Examining potential risk factors, pathways and processes associated with childhood injury in the Longitudinal Study of Australian Children1
Tamara Blakemore
Research and Analysis Branch
Australian Government Department of Families, Community Services and Indigenous Affairs
- 1. Introduction
- 2. Defining childhood injury
- 3. Conceptualising child injury causation
- 4. Contextual factors
- 5. Family factors
- 6. Child factors
- 7. Method
- 8. Results
- 9. Investigating relationships between risk factors
- 10. Discussion
- 11. Conclusion
- Endnotes
- References
1. Introduction
In Australia and in most industrial countries the world over, injury, including poisoning, is the most common cause of child death (Al-Yaman, Bryant & Sargeant 2002; AIHW 2005; Rodriguez 1990; Morrongiello et al. 2004). Among the world's 26 richest nations, injury accounts for 40 per cent of child deaths for children aged between 1–14 years (UNICEF 2000). In 2002, injuries accounted for the deaths of 229 Australian children aged between 1–14 years, 'representing 37 per cent of all deaths for this age group' (ABS 2004, p. 6). Apart from its potentially fatal consequences, childhood injury can also result in significant illness and impairment, with 65,651 children hospitalised for injury in Australia in 2002–03 (AIHW 2005; Hango & Houseknecht 2005).
Given the incidence of childhood injury and its potentially harmful consequences, the task of identifying risk factors for injury is an important prerequisite to forming effective preventative policy and practice parameters. Evidence from existing studies indicates that three domains of experience and characteristics are key influences upon childhood injury (Soubhi, Raina & Kohen 2001). These domains include factors specific to the child, their family, and their broader contextual environment. Because children's lives are shaped by their family environment, which in turn is influenced by the broader contextual environment, it is reasonable to assume that characteristics of each domain are intertwined, influencing injury to varying degrees (Ramsay et al. 2003). Information about the relationships between risk factors and the direct and indirect pathways through which risk is transmitted is, however, limited.
Acknowledging the role of chance in the occurrence of childhood injury, this paper aims to address this gap in the literature by considering the potential of a wide range of child, family and contextual characteristics to act as risk factors for injuries sustained by children in the 'child' cohort of 'Growing up in Australia', the Longitudinal Study of Australian Children (LSAC). Further, the paper aims to explore the interrelationships that may exist between significant risk factors and to examine the possible pathways and processes through which injury risk may occur.
The multifactorial, integrated working model of child injury proposed by Peterson and Brown (1994) is used to guide the investigations conducted. The following sections will review fundamental issues of injury definition and conceptualisation as well as evidence that links characteristics and experiences of children, their family and the broader contextual environment to childhood injury.
2. Defining childhood injury
By definition, childhood injuries result in physical harm or damage to the child's body (KIDSAFE 2004). The harm caused by injury may be minor or may result in significant illness, disability or even death. The cause of childhood injuries may be intentional or unintentional. Clearly, important differences exist between the risk factors and pathways implicated in the occurrence of some forms of intentional and unintentional injuries, child sexual abuse related injuries being a particular case in point. However, whether the distinction between the two types of injury is necessary, useful or practical in both research and prevention contexts relating to physical injuries to young children are debatable (Peterson & Brown 1994). In practice, differentiating some forms of unintentional injury from intentional injury may be very difficult. Injuries not inflicted upon the child may occur as the result of neglect and may therefore not be purely unintentional. Perhaps as a result, the risk factors identified for intentional and unintentional physical injuries to children are numerous, wide-ranging and often remarkably similar (Peterson & Brown 1994). Given these similarities, it may be better to conceptualise injury of either kind as a single entity, indicated by multiple risk factors and potentially reduced by broad-based prevention efforts (Peterson & Brown 1994; Wilson et al. 1991).
3. Conceptualising child injury causation
Historically, within the unintentional injury field, efforts to understand why childhood injuries occur initially concentrated on identifying aspects of the environment that posed a risk for child injury. The resulting environmental risk factor model of child injury has been highly effective in informing the development and adoption of safety standards to separate children from injury hazards. Subsequent work mirrored the direction of research in the child abuse and neglect field and concentrated upon the role human factors play in the occurrence of child injury. However, whereas child abuse and neglect models of injury causation focused on identifying, in particular, parent or family risk factors, models developed within the unintentional injury field concentrated on child-specific risk factors for injury. Identified child-specific risk factors have been conceptualised within a model of 'accident proneness'. This model has met with considerable criticism. Opponents argue that implying that an intrinsic trait is responsible for injury occurrence is not constructive to the injury prevention effort, as it diverts attention away from modifiable aspects of the environment (Klein 1980; Matheny 1987).
The capacity of environmental, parent and family or child-specific risk factor models to explain the occurrence of injury is limited because, on their own, each model relies heavily upon the isolation of single-issue causes. These models fail to take into account the reality of human experience where, rather than occurring in a vacuum, childhood injuries occur in the context of exchanges and interactions between the child and their environment. The integrated working model of child injury proposed by Peterson and Brown (1994) attends to this reality, and is adopted to guide the research conducted for this paper. Informed by ecological theory (Bronfenbrenner 1979) and focused on physical injury, this model argues that most childhood injuries (intentional and unintentional) occur due to the influence of multiple risk factors across domains of experience. Within the model, risk factors that affect global, ongoing and pervasive influences upon the child's life are classed as background contributors to child injury, whereas factors that act as specific triggers for child injury are termed immediate contributors (Peterson & Brown 1994).
Figure 1 below illustrates the conceptual framework developed for this paper. The figure combines background and immediate contributors to injury occurrence and provides details of the variables chosen to represent contextual, family and child factors. The arrows depicted in the figure identify that while contextual factors may influence family and child factors, to some extent this path of influence may be bi-directional, with aspects of the child influencing the family, or the family influencing the environment. This highlights the nested relationship between child, family and contextual factors and their potential influence upon injury occurrence.
Figure 1: conceptual framework
As observed by Peterson and Brown (1994), the factors chosen to represent each domain, and the designation of factors as background versus immediate influences, will vary as a function of the study emphasis and the adopted theoretical base. The flexible quality of the model also inherently mirrors the real-world experiences of children and families. Because of the interrelationships between children's lives, their family environment and the broader contextual environment in which they exist, it is reasonable to assume that characteristics of each will be intertwined, influencing injury to varying degrees (Ramsay et al. 2003). The following sections will briefly review evidence that links contextual, family and child factors to childhood injury.
4. Contextual factors
Contextual factors describe the context within and through which childhood injury occurs. These factors are specific to the child's immediate home environment and their broader community and society. While physical hazards in the child's environment constituted the initial focus of injury prevention research, much less attention has been given to investigating the influence of the broader contextual domain upon injury occurrence (Reading et al. 1999).
Studies that have assessed the relative influence of contextual factors upon childhood injuries have mainly done so in the context of epidemiological investigations based on area-level data (for example, Alwash & McCarthy 1988a; Alwash & McCarthy 1988b; Dougherty, Pless & Wilkins 1990; Towner & Towner 2001). The most consistent finding reported by these studies is that variations in the incidence of childhood injury across areas are due to the differential experience of economic hardship and disadvantage (Jolly, Moller & Volkmer 1993; Nersesian et al. 1985; Reading et al. 1999). Children who live in low-income neighbourhoods are reported to be between two and three times more likely than children in higher-income neighbourhoods to be injured (Durkin et al. 1994; Jolly, Moller & Volkmer 1993; Nersesian et al. 1985).
Children from families experiencing economic hardship may be at an increased risk of injury because they have greater exposure to physical hazards in the home and neighbourhood (Klein 1980). In this way, characteristics of home, neighbourhood and community that are associated with, or are the result of, economic hardship may act as immediate contributors or triggers for many childhood injuries. Child pedestrian injuries for instance, are linked to living in poorer neighbourhoods, high-density housing, streets with heavy traffic flow and limited access to safe playgrounds, all characteristics of public housing estates (Burgess 1995; Rivara & Barber 1985; Roberts et al. 1994). Similarly, child burns and scalds are linked to factors associated with economic hardship, including the use of old or faulty electrical equipment, lack of smoke detectors and electrical faults indicative of poor-quality housing (Burgess 1995). Injury prevention, such as the recall of faulty products or the introduction of safety standards, may be ineffective if the experience of economic hardship means that families need to use old or second-hand products (Burgess 1995).
Social aspects of a child's broader contextual domain may also affect injury risk. Living in a noisy home characterised by a sense of disorganisation, chaos and confusion has been cited as a marker for high risk for injury to young children (Matheny 1987). Economic hardship may underscore the social climate of the home. Economic hardship may also be related to frequent shifts in residence, a further risk factor for child injury. Transience in living arrangements has been found to be related to social isolation and a lack of social support, both of which are noted to exert a significant influence on child injury (Bronfenbrenner 1979; Hecht & Hansen 2001; Peterson & Stern 1997; Wazana et al. 1997). While social isolation is likely to be associated with unsettled living conditions and economic hardship, a complex interaction may also exist between these factors and a number of parent or family factors, including age, marital status, education and parental mental health and wellbeing.
5. Family factors
The high incidence of childhood injuries within the home when children are in the care of their family signifies the importance of understanding the potential influence family factors may have upon child injury (Matheny 1988). The theoretical associations between broad contextual factors and parental or family factors are marked, and may provide some clue as to the pathways through which risk for child injury is transmitted.
Mothers of injured children are on average, observed to be younger, less educated and more likely to be single when compared with mothers of non-injured children (Beautrais, Fergusson & Shannon 1982; McCormick, Shapiro & Starfield 1981; Nersesian et al. 1985; Parker et al. 1991). These demographic characteristics may influence child injury occurrence through their association with social isolation. Young and/or single parents may be isolated from other mothers and a limited education may preclude them from knowing what poses a risk to their child's safety or what is required to prevent injury (Ramsay et al. 2003). Parents with limited education may also not know how to seek out alternative networks for support and information. Social isolation is also found to be significantly associated with other adverse outcomes for parents and families, including health and wellbeing, emotional coping and mental illness, all noted risk factors for child injury (Matheny 1987; Weissman et al. 1986).
Children of mothers who are less active and less emotionally stable have been reported to be at high risk for injury (Matheny 1987), as have children of depressed mothers. Children of depressed mothers are reported to suffer up to four times the number of injuries experienced by children of non-depressed mothers (Brown & Davidson 1978). Where one or both parents have been treated for depression, children have been found to suffer more head injuries and other health complications than children who had parents who did not suffer from depression (Weissman et al. 1986). These results may be indicative of the stress experienced by families where a parent is unwell. Several studies have also noted that stress in the family increases risk for child injury (Beautrais, Fergusson & Shannon 1982). Not only have injured children been observed to come from families characterised by numerous stressful life experiences (Horowitz et al. 1988), but also times of great stress within the family have been identified as key points at which injuries are most likely to occur (Pearn & Nixon 1977).
Increased rates of childhood injury in families characterised by ill health may also reflect the influence such stressors have upon parenting capacity and particularly the ability to accurately assess risk and provide adequate supervision. Research findings indicate that parents are not always accurate in their assessment of their child's abilities and understanding of safety issues (Klein 1980). The extent to which parents overestimate their child's competency and awareness of risk varies as a function of parenting self-efficacy, with poorer parenting self-efficacy being associated with higher occurrence of child injury (Glik, Kronenfeld & Jackson 1993). Radke-Yarrow et al. (1985), suggest that depression in particular may influence a parent's ability to moderate their child's behaviour, which may have serious consequences for the provision of parental supervision. Adequate supervision enables the parent to intervene between the child and the physical environment to prevent exposure to hazards and to ensure safety behaviours are learned (Peterson & Stern 1997). Interactions between parents and children may operate in a two-way fashion; however, and as such, parenting behaviours may also be influenced by factors specific to the child (Fox, Kimmerly & Schaffer 1991).
6. Child factors
One of the most common findings in the injury literature is that boys experience more frequent and more severe injuries than girls (Morrongiello et al. 2004). This systematic variation in the injury population first emerges around two years of age and persists across the life-course (Baker, O'Neill, Ginsberg & Li 1992; Rivara et al. 1982). Gender differences in injury occurrence are most commonly explained by differences between boys and girls' behavioural patterns and perception of risk. Both parental reports and experimental studies observe that boys are more disruptive and engage in more active behaviours when compared to girls (Bijur, Stewart-Brown & Butler 1986; Morrongiello, Ondejko & Littlejohn 2004).
Boys also exhibit more aggressive and hyperactive behaviour than girls (Bijur, Stewart-Brown & Butler 1986). While the relative influence of aggressive versus hyperactive behaviour upon injury occurrence is debated (Davidson 1987), there is general agreement that these two behavioural traits increase the risk of injury due to their association with increased risk-taking behaviour and impulsiveness (Bijur, Stewart-Brown & Butler 1986). When children develop behavioural patterns such as these they tend to respond to stimuli in their environment in a highly energetic and rapid fashion, often without stopping to think before they act (Rothbart, Ahadi & Hershey 1994). These children may be at a greater risk of injury because the speed at which they act limits their capacity to anticipate future adverse consequences or perceive immediate danger. Boys may also incur more injuries than girls, because as well as being more likely to engage in active, impulsive and risk-taking behaviour, they also tend to underestimate injury-related risks and are more likely to attribute injuries to bad luck rather than to their own actions (Alexander et al. 1995; Morrongeillo 1997).
Children's behavioural traits, while varying as a function of gender and developmental stage, are also highly correlated with family and contextual factors that are independently associated with injury, including maternal health and wellbeing, family experience of economic hardship and dysfunction, and confusion and chaos in the home (Bijur, Stewart-Brown & Butler 1986; Campbell et al. 1991). Results from a birth cohort study of around 10,000 children indicate that children from low-income families who lived in crowded or poor-quality housing, who moved frequently and whose mothers were distressed and unhappy were more likely to be hyperactive and aggressive and were also injured at a greater rate (Bijur, Stewart-Brown & Butler 1986). Further evidence of the influence of environmental factors is provided by findings that show that confusion and noise in the home and the absence of regular sleeping and eating patterns and routines are associated with an increased injury risk for young boys (Campbell et al. 1991). Review of the empirical and theoretical origins of the child injury field reveals numerous correlates of injury across multiple domains of experience. To begin to unravel the complex relationships among and between these variables, this paper examines the injury experiences of children in the four year old cohort of LSAC.
7. Method
LSAC is a national prospective longitudinal survey designed to measure child wellbeing, health and development. The Australian Institute of Family Studies (AIFS), on behalf of the Australian Government Department of Families, Community Services and Indigenous Affairs (FaCSIA), conducted the first wave of the study in 2004. LSAC is a random probability sample of Australian residential households with children born within two specified periods, forming an infant cohort (children aged 3 months to 1 year 7 months) and a child cohort (children aged 4 years 3 months to 5 years 7 months). Information was obtained from the person most knowledgeable about the child. This informant (Parent 1) provided basic demographic information about all household members, demographic and socioeconomic information about her/himself and her/his spouse, and extensive information about the selected child. Data were collected using a variety of measures, including face-to-face interview, self-complete questionnaire, direct assessment, and observational measures. At data release 1.1 for Wave 1, data for 4,976 children and for 5,104 infants were available for analysis.
Variables and measures
Variables were selected for use in this paper on the basis of their relevance to the study of child injury as identified in the preceding literature review. Due consideration was given to the psychometric properties of scale scores and new variables were created where necessary. The majority of variables selected for use in this paper were drawn from primary caregiver report measures as these represented the most complete data.
Outcome variable
Child injury was assessed by Parent 1's answer to the question: 'During the last 12 months how many times was child hurt, injured, or had an accident and needed medical attention from a doctor or hospital?' From these data two variables were derived, one identifying injured versus non-injured children and the second identifying number of times injured. Information on the types of injuries sustained and whether the child required hospitalisation as a result of their injury was also collected.
Context variables
A total of 20 background and immediate variables were examined to characterise the broader contextual domains of injured and non-injured children. These variables included indicators specific to the home environment, housing tenure and security, social interaction and engagement, and economic hardship. Variables specific to the home environment included state and regional location, noise, clutter and traffic around the home, the number of people in the home and its general condition as well as the condition of nearby buildings. Characteristics of the home environment, such as noise, clutter and the exterior condition of the dwelling, were assessed by interviewer observation, whereas the child's parent reported on other indicators specific to the home environment. Housing tenure and stability were assessed through home ownership data, details of most recent move and number of homes the child has lived in since birth. A further variable was derived indicating whether the child or their family lived in public housing. Variables measuring neighbourhood characteristics and engagement included parental perception of the neighbourhood as a place to raise children, report of neighbourhood facilities, liveability and belonging. Mean scale scores were used for measures of neighbourhood liveability, belonging and facilities to compensate for missing data. The experience of economic hardship was measured by the total number of items parents endorsed on the economic hardship scale used within the questionnaire. Further variable specification information may be obtained from the online LSAC data dictionary at <http://www.aifs.gov.au/growingup/home.html>.
Family variables
A large number of theoretically relevant variables were selected to characterise children's family domains. Variables selected constituted both background and immediate family influences upon injury occurrence. Variables included: socio-demographic indicators of the family, such as parental age, education and marital status; behavioural indicators, such as report of parenting beliefs, behaviours, skills and difficulties; and indicators of parental health and wellbeing, such as Body Mass Index (BMI), report of depressive symptoms (and post-natal depression), relationship quality, stressful life events, social support, medical conditions, sleep problems and alcohol use. The BMI score reported by parents was included as both a marker of parental wellbeing and also in light of evidence that children whose parents are less active are more likely to be injured (Matheny 1987). Higher BMI scores may also be indicative of general risk in the parent and family environment. The number of stressful life events parents had experienced in the past 12 months was measured using an adaptation of established life events scales. Details of this measure are outlined on the online LSAC data dictionary.2 The number of psychological and psychosocial symptoms parents experienced was assessed using the K6 psychometric measure (Kessler et al. 2003). Only cases with complete data for each depressive scale item were used in analysis.
Child variables
Child-specific variables were selected to characterise the sample of injured and non-injured children, including gender and age, as well as indicators of children's health and behaviour and their social and emotional wellbeing. Indicators of children's health included the presence of sleep problems, medical conditions, experience of Attention Deficit Disorder (ADD) or Attention Deficit Hyperactivity Disorder (ADHD), premature birth and parental report of child's general health. Behavioural patterns were assessed using scores from the prosociality, hyperactivity, persistence, peer approval, conduct and reactivity scales of the 'Strengths and Difficulties Questionnaire' (SDQ) (Goodman 1999). Higher mean scores on the measure of prosocial behaviour indicate positive adjustment, whereas higher scores on the hyperactivity, persistence, peer approval, conduct and reactivity scales give a measure of risk for emotional and behavioural problems. Assessment of parental report of concern over their child's emotional health and/or behavioural wellbeing was also included to characterise children in the sample.
Data analysis procedures
Two main phases of data analysis were conducted: identification of significant risk factors for child injury and examination of the relationships between and among significant risk factors and child injury. The first objective of the analyses was to identify significant risk factors for child injury. This phase of the analyses included comparisons of contextual, family and child factors for injured and non-injured children, using Pearson Chi-Square analyses for binary and ordinal factors, and analysis of variance (ANOVA) procedures for continuous factors. Logistic regression analyses were then used to further assess the significant and net effects of considerably different variables on child injury. The dependent variable for these analyses was the binary form of the injury variable expressed as 'not injured' versus 'injured'. Separate regression models were fitted for contextual, family and child factors, where all variables were entered simultaneously, and models were compared against a constant-only model.
The second objective of the analysis was to examine relationships among and between significant risk factors and child injury. This phase of the analysis included testing two potential models of relationships between risk factors and child injury; the interaction effect model and the third variable effect model. The interaction effects model assessed the potential for significant risk factors to modify the relationship between other potential risk factors and child injury (Baron & Kenny 1986). Logistic regression was used throughout the analyses with childhood injury as the dependent variable. Again the binary form of the injury variable was used. Once the main effects of selected variables were examined, two-way interaction terms between the most significant factors and other terms in the model were entered into the model and tested. Separate regression models were initially examined for each domain. After the three domains of factors were examined, those variables within each domain that made a significant contribution (main effects and interaction terms) to the dependent variable-child injury-were entered into a summary, integrated, regression model.
Third variable analyses were used to examine the direct and indirect effects of significant risk factors (Baron & Kenny 1986). To establish an indirect effect, the independent variable must be significantly related to the dependent variable, the independent variable must be significantly related to the third variable of interest and the third variable must be significantly related to the dependent variable. A series of three logistic regression analyses is performed for each investigation. Providing the conditions of third variable analysis are met, the magnitude and direction of the relationship between the independent variable and the dependent variable are inspected to assess the nature of third variable effect.
8. Results
In total, 883 children, representing almost 18 per cent (17.7 per cent) of the sample of four year olds, were reported by Parent 1 as having sustained a physical injury requiring medical attention in the year prior to interview. Children who were injured sustained between one and eight injuries, with the majority (75 per cent) reported as having been injured only once in the past year. The number of children hospitalised for at least one night as a result of their injuries totalled 77, representing almost 9 per cent (8.7 per cent) of all those injured. The frequency with which each recorded physical injury type was reported is displayed in Figure 2.
Figure 2: Frequency of physical injuries experienced
Differences between injured and non-injured children
The results of comparative analyses characterise injured children as potentially experiencing disadvantage and vulnerability across multiple aspects of their life. Comparisons of continuous and categorical contextual variables identify significant differences between the contextual domains inhabited by injured and non-injured children. As shown in Table 1, injured children and their families moved home more often than non-injured children, experienced greater economic hardship and lived in neighbourhoods that were reported as being less 'liveable' than those lived in by non-injured children.3 No significant differences were found, however, between the two groups in relation to neighbourhood belonging, facilities or crowding in the home.
| Continuous contextual variables | Injured children | Non-injured children | ||
|---|---|---|---|---|
| Mean | (Standard deviation) | Mean | (Standard deviation) | |
| Number of homes lived in since birth | 2.01 | (0.86)* | 1.93 | (0.84)* |
| Number of people in the home | 4.49 | (1.26) | 4.47 | (1.19) |
| Neighbourhood facilities | 2.00 | (0.68) | 1.98 | (0.68) |
| Neighbourhood liveability | 2.03 | (0.49)* | 1.98 | (0.48)* |
| Neighbourhood belonging | 2.36 | (0.65) | 2.35 | (0.63) |
| Number of economic hardship items endorsed | 1.13 | (1.48)* | 0.86 | (1.25)* |
As displayed in Table 2, children whose homes were observed to be in poor condition were found to be at 40 per cent greater risk for injury than children living in homes that were in fair or well-kept condition. Injured children's homes were more likely than those of non-injured children to be cluttered. Primary caregivers of injured children were also more likely to live in public housing.
| Categorical contextual variables |
Contextual indicator | Odds ratio | 95% confidence interval |
|---|---|---|---|
| Housing quality | Home badly deteriorated or in poor condition | 1.40 | 1.12, 1.75 |
| Heavy traffic | Parent 1 agrees or strongly agrees there is heavy traffic around the home |
1.22 | 1.08, 1.38 |
| Household noise | Moderate or loud noise in the home | 1.15 | 1.01, 1.31 |
| Feelings about neighbourhood |
Parent 1 rates neighbourhood as a fair, poor or very poor place to raise children |
1.21 | 1.02, 1.43 |
| Clutter in home | Home is cluttered | 1.34 | 1.11, 1.61 |
| Public housing | Child lives in public housing | 1.28 | 1.01, 1.62 |
As displayed in Table 3, the family environments of injured children were distinguished from those of non-injured children by higher parental BMI scores and greater experience of stressful life events. Parents of injured children were younger than parents of non-injured children and reported less confidence in their parenting role. No significant differences were found between parent's report of depressive symptoms as measured by the K6 depression scale and no differences were found in parental display of warmth or hostility towards their child.
| Variable | Injured children | Non-injured children | ||
|---|---|---|---|---|
| Mean | (Standard deviation) | Mean | (Standard deviation) | |
| Parent 1 Body Mass Index measure | 2.74 | (1.15)* | 2.58 | (1.09)* |
| Parent 1 age in years | 34.12 | (5.44)* | 34.89 | (5.48)* |
| Parent 1 self report of parenting self-efficacy | 3.88 | (0.90)* | 3.94 | (0.89)* |
| Parent 1 stressful life events in the past year | 1.86 | (2.04)* | 1.57 | (1.71)* |
| Parent 1 warmth towards the child | 4.46 | (0.46) | 4.44 | (0.45) |
| Parent 1 hostility towards the child | 2.20 | (0.61) | 2.17 | (0.59) |
| Parent 1 K6 depression inventory score | 9.74 | (3.03) | 9.55 | (2.82) |
Parents of injured children were also distinguished from those non-injured by their greater experience of medical conditions, including post-natal depression after the birth of the study child. Primary caregivers of injured children also reported experiencing poorer quality sleep than parents of non-injured children. No significant differences were found between alcohol use reported by the caregivers of injured and non-injured children. Consistent with findings regarding Parent 1 age and economic hardship the primary caregivers of injured children had lower education levels and were less likely to report being married.
| Variable | Parent/ family indicator | Odds ratio | 95% confidence interval |
|---|---|---|---|
| Parent 1 sleep quality | Fairly or very bad sleep quality | 1.20 | 1.06, 1.36 |
| Parent 1 education | Below year 12 education | 1.26 | 1.12, 1.43 |
| Parent 1 medical conditions | Parent 1 suffers a medical condition | 1.15 | 1.01, 1.32 |
| Parent 1 post-natal depression | Parent 1 suffered post-natal depression after birth of the study child |
1.24 | 1.04, 1.49 |
| Parent 1 alcohol use | Parent 1 alcohol use is problematic | 1.04 | 0.85, 1.28 |
| Parent 1 marital status | Parent not married | 1.26 | 1.11, 1.44 |
Injured children were slightly younger than non-injured children, displayed more hyperactive behaviour and evidenced poorer adjustment across measures of persistence, conduct and prosocial behaviour. Associated with increased rates of hyperactivity is the finding that injured children were more likely to choose active rather than inactive pastimes. No differences were found between the two groups' scores on measures of peer approval or reactivity (see Table 5).
| Variable | Injured children | Non-injured children | ||
|---|---|---|---|---|
| Mean | (Standard deviation) | Mean | (Standard deviation) | |
| Age of child in months | 46.75 | (2.51)* | 56.94 | (2.67)* |
| Child hyperactivity (SDQ scale score) | 3.86 | (2.34)* | 3.43 | (2.27)* |
| Child persistence (SDQ scale score) | 3.84 | (0.97)* | 3.94 | (0.95)* |
| Child prosociality (SDQ scale score) | 7.62 | (1.82)* | 7.76 | (1.78)* |
| Child reactivity (SDQ scale score) | 1.73 | (1.66) | 1.69 | (1.67) |
| Child peer approval (SDQ scale score) | 8.26 | (1.54) | 8.35 | (1.56) |
| Child conduct (SDQ scale score) | 7.21 | (2.13)* | 7.57 | (1.98)* |
As displayed in Table 6, boys were around 30 per cent more likely to be injured than girls and injured children were more likely than non-injured children to have ongoing medical conditions and sleep problems. Children reported as suffering from ADD or ADHD were almost twice as likely as children without diagnosed problems to be injured. Primary caregivers of injured children reported expending greater worry and concern over their child's emotional wellbeing, happiness and/or behaviour than caregivers of non-injured children.
| Variable | Child-specific indicator | Odds ratio | 95% confidence interval |
|---|---|---|---|
| Child's emotional health | Parent 1 has concerns about child's emotional health and wellbeing |
1.17 | 1.03, 1.32 |
| Child's sleep quality | Parent 1 reports child has some sleep problems | 1.16 | 1.03, 1.31 |
| Child gender | Child is male | 1.29 | 1.14, 1.45 |
| Child medical conditions | Child suffers from a medical condition | 1.26 | 1.10, 1.44 |
| Child Attention Deficit Disorder/Attention Deficit Hyperactivity Disorder |
Child suffers from Attention Deficit Disorder or Attention Deficit Hyperactivity Disorder |
1.90 | 1.29, 2.78 |
| Child's choice to spend free time |
Child chooses active pursuits over inactive pursuits in their free time |
1.28 | 1.08, 1.51 |
Risk factors for child injury
Factors identified as differing significantly between injured and non-injured children were analysed using logistic regression to assess further their significant and net effects on childhood injury. Separate logistic regression models were formed to assess which factors from each domain were important in understanding the occurrence of the dependent variable 'child injury'.
Complete data available for analysis differed between models. Data for 1,036 children were available for analysis of contextual factors, 2,509 for analysis of family factors and 2,208 for analysis of child factors. The number of cases with complete data for contextual factors was diminished by the inclusion of the 'public housing' variable. Testing revealed, however, that omission of this factor from the model significantly reduced the model's effectiveness in explaining child injury. Regression models from each domain were found to be statistically significant, indicating that contextual, family and child-specific risk factors, as a set, had some impact on the dependent variable (child injury). The proportion of the variance accounted for by risk factors from each domain was limited, with adjusted estimates ranging from around 2 to 5 per cent. Figure 3 identifies variables within each domain identified as significant risk factors for child injury.
Figure 3: Significant risk factors for child injury
Controlling for the effect of all other contextual domain variables, the factors 'economic hardship' and 'heavy traffic' were statistically significant risk factors for child injury. As a significant background contributor, the experience of economic hardship may have global, pervasive, and ongoing effects upon the child's home and family environment, whereas heavy traffic, as an immediate risk factor, may act to trigger injury events. The factors 'Parent 1 age' and 'Parent 1 BMI' were found to be significantly associated with child injury when the effect of all other family factors was controlled for. The direction of these associations differed, however. The factor 'Parent 1 age' was negatively associated with child injury, with higher parental age decreasing the odds of child injury. The factor 'Parent 1 BMI' was positively associated with child injury, with higher BMI scores associated with greater injury risk. The factor 'number of stressful life events' neared but did not reach statistical significance.
Controlling for the effect of all other child-specific variables the factors 'male gender' and 'hyperactivity' were both found to be significant risk factors for child injury. The association between child injury and a number of other child-specific factors, including 'child sleep problems', 'child prosociality', 'child conduct' and 'child's choice of activity', neared but did not reach statistical significance. Across the individual domain specific regression models, the factor most strongly associated with child injury was 'heavy traffic', with living in a street with heavy traffic increasing the odds for child injury by 50 per cent. However, when the significant variables from each domain were entered simultaneously into an integrated summary regression model, the variables 'heavy traffic' and 'economic hardship' failed to reach significance. All other significant risk factors retained their association with child injury.
9. Investigating relationships between risk factors
Interaction effects
Evidence from the child injury literature suggests that the effect of some variables indicated as risk factors for child injury are likely to interact with the effect of others. The hypothesis examined in testing interaction models is that the effect of some risk factors may be more strongly related to injury for some people or in some circumstances than for others. The model-building strategy suggested by Hosmer and Lemeshow (1989) guided the selection of covariates for the interaction models formed for each domain. Variables were included based upon their theoretical importance and their demonstrated statistical significance (p< 0.05, 95 per cent confidence intervals). Changes to the scale of some variables were done whenever necessary after verification of the assumption of linearity in the logit (Hosmer & Lemeshow 1989). Two-way interaction terms between the most significant factors and all other variables for that domain were entered into the model and tested.
Significant interaction terms from the contextual domain, when added to the regression model, did not make a significant contribution to the dependent variable: child injury. Interaction terms from the parent/family domain, however, did reliably contribute to the model; in particular the analysis revealed a significant interaction between the variables 'Parent 1 BMI' and 'Parent 1 self report of parenting self-efficacy'. The effect of 'Parent 1 BMI' upon child injury was greater when parents reported less confidence in their parenting self-efficacy. The interaction between number of stressful life events and parenting self-efficacy neared but did not reach significance. Inclusion of the interaction terms in the model meant that the variable 'number of stressful life events' reached statistical significance, but that 'Parent 1 BMI' was no longer significantly associated with the dependent variable. The variable 'Parent 1 age' maintained its significant negative association with child injury when interaction terms were included in the model.
Analysis also revealed a significant interaction between the child domain variables 'hyperactivity' and 'child sleep problems' and also 'emotional health a worry' and 'male gender'. The effect of the variable 'hyperactivity' upon child injury was greater for children whose sleep problems were reported to be problematic rather than non-problematic. The combined effect of 'emotional health a worry' and 'male gender' decreased the odds of child injury. Inclusion of these significant interaction terms meant that the variable 'child's choice of activity' reached statistical significance. The variables 'male gender' and 'hyperactivity' retained their significant effect upon child injury when the effect of all other child factors and significant interaction terms were controlled for. Variables significantly associated within child injury were then assessed for their potential to interact with factors from other domains. Interaction terms were added to a logistic regression model containing their component terms. All variables (main effects and interaction terms) found to make a significant contribution to the dependent variable (child injury) were simultaneously entered into a summary logistic regression model. As a whole the factors in this final summary model were significant predictors of child injury, accounting collectively for around 7 per cent of the variance in the dependent variable. Table 7 displays factors and interaction terms in the final summary model.
| Variables entered | Odds ratio | 95% confidence interval |
|---|---|---|
| Economic hardship | 0.971 | 0.853, 1.106 |
| Heavy traffic in the street | 0.572 | 0.316, 1.037 |
| Parent 1 age | 0.966 | 0.940, 0.993 |
| Parent 1 Body Mass Index | 0.769 | 0.552, 1.070 |
| Number of stressful life events in past year (parent) | 1.174 | 0.990, 1.394 |
| Parent report of parenting self-efficacy | 0.676 | 0.445, 1.029 |
| Child's mean hyperactivity score | 1.063 | 0.996, 1.135 |
| Male gender (child) | 1.174 | 0.878, 1.570 |
| Child chooses active ways to spend free time | 1.290 | 0.969, 1.719 |
| Child's emotional health a worry to Parent 1 | 0.941 | 0.695, 1.374 |
| Child's sleep patterns problematic | 1.520 | 1.047, 2.207 |
| Parent 1 Body Mass Index * self report of parenting self-efficacy | 1.208 | 1.051, 1.389 |
| Number of life events * parenting self-efficacy | 0.948 | 0.879, 1.023 |
| Child's emotional health a worry * heavy traffic | 0.509 | 0.281, 0.920 |
| Economic hardship * child sleep problems | 0.742 | 0.587, 0.938 |
| Child's mean hyperactivity score * heavy traffic | 1.206 | 1.064, 1.366 |
| Child chooses active pastime * male gender | 1.679 | 0.947, 2.977 |
As previously found, the variable 'Parent 1 age' was negatively associated with child injury, with older parent age associated with a decreased risk of child injury. The variable 'child sleep problems' was positively associated with child injury, with children reported as having problematic sleep being at greater risk for injury than those with non-problematic sleep patterns. The direct association between injury and several other variables, including 'heavy traffic', 'number of stressful life events', 'hyperactivity', 'child's choice of activity' and 'self report of parenting self-efficacy', neared but did not reach statistical significance. The interactive effects of the variables 'child's emotional health a worry' and 'heavy traffic' and the variables 'economic hardship' and 'child sleep problems' both reduced the odds of child injury, whereas the interactive effects of the variables 'Parent 1 BMI' and 'self report of parenting self-efficacy' and 'hyperactivity' and 'heavy traffic' both acted to increase the odds of child injury.
Third variable effects
Third variable analyses were used to examine the potential of significant risk factors influencing injury occurrence through indirect or 'third variable' effects. These effects can be thought of as influencing outcomes through different pathways, including mediation pathways, confounding pathways and pathways of suppression. Mediation pathways, by definition, imply that the independent variable causes the third variable (the mediator), which, in turn, causes the dependent variable (MacKinnon, Krull & Lockwood 2000). By contrast confounding pathways suggest that the third variable (the confounder) explains the relationship between the independent and dependent variable, but a causal relationship is not necessarily implied (MacKinnon, Krull & Lockwood 2000). Where addition of the third variable results in an increased relation between the independent variable and the dependent variable, pathways may involve suppression effects. For the purpose of these analyses, necessary continuous variables were reduced to a binary form with 0 indicating below mean scores and 1 indicating above mean scores. The significance of third variable effects were assessed using Sobel's test method (Sobel 1986).
Significant third variable effects
Analyses revealed that two background risk factors, 'economic hardship' and 'stressful life events', were involved in third variable or indirect pathways to injury. However, of these two factors the contextual variable 'economic hardship' was indicated in more pathways, with greater strength and reliability than was 'stressful life events' and is discussed in the following section.
Third variable effects involving 'economic hardship'
Including economic hardship as a third variable in the regression model significantly reduced the magnitude of the relation between child injury and a number of factors, suggesting economic hardship may influence child injury through mediation or confounding pathways (see Table 8).
| Economic hardship confounds the relationship between parent 1 education and child injury | ||
|---|---|---|
| Odds Ratio | 95% confidence interval | |
| Testing Step 1 | ||
| Outcome: Injury status (non-injured versus injured) | ||
| Predictor: Education | 1.33 | 1.15, 1.54 |
| Testing Step 2 | ||
| Third variable: Economic hardship | ||
| Predictor: Education | 2.04 | 1.82, 2.29 |
| Testing Step 3 | ||
| Outcome: Injury status (non-injured versus injured) | ||
| Third variable: Economic hardship | 1.26 | 1.09, 1.47 |
| Predictor: Education | 1.28 | 1.11, 1.49 |
| Sobel's test of significance: 3.01, p=0.003 | ||
| Economic hardship confounds the relationship between parenting self-efficacy and child injury | ||
| Testing Step 1 | ||
| Outcome: Injury status (non-injured versus injured) | ||
| Predictor: Parenting self-efficacy | 1.09 | 1.01, 1.18 |
| Testing Step 2 | ||
| Third variable: Economic hardship | ||
| Predictor: Parenting self-efficacy | 1.16 | 1.09, 1.24 |
| Testing Step 3 | ||
| Outcome: Injury status (non-injured versus injured) | ||
| Third variable: Economic hardship | 1.30 | 1.13, 1.51 |
| Predictor: Parenting self-efficacy | 1.10 | 1.01, 1.17 |
| Sobel's test of significance: 2.82, p=0.004 | ||
| Economic hardship suppresses the relationship between child sleep problems and child injury | ||
| Testing Step 1 | ||
| Outcome: Injury status (non-injured versus injured) | ||
| Predictor: Child sleep problems | 1.20 | 1.04, 1.40 |
| Testing Step 2 | ||
| Third variable: Economic hardship | ||
| Predictor: Child sleep problems | 1.37 | 1.22, 1.55 |
| Testing Step 3 | ||
| Outcome: Injury status (non-injured versus injured) | ||
| Third variable: Economic hardship | 1.30 | 1.13, 1.51 |
| Predictor: Child sleep problems | 1.18 | 1.01, 1.37 |
| Sobel's test of significance: 2.94, p=0.003 | ||
| Economic hardship suppresses the relationship between child sleep problems and child injury | ||
| Testing Step 1 | ||
| Outcome: Injury status (non-injured versus injured) | ||
| Predictor: Child sleep problems | 1.20 | 1.04, 1.40 |
| Testing Step 2 | ||
| Third variable: Economic hardship | ||
| Predictor: Child sleep problems | 1.37 | 1.22, 1.55 |
| Testing Step 3 | ||
| Outcome: Injury status (non-injured versus injured) | ||
| Third variable: Economic hardship | 1.30 | 1.13, 1.51 |
| Predictor: Child sleep problems | 1.18 | 1.01, 1.37 |
| Sobel's test of significance: 2.94, p=0.003 | ||
From Table 8, an example of a mediating pathway may be where low levels of education lead to economic hardship, which in turn is associated with injury. The magnitude of the relationship between education and child injury is reduced because economic hardship explains part or all of the relationship between education and injury.
Further, economic hardship may confound the relationship between parenting self-efficacy and child injury in the following way: parents experiencing greater economic hardship may express less confidence in their parenting self-efficacy than parents who were under less economic strain, and children from families experiencing economic hardship may also be more likely to be injured. Parenting self-efficacy and child injury are thus related through a common confounder, economic hardship. Parenting self-efficacy does not cause economic hardship, which then causes injury, but the relationship between parenting self-efficacy and child injury is reduced in magnitude because the distortion due to economic hardship is removed.
Including economic hardship in the regression model also significantly increases the magnitude of the relationship between 'child sleep problems' and child injury, suggesting that apart from being involved in confounding and mediation effects, economic hardship may also operate as suppressor variable. The increase in the magnitude of the relationship between child sleep problems and child injury may be because economic hardship explains the variability in sleep problems, or that sleep problems are more common among children from families experiencing economic hardship.
Economic hardship also significantly influences child injury through indirect effects where 'Parent 1 BMI', 'neighbourhood liveability', 'heavy traffic in the street' and 'housing quality' are entered as third variables (see Table 9). In these instances the relationship between economic hardship and child injury is explained by the causal relationship between economic hardship and factors such as housing quality.
| Parent 1 Body Mass Index mediates the relationship between economic hardship and child injury | ||
|---|---|---|
| Odds ratio | 95% confidence interval | |
| Testing Step 1 | ||
| Outcome: Injury status (non-injured versus injured) | ||
| Predictor: Economic hardship | 1.32 | 1.14, 1.53 |
| Testing Step 2 | ||
| Third variable: Parent 1 Body Mass Index | ||
| Predictor: Economic hardship | 1.47 | 1.26, 1.71 |
| Testing Step 3 | ||
| Outcome: Injury status (non-injured versus injured) | ||
| Third variable: Parent 1 Body Mass Index | 1.29 | 1.06, 1.57 |
| Predictor: Economic hardship | 1.22 | 1.06, 1.49 |
| Sobel's test of significance: 2.28, p=0.020 | ||
| Neighbourhood liveability mediates the relationship between economic hardship and child injury | ||
| Testing Step 1 | ||
| Outcome: Injury status (non-injured versus injured) | ||
| Predictor: Economic hardship | 1.32 | 1.14, 1.53 |
| Testing Step 2 | ||
| Third variable: Neighbourhood liveability | ||
| Predictor: Economic hardship | 1.82 | 1.62, 2.05 |
| Testing Step 3 | ||
| Outcome: Injury status (non-injured versus injured) | ||
| Third variable: Neighbourhood liveability | 1.17 | 1.01, 1.37 |
| Predictor: Economic hardship | 1.29 | 1.11, 1.50 |
| Sobel's test of significance: -2.54, p=0.001 | ||
| Housing quality mediates the relationship between economic hardship and child injury | ||
| Testing Step 1 | ||
| Outcome: Injury status (non-injured versus injured) | ||
| Predictor: Economic hardship | 1.32 | 1.14, 1.53 |
| Testing Step 2 | ||
| Third variable: Housing quality | ||
| Predictor: Economic hardship | 2.76 | 2.12, 3.60 |
| Testing Step 3 | ||
| Outcome: Injury status (non-injured versus injured) | ||
| Third variable: Housing quality | 1.41 | 1.05, 1.90 |
| Predictor: Economic hardship | 1.28 | 1.11, 1.50 |
| Sobel's test of significance: 2.18, p=0.020 | ||
| Heavy traffic in the street mediates the relationship between economic hardship and child injury | ||
| Testing Step 1 | ||
| Outcome: Injury status (non-injured versus injured) | ||
| Predictor: Economic hardship | 1.32 | 1.14, 1.53 |
| Testing Step 2 | ||
| Third variable: Traffic | ||
| Predictor: Economic hardship | 1.44 | 1.28, 1.63 |
| Testing Step 3 | ||
| Outcome: Injury status (non-injured versus injured) | ||
| Third variable: Traffic | 1.25 | 1.08, 1.46 |
| Predictor: Economic hardship | 1.29 | 1.12, 1.50 |
| Sobel's test of significance: 2.97, p=0.003 | ||
10. Discussion
Recognition of the prevalence and potentially fatal consequences of childhood injury have driven researchers, practitioners and policy makers alike to search for those factors associated with increased risk for child injury. While evidence suggests that factors specific to the child, their family and their broader contextual environment are key influences upon childhood injury, few studies have assessed the relationships between risk factors and the direct and indirect pathways through which risk is transmitted.
Using data from the four-year old cohort of LSAC, this paper presents an empirical application of Peterson and Brown's (1994) integrated working model of child injury. It considered the potential of a wide range of child, family and contextual characteristics to act as risk factors for child injury and for interrelationships between significant risk factors to provide some insight into the pathways and processes through which injury occurs.
Results of analyses conducted indicate that apart from their injury experience, injured children significantly differed from non-injured children across contextual, family and child-specific factors. These differences collectively characterise injured children in the study sample as potentially experiencing disadvantage and vulnerability in many aspects of their life. Injured children's homes were more likely than those of non-injured children to be cluttered, noisy and close to heavy traffic. The primary caregivers of injured children reported greater economic hardship than those of non-injured children and also described their neighbourhoods as less liveable and less desirable as a place to bring up children. Children living in homes that were in poor condition were 40 per cent greater risk for injury compared to children living in homes in better condition. Children living in public housing were 28 per cent more likely to be injured than children living in other rental accommodation. The primary caregivers of injured children were slightly younger and less educated than those of non-injured children. They were also generally less healthy than caregivers of non-injured children, being more likely to have ongoing medical conditions, poorer sleep quality, and higher BMI scores. When compared to their non-injured counterparts, injured children were more likely to choose active pastimes and display more hyperactive behaviour. Injured children were also less healthy than non-injured children, having more ongoing medical conditions and more problematic sleep patterns. Boys were at 29 per cent greater risk of injury than girls and children with ADD or ADHD were almost twice as likely as those without attention problems to be injured.
Further examination identified that within each domain of life experience, statistically significant background and immediate risk factors for childhood injury exist. These findings provide support to the Peterson and Brown (1994) integrated working model of child injury and indicate that rather than caused by any one single factor, child injury is potentially associated with multiple risk factors across contextual, family and child-specific domains. Significant contextual risk factors identified for childhood injury included 'heavy traffic' and 'economic hardship', and significant family risk factors included 'Parent 1 age' and 'Parent 1 BMI'. Significant child risk factors for injury were 'male gender' and 'hyperactivity'.
Given the nested nature of child, family and contextual domain factors, and the fact that childhood injuries invariably occur within and through the context of multiple risk factors, the manner in which risk factors influence injury may not be direct or linear. The proportion of variance accounted for by the multivariate models is noted to be small. This is likely to be because injury occurs within the context of multiple exchanges between the child and their environment and that the contribution these exchanges make to injury experience is unlikely to be easily captured. Supporting this, assessment of the relationships among and between significant risk factors and child injury reveals that both interaction and third variable models are likely pathways or processes through which injury risks are transmitted. The role of chance must also be acknowledged as exerting a significant and large influence upon injury occurrence.
Significant interaction models identified that the effect of some risk factors was more strongly related to injury for some people or in some circumstances than for others. For example, the effect of 'Parent 1 BMI' upon injury was greater when less confidence was reported in the parenting role and 'hyperactivity' posed a greater risk for child injury when the child lived in a street with heavy traffic. Interactive effects also reduced the likelihood of injury. When the effects of worrisome emotional health and heavy traffic were combined a decreased risk of child injury was observed. Similar results were found when the effects of problematic sleep and economic hardship were combined. This is likely to be due to the fact that when combined, the individual effects of these risk factors cancel each other out.
Third variable analyses revealed that the risk factors 'economic hardship' and 'stressful life events' were implicated in numerous third variable or indirect pathways to child injury. These factors influenced the occurrence of child injury through their interrelationships with other significant risk factors. Importantly, economic hardship may have had an indirect influence upon child injury because of its relationship with a number of other significant risk factors across domains of experience and characteristics. This finding is consistent with other studies that suggest economic disadvantage has moderate to strong influence upon the injuries experienced by Australian children (Jolly, Moller & Volkmer 1993) and that economic disadvantage may influence the occurrence of child injury through a combination of mechanisms and processes (Jencks & Mayer 1990; Platt & Pharaoh 1996).
The findings presented in this paper support the general conclusion that childhood injury is likely to be influenced by multiple risk factors, and that these risk factors are likely to influence the occurrence of injury through interactive and indirect pathways. These findings should, however, be considered in light of limitations related to measurement and methodology. Firstly, while it is commonly used in the injury literature, caregiver report of child injury requiring medical attention may be subject to recall bias (Morrongiello 1997; Schwebel et al. 2004). The measure may also be biased by the effect of factors such as geographical location, access to services, caregiver health and wellbeing and caregiver knowledge of injury care. Secondly, the outcome measure used was limited to the occurrence of any injury in the 12 months prior to the survey. Future research could refine the analyses conducted to take into account the types of injuries involved, as some risk indicators may be related to injuries in general, but not to specific types of injuries (Wazana et al. 1997). Finally, the research conducted did not consider the potential for factors related to Parent 2 to influence injury. It may be that Parent 2 characteristics and experiences also affect injury through both direct and indirect pathways and processes.
11. Conclusion
The findings presented in this paper support the conceptualisation of childhood injury as the result of exchanges and interactions between the child and their family and their broader contextual environment. These findings may also warrant examination of the most effective mix of programs and interventions to prevent childhood injury. An important implication of the findings presented may be that aspects of the broader contextual environment should not be considered in isolation from family and child factors. Because interrelationships across domains may explain the pathways through which injury occurs, preventative strategies that focus on risk factors from any one domain without paying attention to related risk factors from other domains may be limited in their capacity to reduce injury incidence.
Endnotes
1. This paper builds upon preliminary work completed by the author using and presented at the Australian Institute of Family Studies (AIFS) conference in February 2005. The paper was presented at the Social Policy Research Centre conference in July 2005 and an electronic copy of the paper was posted on the website thereafter. A subset of the paper was also presented at the Australian Research Alliance for Children and Youth (ARACY) conference in August 2005. The author is grateful for the assistance and feedback provided in preparation of this paper by Helen Moyle, Justine Gibbings and Emily Bell, Research Projects Section, Australian Government Department of Families, Community Services and Indigenous Affairs.
2. Details can be found at the AIFS LSAC homepage: <http://www.aifs.gov.au/growingup/home.html>
3. Higher scores on the neighbourhood liveability index represent a less liveable neighbourhood.
References
Alexander, CS, Somerfield, MR, Ensminger, ME, Kim, YJ & Johnson, KE 1995, 'Gender differences in injuries among rural youth', Injury Prevention, vol. 1, pp. 188–93.
Alwash, R & McCarthy, M 1988a, 'Accidents in the home among children under 5: ethnic differences or social disadvantage?', British Medical Journal, vol. 196, pp. 1450–53.
Alwash, R & McCarthy, M 1988b, 'Measuring severity of injuries to children from home accidents', Archives of Disease in Childhood, vol. 63, pp. 635–38.
Al-Yaman, F, Bryant, M & Sargeant, H 2002, Australia's children: their health and wellbeing, Australian Institute of Health and Welfare (AIHW) cat. no. PHE 36, AIHW, Canberra.
Australian Bureau of Statistics (ABS) 2004, Health of children, 2004, cat. no. 4829.0.55.001, ABS, Canberra.
Australian Institute of Family Studies (AIFS) 2005, LSAC online Data Dictionary, viewed at <http://www.aifs.gov.au/growingup/home.html>.
Australian Institute of Health and Welfare (AIHW) 2005, A picture of Australia's children, AIHW, Canberra.
Baker, SP, O'Neill, B, Ginsberg, M & Li, G 1992, The injury fact book, 2nd edition, Oxford University Press, New York.
Baron, RM & Kenny, DA 1986, 'The moderator-mediator variable distinction in social psychology research: conceptual, strategic and statistical considerations', Journal of Personality and Social Psychology, vol. 51, pp. 1173–82.
Beautrais, AL, Fergusson, DM & Shannon, DT 1982, 'Childhood accidents in a New Zealand birth cohort', Australian Pediatric Journal, vol. 18, pp. 238–42.
Bijur, PE, Stewart-Brown, S & Butler, N 1986, 'Child behavior and accidental injury in 11,966 preschool children', American Journal of Diseases of Children, vol. 140, pp. 487–92.
Bronfenbrenner, U 1979, The ecology of human development: experiments by nature and design, Harvard University Press, Cambridge, MA.
Brown, GW & Davidson, S 1978, 'Social class, psychiatric disorder of mother, and accidents to children', Lancet, vol. 18, pp. 378–81.
Burgess, T 1995, 'A discussion of the relationship between socio-economic factors and the prevention of childhood injuries', The Forum, vol. 3, pp. 63–8.
Campbell, SB, March, CL, Pierce, EW, Ewing, LJ & Szumowski, EK 1991, 'Hard-to-manage preschool boys: family context and the stability of externalizing behavior', Journal of Abnormal Child Psychology, vol. 19, pp. 301–18.
Davidson, LL 1987, 'Hyperactivity, antisocial behaviour, and childhood injury: a critical analysis of the literature', Development and Behavioural Pediatrics, vol. 8, pp. 335–40.
Durkin, MS, Davidson, LL, Kuhn, L, O'Connor, P & Barlow, B 1994, 'Low-income neighborhoods and the risk of severe pediatric injury: a small-area analysis in northern Manhattan', American Journal of Public Health, vol. 84, pp. 587–92.
Dougherty, G, Pless, IB & Wilkins, R 1990, 'Social class and the occurrence of traffic injuries and deaths in urban children', Canadian Journal of Public Health, vol. 81, pp. 204–09.
Fox, NA, Kimmerly, N & Schaffer, W 1991, 'Attachment to mother/attachment to father: a meta-analysis', Child Development, vol. 62, no. 3, pp. 210–15.
Glik, D, Kronenfeld, J & Jackson, K 1993, 'Sex differences in children's risk taking behaviour', Child Development, vol. 53, pp. 426–28.
Goodman, R 1999, Strengths and Difficulties Questionnaire (SDQ), viewed at <www.sdqinfo.com>.
Hango, DW & Houseknecht, SK 2005, 'Marital disruption and accidents/injuries among children', Journal of Family Issues, vol. 26, pp. 3–31.
Hecht, DB & Hansen, DJ 2001, 'The environment of child maltreatment: contextual factors and the development of psychopathology', Aggression and Violent Behavior, vol. 6, pp. 433–57.
Horowitz, SM, Morgenstern, H, DiPietro, L & Morrison, CL 1988, 'Determinants of pediatric injuries', American Journal of Diseases in Childhood, vol. 142, pp. 605–11.
Hosmer, DW & Lemeshow, S 1989, Applied Logistic Regression, John Wiley & Sons, New York.
Jencks, C & Mayer, SE 1990, 'The social consequences of growing up in a poor neighbourhood', in LE Lynn Jr & GH McGeary (eds), Inner city poverty in the United States, National Academy Press, Washington DC, pp. 111–86.
Jolly, DL, Moller, JN & Volkmer, RE 1993, 'The socioeconomic context of child injury in Australia', Journal of Paediatrics and Child Health, vol. 29, pp. 438–44.
Kessler, RC, Barker, PR, Colpe, LJ, Epstein, JF, Gfroerer, JC, Hiripi, E, Howes, MJ, Normand, SLT, Manderscheid, RW, Walters, EE & Zaslavsky, AM 2003, 'Screening for serious mental illness in the general population', Archives of General Psychiatry, vol. 60, pp. 184–89.
Kidsafe 2004, About Kidsafe, viewed at <http://www.kidsafe.com.au>.
Klein, D 1980, 'Societal influences on childhood accidents', Accident and Prevention, vol. 12, pp. 275–81.
McCormick, MC, Shapiro, S & Starfield, BH 1981, 'Injury and its correlates among 1-year-old children', American Journal of Diseases of Children, vol. 135, pp. 159–63.
MacKinnon, DP, Krull, JL & Lockwood, CM 2000, 'Equivalence of the mediation, confounding and suppression effect', Prevention Science, vol. 1, pp. 173–81.
Matheny, AP 1987, 'Psychological characteristics of childhood accidents', Journal of Social Issues, vol. 43, pp. 45–60.
Matheny, AP 1988, 'Accidental injuries', in D Routh (ed), Handbook of pediatric psychology, Guilford, New York.
Morrongiello, BA 1997, 'Children's perspectives on injury and close-call experiences: sex differences in injury-outcome processes', Journal of Pediatric Psychology, vol. 22, pp. 499–512.
Morrongiello, BA, Ondejko, L & Littlejohn, A 2004, 'Understanding toddlers in-home injuries: contexts, correlates and determinants', Journal of Pediatric Psychology, vol. 29, pp. 415–31.
Nersesian, WS, Petit, MR, Shaper, R, Lemieux, D & Naor, E 1985, 'Childhood death and poverty: a study of all childhood deaths in Maine, 1975 to 1980', Pediatrics, vol. 75, pp. 41–50.
Parker, RM, Rescorla, LA, Finkelstein, JA, Barnes, N, Holmes, JH & Stolley, PD 1991, 'A survey of the health of homeless children in Philadelphia shelters', American Journal of Disease in Children, vol. 145, pp. 520–26.
Pearn, J & Nixon, J 1977, 'Prevention of childhood drowning accidents', Medical Journal of Australia, vol. 1, pp. 616–18.
Peterson, L & Brown, D 1994, 'Integrating child injury and abuse–neglect research: common histories, etiologies and solutions', Psychological Bulletin, vol. 116, pp. 293–315.
Peterson, L & Stern, BL 1997, 'Family processes and child risk for injury', Behaviour Research and Therapy, vol. 35, no. 3, pp. 179–90.
Platt, MJ & Pharaoh, POD 1996, 'Child health statistical review, 1996', Archives of Disease in Childhood, vol. 75, pp. 527–33.
Radke-Yarrow, M, Cummings, EM, Kuczynski, L & Chapman, M 1985, 'Patterns of attachment in two-and three-year olds in normal families and families with parental depression', Child Development, vol. 56, pp. 884–95.
Ramsay, LJ, Moreton, G, Gorman, DR, Blake, E, Goh, D, Elton, RA & Beattie, TF 2003, 'Unintentional home injury in preschool-aged children: looking for the key-an exploration of the inter-relationship and relative importance of potential risk factors', Public Health, vol. 117, pp. 404–11.
Reading, R, Langford, IH, Haynes, R & Lovett, A 1999, 'Accidents to preschool children: comparing family and neighbourhood risk factors', Social Science and Medicine, vol. 48, pp. 321–30.
Rivara, FP & Barber, M 1985, 'Demographic analysis of childhood pedestrian injuries', Pediatrics, vol. 76, no. 3, pp. 375–81.
Rivara, F, Bergman, A, LoGerfo, J & Weiss, N 1982, 'Epidemiology of childhood injuries: sex differences in injury rates', American Journal of Diseases of Children, vol. 136, pp. 502–06.
Roberts, I, Norton, R, Dunn, R, Hassall, I & Lee-Joe, T 1994, 'Environmental factors and child pedestrian injuries', Australian Journal of Public Health, vol. 18, pp. 43–6.
Rodriguez, J 1990, 'Childhood injuries in the United States: a priority issue', American Journal of Social Issues, vol. 43, pp. 105–18.
Rothbart, MK, Ahadi, SA & Hershey, KL 1994, 'Temperament and social behavior in children', Merrill-Palmer Quarterly, vol. 40, no. 1, pp. 21–39.
Schwebel, DC, Brezausek, CM, Ramey, SL & Ramey, CT 2004, 'Interactions between child behavior patterns and parenting: implications for children's unintentional injury risk', Journal of Pediatric Psychology, vol. 29, pp. 93–104.
Sobel, ME 1986, 'Some new results on indirect effects and their standard errors in covariance structure models', in N Tuma (ed), Sociological methodology, American Sociological Association, Washington DC, pp. 159–86.
Soubhi, H, Raina, P & Kohen, D 2001, Effects of neighbourhood, family, and child behaviour on childhood injury in Canada, Applied Research Branch, Strategic Policy, Human Resources Development Canada.
Towner, E & Towner, J 2001, 'The prevention of childhood unintentional injury', Current Paediatrics, vol. 11, pp. 403–08.
United Nations Children's Fund (UNICEF) 2000, A league table of child deaths by injury in rich nations, UNICEF, Florence.
Wazana, A, Krueger, P, Raina, P, Chambers, L 1997, 'A review of risk factors for child pedestrian injuries: are they modifiable?', Injury Prevention, vol. 3, pp. 295–304.
Weissman, MN, John, K, Merikangas, KR, Prusoff, BA, Wickramaratne, P, Gammon, D, Angold, A & Warner, V 1986, 'Depressed parents and their children: general health, social, and psychiatric problems', American Journal of Diseases in Children, vol. 140, pp. 801–05.
Wilson, M, Baker, S, Teret, S, Shock, S & Gabarino, J 1991, Saving children: a guide to injury prevention, Oxford University Press, New York.