injury risk over and above individual and household factors supports the use of area-based policies to reduce injuries in children.

Introduction

Injury is the commonest cause of death in childhood in developed countries (UNICEF, 2001). The burden of resulting disability, morbidity and health service use has been described as one of the most pressing global public health problems affecting children (Stone, Jarvis, & Pless, 2001). One consistent epidemiological feature of childhood injury is the steep social gradient seen in all types of injury, at all ages and in all the developed countries in which it has been studied. In recent years these social differences are becoming wider (Roberts & Power, 1996; Scholer, Hickson, & Ray, 1999).

Injury rates have been shown to differ according to the socio-economic circumstances of families and households (Alwash & McCarthy, 1988; Roberts, 1997; Scholer, Mitchel, & Ray, 1997; Scholer, Hickson, Mitchel, & Ray, 1998), but the bulk of research on social inequalities in childhood injury has used area of residence as the basis on which to compare rates. Injury rates are higher in children living in deprived areas compared to those living in affluent areas as determined by census data on material deprivation (Sharples, Storey, Aynsley-Green, & Eyre, 1990; Walsh & Jarvis, 1992; Laing & Logan, 1999), income (Joly, Foggin, & Pless, 1991; Jolly, Moller, & Volkmer, 1993; Durkin, Davidson, Kuhn, O’Connor, & Barlow, 1994; Faelker, Pickett, & Brison, 2000; Pomerantz, Dowd, & Buncher, 2001), and proportions of lone parent families (Braddock, Lapidus, Gregorio, Kapp, & Banco, 1991; Haynes, Lovett, Reading, Langford, & Gale, 1999).

These studies have brought the issue of social inequalities in childhood injury to widespread notice but they depend on the assumption that the area characteristics identified reflect the household characteristics associated with increased risk of injury. This can be questioned on two counts. Firstly it risks the ecological fallacy, namely that the characteristics of the households in which children are injured may not be the ones measured at the area level. This possibility has been much discussed in the literature, but is probably not relevant to childhood injury studies because of the consistency of findings between different settings and between area-based and individual household-based studies. More importantly, comparisons between areas or between families cannot distinguish the level at which these social factors operate. Is a child's risk of injury increased by coming from a poor family or living in a poor area?

The distinction is important in understanding the causes of social inequalities in injury. Neighbourhood factors might well play a role in the aetiology of injury alongside factors related to individuals and families because injuries occur in a place as well as to a person. The risks associated with deprived neighbourhoods might be environmental, such as dangerous streets, poor quality housing stock and unprotected industrial and building sites, but might also reflect social and culturally determined attitudes and practices such as those relating to supervision, independence and appropriate play activities for children (see for example Soori and Bhopal, 2002). Macintyre, Ellaway, and Cummins (2002) have described these types of influences as “collective social functioning” and argue that they are an important aspect of the contextual area based influences on health. Although these collective factors are often related to area deprivation, there is no a priori reason why they need to be.

The distinction between neighbourhood and household risk factors is also relevant to preventive policy. Many of the current initiatives for reducing child health inequalities in the UK are area-based (DfSS, 1999; DoH, 2001). For instance, Sure Start, a Government-funded scheme based on early intervention, education and family support, is designed to improve the life chances of babies and young children from disadvantaged backgrounds and has a particular target of reducing accident rates. Sure Start programmes are established in socially deprived areas, aiming to involve each local community in both planning and implementation (DfEE, 1999; Roberts, 2000). The rationale for targeting interventions on areas with high levels of deprivation is that it provides the most effective way of reaching families most in need. Critics of this approach point to the evidence that only a minority of the poorest families live in the most deprived areas (McLoone, 2001). However, if injury risk is related to neighbourhood factors, then there may be an additional specific benefit of area-based preventive interventions.

There have been relatively few studies separating out neighbourhood from household influences on injury risk. In comparison, there is considerable interest in the interaction between people and places in the cause of other health problems such as adult mortality (Ecob & Jones, 1998), limiting illness (Shouls, Congdon, & Curtis, 1996), smoking behaviour (Duncan, Jones, & Moon, 1999), child behaviour (Kalff et al., 2001), child development (Brooks-Gunn, Duncan, Klebanov, & Sealand, 1993; McCulloch & Joshi, 2001) and immunisation coverage (Jones, Moon, & Clegg, 1991). Such studies have been facilitated by the development of multilevel modelling, a statistical method for separating effects at different levels of aggregation (Duncan, Jones, & Moon, 1998).

We know of only three studies so far published that have used appropriate multilevel modelling techniques to disentangle individual (or household) effects from neighbourhood effects on the risk of injury. One was on injury mortality in a large sample of adults in the US (Cubbin, LeClere, & Smith, 2000), another on self reported injury-producing events in a small survey of families with infants and young children from Baltimore, Maryland (O’Campo, Rao, Gielen, Royalty, & Wilson, 2000) and the third was a study by ourselves from Norfolk, UK, on hospital attendances for injury in a large total population of preschool aged children (Reading, Langford, Haynes, & Lovett, 1999). In each of these studies both individual and neighbourhood socio-economic characteristics independently influenced the risk of injury. Coming from a socially disadvantaged family increased the injury risk, but so did living in a deprived neighbourhood regardless of the family circumstances. Our study showed the well-recognised associations of injury risk with individual and household factors such as family size, maternal age, lone parenthood and sex. An extra independent risk related to neighbourhood was partly explained by area deprivation, but some residual variation between neighbourhoods remained unexplained. Thus, characteristics of neighbourhoods other than deprivation also influenced injury risk.

In the current study we have extended the age group to look at older children. We have used the same geographical methods as developed previously (Haynes et al., 1999) although this study only considers the built up area in and around the city of Norwich because in the previous study we found little neighbourhood variation in the surrounding rural areas and the rural injury rates were affected by distance from the hospital. We have examined a greater range of possible neighbourhood explanatory factors than before and we have also included three levels in the model, corresponding to the household, the very localised immediate vicinity and the wider neighbourhood. Thus in all important respects the results can be compared with those found previously in the younger age group, but we believe this study has a number of methodological improvements.