1

Research Projects Section

Research and Data Management Branch

Department of Family and Community Services

“Examining Potential Risk Factors, Pathways and Processes Associated with Childhood Injury in the Longitudinal Study of Australian Children”

Tamara Blakemore

Note The views expressed in this paper are the author’s and do not necessarily reflect the opinions of the Department of Family and Community Services, the minister or of the federal government.

Acknowledgements The author is grateful to Justine Gibbings and Mark Reid for the valuable advice and assistance they provided in the preparation of this paper.

Abstract In Australia, injury including poisoning is reported as the most common cause of death from early childhood through to middle age (AIHW, 2002). Apart from its potentially fatal consequences, childhood injury can also affect children’s development via its association with disability, impairment and illness. Considerable debate exists around understandings of injury causation and the division between intentional and unintentional injury. This hampers the efforts of policy makers and practitioners to prevent injury. It is argued that the risk factors associated with physical injuries of either kind are similar, existing across multiple domains of human experience and their effect influenced by their interaction with each other. This paper builds upon the findings of preliminary work using data from first wave of Growing Up in Australia, the Longitudinal Study of Australian Children, and presents the results of work exploring potential risk factors, and the direct and indirect pathways and processes implicated in injury occurrence. The paper uses an integrated, multi-dimensional conceptual framework to guide investigation, conceptualising childhood injury as the result of exchanges and interactions between the child and their family and the broader contextual environment. Understanding the function of risk factors associated with childhood injury will aid in efforts to mitigate the risk associated with such life events and is imperative for the formation of effective injury prevention policy and practice parameters.

2

Introduction

In Australia and in most industrial countries the world over, injury including

poisoning is the most common cause of child death (AIHW, 2002; Rodriguez, 1990;

Morongiello, Ondejko & Littlejohn, 2004). Amongst the world’s 26 richest nations,

injury accounts for 40 percent (40%) of child deaths for children aged between 1 and

14 years (UNICEF, 2000). In 2002, injuries accounted for the deaths of 229

Australian children aged between 1 and 14 years, representing 37 percent (37%) of all

deaths for this age group (ABS, 2004). Apart from its potentially fatal consequences,

childhood injury can also result in significant illness and impairment with around

65650 children hospitalised for injury in Australia in the year 2002 - 2003

(AIHW,2005; Hango & Houseknect, 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

in the injury literature indicates 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 via which risk is transmitted is however

limited.

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

child injury and by examining inter-relationships between significant risk factors. The

multifactorial, integrated working model of child injury proposed by Peterson and

Brown (1994), is used to guide investigation and data from the four-year-old cohort of

‘Growing up in Australia’, the Longitudinal Study of Australian Children (LSAC)’ is

used.

3

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.

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 is 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).

Conceptualising child injury causation

Historically, within the field of ‘unintentional’ injury, 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 in the field went on to reflect the direction of research in the child abuse and

neglect field and consider the role human factors play in the occurrence of child

injury. While models of injury causation within the child abuse and neglect field

focused on parent or family risk factors, models developed within the ‘unintentional’

injury field concentrated on child-specific risk factors for injury. Identified risk

4

factors were 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 address the

reality of human experience where, rather than occur 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 (Bronfebrenner, 1979) and focussed 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 affects global, ongoing, and pervasive

influences upon the child’s life, are classed ‘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.

Figure 1

C o n c e p t u a l F r a m e w o r k

C o n t e x t u a l F a c t o r s

I n j u r y

F a m i l y F a c t o r s C h i l d F a c t o r s

B a c k g r o u n d I n f l u e n c e s

B a c k g r o u n d I n f l u e n c e s

B a c k g r o u n d I n f l u e n c e s

I m m e d i a t eI n f l u e n c e s

I m m e d i a t eI n f l u e n c e s

I m m e d i a t eI n f l u e n c e s

5

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 adopted theoretical base. The

flexible quality of the model also inherently mirrors the real-world experiences of

children and families. Because of the inter-relationships 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.

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 (e.g., Alwash & McCarthy, 1988; 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; Neresian, 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; Neresian, 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 the result of economic hardship may act as

immediate contributors or triggers for many childhood injuries. Child pedestrian

6

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). Preventative efforts aimed at reducing injury whereby faulty products are

recalled or safety standards introduced may be ineffective if the experience of

economic hardship means that families 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 disorganization, chaos, and

confusion has been cited as a marker for high risk for injury to young children

(Matheny, 1987). The social climate of the home may be underscored by economic

hardship, as may 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, 1986; Hecht & Hanson, 2001; Peterson & Stern, 1997;

Wazana, Krueger, Raina et al., 1997). While social isolation is likely to be associated

with the experience of economic hardship and unsettled living conditions, 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 well-

being.

Family Factors

The high incidence of childhood injuries within the home and whilst 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 via 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

7

(Beautrais et al., 1982; McCormick, Shapiro, & Starfield, 1981; Nersesian et al.,

1985; Parker et al., 1991). As stated, these demographic characteristics may influence

child injury occurrence via 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 well-being, 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 noted

to be at high risk for child 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 et al., 1982). Not only have injured children been observed

to come from families characterised by numerous stressful life experiences (Horowitz,

1988), times of great stress within the family have also 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 skill, with poorer parenting skills being associated with higher occurrence

of child injury (Glik, Kronefeld & Jackson, 1993). Radke-Yarrow et al (1985),

suggest that depression in particular, may influence a parent’s ability to moderate their

8

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 however, operate in a two-way fashion and as such, parenting

behaviours may also be influenced by factors specific to the child (Fox, Kimmerly, &

Schaffer, 1991).

Child factors

The role child factors play in influencing injury occurrence remains a contentious

issue. While injured children were previously thought to possess a trait that made

them ‘accident-prone’ it is now considered that child factors probably influence injury

occurrence via their association with broader family and contextual factors.

One of the most common findings in the injury literature is that boys experience more

frequent and more severe injuries than girls (Morrongiello, Ondejko & Littlejohn,

2004). This systematic variation in the injury population first emerges around 2 years

of age and persists across the life-course (Baker, O’Neill, & Ginsberg, 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, et al., 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 behaviour 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

9

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 that do occur to bad luck rather than to their own actions (Alexander, et al.,

1995; Morongeillo, 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 well-being,

family experience of economic hardship and dysfunction, confusion and chaos in the

home (Bijur, Stewart-Brown, & Butler, 1986; Campbell, et al., 1991). Results from a

birth cohort study of 10, 394 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 which 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.

Method

LSAC is a national prospective longitudinal survey designed to measure child well-

being, health, and development. The Australian Institute of Family Studies (AIFS), on

behalf of the Commonwealth Department of Family and Community Services

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 a infant cohort (children aged 3 months – 1 yr 7 months) and a child cohort

(children aged 4yrs 3 months and 5 years 7 months). Information was obtained from

the person most knowledgeable about the child. This informant (Parent 1) provided

10

basic demographic information about all household members, socio-economic

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.

In Wave 1, complete data were collected for 5104 infants and 4976 children.

Variables and measures Variables were selected for use in this paper on the basis of their theoretical relevance

to the study of child injury. 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 this 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 twenty background and immediate variables were examined to characterise

the broader contextual domains of injured and non-injured children; these variables

(displayed in Figure 2 below) included indicators of home environment, housing

quality, tenure and stability, economic hardship and social interaction with the broader

environment. Mean scale scores are used for measures of ‘neighbourhood liveability’,

‘neighbourhood belonging’ and ‘neighbourhood facilities’ to compensate for missing

data. The total number of items endorsed on the ‘economic hardship’ scale is used as

an indicator of economic hardship. Interviewer observation was used to assess

characteristics of the home environment including clutter, noise, and the condition of

the exterior of the home. From information provided on housing tenure a further

variable was derived indicating whether the child or their family lived in public

housing. Further variable specification information may be obtained from the online

LSAC data dictionary (http://www.aifs.gov.au/growingup/home.html).

11

Figure 2: Contextual factors examined

• H o u s i n g q u a l i t y• T r a f f i c • N o i s e i n t h e h o m e• C l u t t e r i n t h e h o m e• P a r e n t 1 p e r c e p t i o n o f t h e n e i g h b o u r h o o d a s a p l a c e

t o r a i s e c h i l d r e n . • N u m b e r o f h o m e s c h i l d h a s l i v e d i n s i n c e b i r t h• P a r e n t 1 r e p o r t o f n e i g h b o u r h o o d l i v e a b i l i t y• P a r e n t 1 r e p o r t o f n e i g h b o u r h o o d f a c i l i t i e s• P a r e n t 1 r e p o r t o f n e i g h b o u r h o o d b e l o n g i n g• G e n e r a l c o n d i t i o n o f b u i l d i n g s n e a r b y• D e t a i l s o f m o s t r e c e n t m o v e• D e t a i l s o f h o m e o w n e r s h i p• S t a t e a n d r e g i o n a l l o c a t i o n i n w h i c h c h i l d l i v e s• N u m b e r o f p e o p l e i n t h e h o m e• E c o n o m i c h a r d s h i p

C o n t e x t u a l F a c t o r s

B a c k g r o u n d I n f l u e n c e s

I m m e d i a t eI n f l u e n c e s

Family variables A large number of theoretically relevant variables were selected to characterise the

child’s family domain. Both background and immediate variables were selected

including socio-demographic indicators, behavioural indicators and indicators of

reported health and wellbeing. These variables are displayed in Figure 3 below.

Responses to the scale measures ‘number of stressful life events’ and ‘Parent 1 K6

depression score’ were independently summed to form variables representing ‘number

of stressful life events reported’, and ‘number of depressive symptoms endorsed’.

Only cases with complete data for each depressive scale items were used.

Figure 3: Family factors examined

F a m i l y F a c t o r s

B a c k g r o u n d I n f l u e n c e s

I m m e d i a t eI n f l u e n c e s

• P a r e n t 1 a g e• P a r e n t 1 B M I• P a r e n t 1 e d u c a t i o n• P a r e n t 1 l e g a l / r e g i s t e r e d m a r i t a l s t a t u s• P a r e n t 1 r e p o r t o f p o s t n a t a l d e p r e s s i o n P a r e n t 1

D e p r e s s io n i n v e n t o r y s c o r e• P a r e n t 1 n u m b e r o f s t r e s s f u l l i f e e v e n t s• P a r e n t 1 p a r e n t i n g b e h a v io u r s • P a r e n t 1 r e l a t i o n s h i p q u a l i t y• P a r e n t 1 s e l f r e p o r t o f p a r e n t i n g s k i l l s• P a r e n t 1 r e p o r t o f b e l i e f s a b o u t s u p e r v i s i o n• P a r e n t 1 p r o b l e m s m a n a g i n g s t u d y c h i l d• P a r e n t 1 s o c i a l s u p p o r t• P a r e n t 1 m e d i c a l c o n d i t i o n s• P a r e n t 1 s l e e p p r o b l e m s• P a r e n t 1 a l c o h o l u s e

12

Child variables Child-specific variables selected to characterise the sample of injured and non-injured

children included gender and age and indicators of children’s health and behaviour

and their social and emotional well-being. Behavioural patterns were assessed using

scales from the “Strengths and Difficulties Questionnaire”. Higher mean scores on the

measure of prosocial behaviour indicate positive adjustment whereas scores on

hyperactivity, and persistence, conduct and peer approval scales give a measure of

risk for emotional and behavioural problems. Child-specific factors examined are

displayed in Figure 4 below.

Figure 4: Child-specific factors examined

C h i l d F a c t o r s

B a c k g r o u n d I n f l u e n c e s

I m m e d i a t eI n f l u e n c e s

• C h i l d g e n d e r• C h i l d s l e e p p r o b l e m s• C h i l d m e d i c a l c o n d i t i o n s• C h i l d s u f f e r s f r o m A D D / A D H D• C h i l d ’ s c h o i c e o f a c t i v i t y• C h i l d a g e i n m o n t h s• C h i l d p r o s o c i a l i t y s c a l e s c o r e• C h i l d h y p e r a c t i v i t y s c a l e s c o r e• C h i l d p e r s i s t e n c e s c a l e s c o r e• C h i l d c o n d u c t s c a l e s c o r e• C h i l d r e a c t i v e s c a l e s c o r e• I n d i c a t o r o f p r e m a t u r e b i r t h• I n d i c a t o r o f s c h o o l a t t e n d a n c e• C h i l d g e n e r a l h e a l t h• P a r e n t 1 r e p o r t o f c o n c e r n o v e r c h i l d ’ s e m o t i o n a l

h e a l t h a n d / o r b e h a v i o u r a l w e l l b e i n g

Data analysis procedures

The analyses conducted included two main phases: 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

comparison 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 assess further the significant and net effects of

significantly differing variables on child injury. Separate regression models were

fitted for contextual, family and child factors, wherein all variables were entered

simultaneously, and models were compared against a constant only model.

13

The second object 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.

With the interaction effects model, the potential for significant risk factors to modify

the relationship between other potential risk factors and child injury was assessed

(Baron & Kenny, 1986). Logistic regression was used throughout the analyses with

childhood injury as the dependent variable. 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 first formed 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 models were entered into a

summary integrated regression model.

Third variable analyses were used to sort out the direct and indirect effects of

significant risk factors using a series of logistic regression analyses (Baron & Kenny,

1986). To illustrate, to assess the indirect influence of Parent 1 economic hardship (as

an independent variable) on injury (as a dependent variable) using Parent 1 BMI as a

third variable, the following regression equations are estimated:

Equation 1: Injury = β01 + β11 economic hardship

Equation 2: Parent 1 BMI = β02 + β12 economic hardship + error 2

Equation 3: Injury = β03 + β13 Parent 1 BMI + β23 economic hardship

To establish an indirect effect, the independent variable must be significantly related

to the dependent variable in the first equation, the independent variable must be

significantly related to the third variable in the second equation and the third variable

must be significantly related to the dependent variable in the third equation. 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.

14

Results

In total, 883 children, representing almost 18 percent (17.7%) 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%) 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 percent (8.7%) of all

those injured. The frequency with which each recorded physical injury type was

reported is displayed in Chart 1 below.

Chart 1. Frequency of physical injuries experienced

0 50 100 150 200 250 300 350 400 450

Cut or scrapeOther

Broken bonesHead injury

Sprain or strainBurn or scaldDental injury

DislocationAccidental poisoningOther internal injury

Frequency (n)

Differences between injured and non-injured children

Comparative analysis characterises injured children as potentially experiencing

disadvantage and vulnerability across multiple aspects of their life. Figure 7 lists the

contextual, family and child-specific factors found to differ significantly between

injured and non-injured children. Extended data tables reporting full results are

contained in the Appendix.

Comparison of the broad contextual domains inhabited by injured and non-injured

children reveal that significant differences exist between the two groups in relation to

both background and immediate contextual factors. Injured children were more likely

to live in neighbourhoods considered by Parent 1 as less liveable and less desirable as

a place to raise children.

15

Figure 7. Significant differences between injured and non-injured children

C o n t e x t u a l F a c t o r s

• N e ig h b o u r h o o d l iv e a b i l i t y• N u m b e r o f h o m e s• E c o n o m ic h a r d s h ip• H o u s in g q u a l i t y• H e a v y t r a f f ic• N o is e in t h e h o m e• C lu t t e r in t h e h o m e• N e ig h b o u r h o o d a s a p la c e t o r a is e c h i ld r e n• P u b l ic h o u s in g t e n a n c y

F a m i ly F a c t o r s

• P a r e n t 1 B M I• P a r e n t 1 a g e• P a r e n t in g s k i l l s• L if e e v e n t s• P a r e n t 1 S le e p p r o b le m s• P a r e n t 1 e d u c a t io n• P a r e n t 1 m e d ic a l c o n d i t io n s• P a r e n t 1 p o s t n a t a l d e p r e s s io n • P a r e n t 1 m a r i t a l s t a t u s

C h i ld F a c t o r s

• A g e o f c h i ld• C h i ld h y p e r a c t iv i t y• C h i ld p e r s i s t e n c e• C h i ld p r o s o c ia l i t y• C h i ld c o n d u c t• C h i ld ’ s e m o t io n a l h e a l t h o f c o n c e r n• C h i ld s le e p in g p r o b le m s• C h i ld g e n d e r• C h i ld A D D /A D H D• C h i ld m e d ic a l c o n d i t io n s• C h i ld ’ s c h o ic e o f a c t iv i t y

No significant differences were found however between the two groups in relation to

neighbourhood belonging, facilities or crowding in the home. Primary caregivers of

injured children are more likely than the caregivers of non-injured children to report

greater experience of economic hardship, were more likely to report more changes in

residence since the birth of the study child and were more likely to live in public

housing. Interviewer observations of the home environment found no differences in

the condition of buildings close to or nearby injured and non-injured children’s

homes. However children whose own homes were in poor condition were found to be

40 percent (40%) greater risk for injury than children living in homes that were in fair

or well-kept condition. Injured children’s homes were also more likely than those of

non-injured children to be cluttered, noisy, or close to heavy traffic.

The family environments of injured children were distinguished from those of non-

injured children by greater report of primary caregiver medical conditions, including

post-natal depression after the birth of the study child. No significant differences were

found between parent’s report of depressive symptoms as measured by the K6

depression scale. Primary caregivers of injured children were found to have higher

body mass index (BMI) scores when compared to parents of non-injured children and

also reported experiencing poorer quality sleep. No significant differences were found

16

between alcohol use reported by the caregivers of injured and non-injured children.

Consistent with findings regarding economic hardship, primary caregivers of injured

children were also slighter younger than those of non-injured children, had lower

education levels and were less likely to report being married. Caregivers of injured

children also reported experiencing more stressful life events in the year prior to

interview. No significant differences were found between beliefs regarding

supervision or the warmth, consistency, or hostility of parenting behaviours between

caregivers of injured and non-injured children. Caregivers of injured children were

however less likely to report confidence in their parenting skill or ability when

compared to caregivers of non-injured children.

Injured children were slightly younger than non-injured children and boys were

around 30 percent (29%) more likely to be injured than girls. Injured children were

more likely than non-injured children to have ongoing medical conditions and sleep

problems. Injured children displayed more hyperactive behaviour and evidenced

poorer adjustment across measures of persistence, conduct and prosocial behaviour.

No differences were found between the two groups’ scores on measures of peer

approval or reactivity but 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. Associated with increased rates of

hyperactivity is the finding that injured children were more likely to choose active

rather than inactive pastimes children than non-injured children and those reported as

suffering from ADD or ADHD were almost twice as likely as children without

diagnosed problems to be injured.

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 1036 children

were available for analysis of contextual factors, 2509 for analysis of family factors

17

and 2208 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

percent. Figure 8 below identifies variables within each domain identified as

significant risk factors for child injury.

Figure 8. Significant risk factors for child injury

S i g n i f i c a n t C o n t e x t u a l

R i s k F a c t o r s

E c o n o m i c h a r d s h i p

H e a v y t r a f f i c

S i g n i f i c a n t F a m i l y

R i s k F a c t o r s

P a r e n t 1 a g e

P a r e n t 1 B M I

S i g n i f i c a n t C h i l d

R i s k F a c t o r s

M a l e g e n d e r

H y p e r a c t i v i t y

I n j u r y

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

18

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 percent (50%). 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.

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% 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

19

entered into the model and tested. Extended data tables reporting full results are

contained in the Appendix.

For the contextual domain interaction terms included the product of the background

risk factor ‘economic hardship’ and all other contextual variables and the product of

the immediate risk factor ‘heavy traffic’ and all other contextual variables. Addition

of these terms to the regression model found no significant interaction effects.

Significant interaction effects were however observed within the family and child

domains.

Interaction terms for the family domain included the product of ‘Parent 1 BMI’ and

all other family variables, the product of ‘Parent 1 age’ and all other family variables

and the product of ‘number of stressful life events’ and all other family variables.

Addition of the interaction terms to the model of contextual factors revealed a

significant interaction between the variables ‘Parent 1 BMI’ and ‘Parent 1 self report

of parenting skills’. The effect of ‘Parent 1 BMI’ upon child injury was greater when

parents reported less confidence in their parenting skills. The interaction between

number of stressful life events and parenting skills 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.

Interaction terms for the child domain included the product of ‘male gender’ and all

other child variables and the product of ‘hyperactivity’ and all other child variables.

Addition of these interaction terms to the child domain regression model revealed a

significant interaction between the variables ‘hyperactivity’ and ‘child sleep

problems’ and also between ‘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

20

‘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 including the

product of ‘male gender’, ‘hyperactivity’, ‘heavy traffic’, ‘Parent 1 age’, ‘Parent 1

BMI’, ‘number of stressful life events’, ‘economic hardship’ with all other variables

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 percent

(7%), of the variance in the dependent variable. Figure 9 below displays factors and

interaction terms in the final summary model. Factors and terms marked with an

asterisk were significantly associated with child injury.

Figure 9. Final summary model

E c o n o m ic h a r d s h i pT ra ff icP a re n t 1 a g e *P a re n t 1 B M IL ife e v e n tsP a re n ti n g S k il lsH yp e ra c t iv it yM a le g e n d e rC h o ic e o f a c t iv it yE m o t io n a l h e a l th c o n c e rn sC h il d s le e p p ro b le m s *

E m o t io n a l h e a l th c o n c e rn s x T ra f f ic *E c o n o m ic h a r d s h i p x C h i ld s le e p p r o b s .*

H yp e ra c tiv it y x T ra ffic *C h o ic e o f a c t iv it y x M a le g e n d e r

P a re n t 1 B M I x P a re n t i n g s k ills *L ife e v e n ts x P a re n ti n g s k il ls

I n ju r y

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

21

several other variables including ‘heavy traffic’, ‘number of stressful life events’,

‘hyperactivity’, ‘child’s choice of activity’ and ‘self report of parenting skills’ 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

skills’ 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

to influence injury occurrence through indirect or ‘third variable’ effects. To do so, a

series of three logistic regression analyses are 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. Where addition of

the third variable diminishes the relation between the independent variable and the

dependent variable, mediation or confounding pathways may be indicated. 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). 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). Where addition of the third variables results in an increased relation

between the independent variable and the dependent variable, pathways may involve

suppression effects. For the purpose of these analyses where necessary continuous

variables were reduced to a binary form with 0 indicating below mean scores and 1

indicating above mean scores. Significance of third variable effects were assessed

using Sobel’s test method (Sobel, 1986). Full details of all analyses conducted are

presented in the Appendix.

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.

22

Third variable effects involving ‘economic hardship’ Third variable analyses reveal that the background risk factor ‘economic hardship’ is

implicated in numerous indirect pathways to child injury. Including economic

hardship as a third variable in the regression model significantly reduced the

magnitude of the relation between child injury and the contextual factors; ‘public

housing’ ‘how you feel about your neighbourhood as a place to raise children’, and

‘background noise in the home’, the family factors; ‘Parent 1 education’, ‘self report

of parenting skills’ ‘number of stressful life events’ and the child factor ‘emotional

health a worry’. These results suggest economic hardship may influence child injury

through mediation or confounding pathways.

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. Economic

hardship may confound the relationship between parenting skill and child injury in the

following way; parents experiencing greater economic hardship may express less

confidence in their parenting skill than parents who were under less economic strain,

and children from families experiencing economic hardship may also more likely to

be injured. Parenting skills and child injury are thus related through a common

confounder, economic hardship. Parenting skills do not cause economic hardship,

which then causes injury, but the relationship between parenting skills and child

injury is reduced in magnitude because the distortion due to economic hardship is

removed.

Including economic hardship in the regression model, significantly increased the

magnitude of the relationship between ‘child sleep problems’ and child injury,

suggesting 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.

23

Economic hardship also significantly influences child injury through indirect effects

where; ‘Parent 1 BMI’, ‘neighbourhood liveability’, ‘clutter in the home’ ‘heavy

traffic in the street’ ‘Parent 1 sleep quality’ and ‘housing quality’ are entered as third

variables. In these instances the relationship between economic hardship and child

injury is explained by the causal relationship between economic hardship and factors

such as clutter in the home etc. Figure 10. illustrates the pathways via which

economic hardship may be associated with child injury. The letters (M), (C), and (S)

indicate that economic hardship significantly mediates (M), confounds (C), or

suppresses (S) the influence of the independent variable upon the dependent variable

(child injury). Variables marked with an asterisk were found to significantly mediate

the relation between economic hardship and child injury.

Figure 10. Third variable effects involving ‘economic hardship’

E c o n o m i c h a r d s h i p

• P u b l ic h o u s in g ( C )• N e i g h b o u r h o o d a s a p la c e t o r a is e c h ild r e n ( C )• N o is e ( C )• P a r e n t in g s k i l l ( C )• C h i ld ’ s e m o t io n a l h e a lt h a w o r r y ( C )• P a r e n t 1 e d u c a t io n ( M )• S t r e s s f u l l if e e v e n t s ( M )• C h i ld s l e e p p r o b l e m s ( S )• N e i g h b o u r h o o d l iv e a b i l it y *• C lu t t e r in t h e h o m e *• H e a v y t r a f f ic *• H o u s in g q u a l it y *• P a r e n t 1 B M I *• P a r e n t 1 s le e p q u a l it y *• H y p e r a c t iv it y *

I n ju r y

Third variable effects involving ‘number of stressful life events’ The family risk factor ‘number of stressful life events’ was also found to be involved

in fewer indirect pathways to injury than ‘economic hardship’. Including ‘stressful life

events’ as a third variable in the regression model significantly reduced the magnitude

of the relation between child injury and the family factors ‘Parent 1 sleep quality’ and

‘self report of parenting skill’. Similar effects were observed for the child factors

‘child sleep problems’ and ‘emotional health a worry’. Including ‘stressful life events’

in the regression model, significantly increased the magnitude of the relationship

between ‘Parent 1 BMI’ and child injury, suggesting ‘stressful life events’ may also

operate as suppressor variable. The factor ‘stressful life events’ was also found to

24

significantly influence child injury through indirect effects where the factors ‘marital

status’, and ‘Parent 1 education’ were entered as third variables. Figure 11. illustrates

the pathways via which ‘stressful life events’ may be associated with child injury,

again the letters (M), (C), and (S) indicate that ‘stressful life events’ significantly

mediate (M), confound (C), or suppress (S) the influence of the independent variable

upon the dependent variable (child injury). Variables marked with one asterisk were

found to significantly mediate the relation between ‘stressful life events’ and child

injury, whereas those marked with two asterisks were found to significantly confound

the relation between the ‘stressful life events’ and child injury.

Figure 11. Third variable effects involving ‘stressful life events’

S tr e s s f u l l if e e v e n ts

• P a re n t 1 s l e e p q u a l i t y (C )• P a re n t i n g s k i l l (C )• C h i l d sl e e p p ro b l e m s (C )• C h i l d e m o ti o n a l h e a l t h a w o r ry (C )• P a re n t 1 B M I (S )• M a r t i a l s t a t u s *• P a re n t 1 e d u c a t i o n * *

In j u r y

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 via 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

25

characteristics to act as risk factors for child injury and for inter-relationships between

significant risk factors to provide some insight into the pathways and processes via

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 percent (40%) greater risk for injury compared to children living in

homes in better condition. Children living in public housing were almost 30 percent

(28%) 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 around 30 percent (29%) 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

26

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. Assessment of the relationships among and between significant risk factors and

child injury revealed both interaction and third variable models as likely pathways or

processes via which injury risk was transmitted.

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 via their inter-

relationships with other significant risk factors. Importantly, economic hardship had

an indirect influence upon child injury via its causal relationship with a number of

other significant risk factors across domains of experience and characteristic. 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).

27

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 via interactive and indirect pathways.

These findings should, however, be considered in light of limitations related to

measurement and methodology. Firstly, while commonly used in the injury literature,

caregiver report of child injury requiring medical attention may be subject to recall

bias (Morongiello, 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 further refine the analyses conducted to needed 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, 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.

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 via

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. Government early childhood and

parenting programs that target some of the family and child risk factors identified in

this paper may be an important complement to existing preventative programs

addressing contextual risks.

28

References

• Alexander, C.S., Somerfield, M.R., Ensminger, M.E., Kim, Y.J., & Johnson, K.E. (1995),

Gender differences in injuries among rural youth. Injury Prevention, 1,188-193.

• Alwash, R., & McCarthy, M. (1988). Accidents in the home among children under 5: Ethnic

differences or social disadvantage? British Medical Journal, 196; 1450-3.

• Alwash, R., & McCarthy, M. (1988). Measuring severity of injuries to children from home

accidents. Archives of Disease in Childhood, 63, 635-638.

• Al-Yaman, F., Bryant, M., & Sargeant, H. (2002). Australia’s Children : Their Health and

Wellbeing. AIHW Cat.No. PHE 36. Canberra: AIHW.

• Australian Bureau of Statistics. (2004). The Health of Children. Canberra : Australian Bureau

of Statistics.

• Baron, R.M., & Kenny, D.A. (1986). The moderator-mediator variable distinction in social

psychology research: Conceptual, strategic and statistical considerations. Journal of

Personality and Social Psychology, 51, 1173 – 1182.

• Beautrais, A. L., Fergusson, D. M., & Shannon, D. T. (1982). Childhood accidents in a New

Zealand birth cohort. Australian Pediatric Journal, 18, 238-242.

• Bijur, P. E., Stewart-Brown, S., & Butler, N. (1986). Child behavior and accidental injury in

11,966 preschool children. American Journal of Diseases of Children, 140, 487-492.

• Bronfenbrenner, U. (1979). The ecology of human development: Experiments by nature and

design. Cambridge, MA: Harvard University Press.

• Burgess, T. (1995). A discussion of the relationship between socio-economic factors and the

prevention of childhood injuries. The Forum, 3, 63-68.

• Campbell, S.B., March, C.L., Pierce, E.W., Ewing, L.J., & Szumowski, E.K.(1991). Hard-to-

manage preschool boys : Family context and the stability of externalizing behavior. Journal of

Abnormal Child Psychology, 19, 301-318.

• Durkin, M. S., Davidson, L. L., 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, 84, 587-592.

• Dougherty, G., Pless, I.B., & Wilkins, R. (1990). Social class and the occurrence of traffic

injuries and deaths in urban children. Canadian Journal of Public Health, 81, 204-209.

• Garbarino, J. (1977). The human ecology of child maltreatment: A conceptual model for

research. Journal of Marriage and Family, 39,721-735.

• Hango, D.W., & Houseknecht, S.K., (2005), Marital disruption and accidents/injuries among

children. Journal of Family Issues, 26, 3-31.

• Hosmer, D.W., & Lemeshow, S. (1989).Applied Logistic Regression. New York : John Wiley

& Sons.

• Glik, D., Kronenfeld, J., & Jackson, K.(1993). Sex differences in children’s risk taking

behavior. Child Development, 53, 426-428.

29

• Hecht, D. B., & Hansen, D. J. (2001). The environment of child maltreatment: Contextual

factors and the development of psychopathology. Aggression and Violent Behavior, 6, 433-

457.

• Horowitz, S.M., Morgenstern, H., DiPietro, L., & Morrison, C.L. (1988). Determinants of

pediatric injuries. American Journal of Diseases in Childhood, 142, 605-611.

• Jencks, C., & Mayer, S.E.(1990).The social consequences of growing up in a poor

neighbourhood. In L.E.Lynn , Jr. and G.H.McGeary (Eds.), Inner City Poverty in the United

States (pp. 111-186). Washington DC: National Academy Press.

• Jolly, D.L., Moller, J.N., & Volkmer, R.E. (1993). The socioeconomic context of child injury

in Australia. Journal of Paediatrics and Child Health, 29, 438-444.

• KIDSAFE (2004). About Kidsafe, http://www.kidsafe.com.au/,

• Klein, D. (1980). Societal influences on childhood accidents. Accident and Prevention, 12,

275-281.

• Matheny, A. P. (1987). Psychological characteristics of childhood accidents. Journal of Social

Issues, 43, 45-60.

• MacKinnon, D.P., Krull, J.L., & Lockwood, C.M., (2000). Equivalence of the mediation,

confounding and suppression effect, Prevention Science, 1,173-181.

• Matheny, A.P. (1988) Accidental injuries. In. Routh, D. (Ed).Handbook of pediatric

psychology. New York: Guilford.

• McCormick, M.C., Shapiro, S., & Starfield, B.H. (1981). Injury and its correlates among 1-

year-old children. American Journal of Diseases of Children, 135, 159-163.

• Morongiello, B. A. (1997). Children's perspectives on injury and closs-call experiences : Sex

differences in injury-outcome processes. Journal of Pediatric Psychology, 22, 499-512.

• Morongiello, B. A., Ondejko, L., & Littlejohn, A. (2004). Understanding toddlers in-home

injuries : I Contexts, correlates and determinants. Journal of Pediatric Psychology, 29, 415-

431.

• Nersesian, W.S., Petit, M.R., Shaper, R., Lemieux, D., & Naor, E. (1985) Childhood death and

poverty : A study of all childhood deaths in Maine, 1975 to 1980. Pediatrics, 75, 41-50.

• Parker, R.M., Rescorla, L.A., Finkelstein, J.A., Barnes, N., Holmes, J.H., & Stolley,

P.D.(1991). A survey of the health of homeless children in Philadelphia shelters. American

Journal of Disease in Children, 145, 520-526.

• Pearn, J., & Nixon, J. (1977). Prevention of childhood drowning accidents. Medical Journal of

Australia, 1, 616-618.

• Peterson, L., & Brown, D. (1994). Integrating child injury and abuse-neglect research :

Common histories, etiologies and solutions. Psychological Bulletin, 116, 293-315.

• Platt, M.J., & Pharaoh, P.O.D. (1996) Child health statistical review, 1996. Archives of

Disease in Childhood, 75, 527-533.

• Plumert, J. M., & Schwebel, D. C. (1997). Social and temperamental influences on children's

overestimation of their physical abilities: Links to accidental injuries. Journal of Experimental

Child Psychology, 67, 317-337.

30

• Radke-Yarrow, M., Cummings, E.M., Kuczynski, L., & Chapman, M. (1985). Patterns of

attachment in two- and three-year olds in normal families and families with parental

depression. Child Development, 56, 884-895.

• Ramsay, L. J., Moreton, G., Gorman, D. R., Blake, E., Goh, D., Elton, R. A., & Beattie, T. F.

(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, 117, 404-411.

• Reading, R., Langford, I. H., Haynes, R., & Lovett, A. (1999). Accidents to preschool children

: Comparing family and neighbourhood risk factors. Social Science & Medicine, 48, 321-330.

• Rivara, F., Bergman, A., LoGerfo, J., & Weiss, N. (1982). Epidemiology of childhood injuries

II. Sex differences in injury rates. American Journal of Diseases of Children, 136, 502-506.

• Roberts, I., Norton, R., Dunn, R., Hassall, I., & Lee-Joe, T. (1994). Environmental factors and

child pedestrian injuries. Australian Journal of Public Health, 18, 43-46.

• Rodriguez, J. (1990). Childhood injuries in the United States : A priority issue. American

Journal of Social Issues, 43, 105-118.

• Schwebel, D. C. (2004). Temperamental risk factors for children's unintentional injury : The

role of impulsivity and inhibitory control. Personality and Individual Differences, 37, 567-

578.

• Schwebel, D. C., Brezausek, C. M., Ramey, S. L., & Ramey, C. T. (2004). Interactions

between child behavior patterns and parenting : Implications for children's unintentional injury

risk. Journal of Pediatric Psychology, 29, 93-104.

• Sobel, M.E. (1986). Some new results on indirect effects and their standard errors in

covariance structure models. In N. Tuma (Ed.), Sociological methodology, (pp.159-186).

Washington, DC: American Sociological Association.

• 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, 11, 403-408.

• United Nations Children's Fund. (2000). A league table of child deaths by injury in rich

nations. Florence: UNICEF.

• Wazana, A. (1997). Are there injury-prone children? A critical review of the literature.

Canadian Journal of Psychiatry, 42, 602-10.

• Weissman, M.N., John, K., Merikangas, K.R., Prusoff, B.A., Wickramaratne, P., Gammon,

D., Angold, A., and Warner, V. (1986). Depressed parents and their children. General health,

social, and psychiatric problems. American Journal of Diseases in Children, 140, 801-805.

• Wilson, M., Baker, S., Teret, S., Shock, S., & Gabarino, J. (1991), Saving children: A guide to

injury prevention. New York: Oxford University Press.

31

Comparative Analysis ~ Parent/family characteristics, injured vs. non-injured children Non-Injured Children -Injured Children Parent/family variables (continuous).

Significance calculated using unweighted data Unweighted sample Weighted sample Unweighted sample Weighted sample N Mean S.D N Mean S.D N Mean S.D N Mean S.D Parent 1 “Body Mass Index Measure”(banded measure) Significance: F(1, 2812) =9.375, p=.002.

2317 2.58 1.090 149625 2.59 1.100 496 2.74 1.147 25102 2.78 1.167

Parent 1 “Age in years” Significance: F(1, 4974) =14.279, p=<.001

4093 34.89 5.481 207064 34.78 5.545 882 34.12 5.435 45707 33.98 5.454

Parent 1 “Self report of parenting skills” Significance: F(1, 4961) =4.053, p=.044

4081 3.94 .885 206382 3.95 .895 881 3.88 .899 45655 3.89 .899

Parent 1 “Number of live events in past year” Significance: Kruskal Wallis ( ²=9.456, df(1), N(4225),p..002)

3492 1.57 1.716 175607 1.59 1.746 733 1.86 2.042 38168 1.92 2.089

Non-Injured Children Injured Children Unweighted Weighted Unweighted Weighted

Parent/family variables (categorical)

N % N % N % N %

Significance

(calculated using unweighted data) Parent 1 “Sleep problems” Very good Fairly good Fairly bad Very bad

777 2043 898 376

19.0% 49.9% 21.9% 9.2%

39398 102552 45383 19771

19.0% 49.5% 21.9% 9.5%

139 424 222 97

15.8% 48.1% 25.2% 11.0%

7177 21839 11479 5194

15.7% 47.8% 25.1% 11.4%

Pearson ( ²=10.462, df(3), N(4976),p.015)

Relative Risk using recoded variable (very or fairly good vs. fairly or very bad)

= 1.20 (95% c.i <1.06, 1.36>) Parent 1 “Highest year of education” Year 12 Year 11 Year 10 Year 9 Year 8 or below.

2431 532 847 174 107

59.4% 13.0% 20.7% 4.3% 2.6%

102516 33467 53251 10742 7005

49.5% 16.2% 25.7% 5.2% 3.4%

461 144 214 39 23

52.3% 16.3% 24.3% 4.4% 2.6%

18909 8659 14011 2613 1527

41.1% 18.9% 30.6% 5.7% 3.3%

Pearson ( ²=16.647, df(4), N(4972),p.002) Relative Risk using recoded variable

( Yr 12 vs. Below Yr 12) = 1.26, (95% c.i. <1.12, 1.43>)

Parent 1 “Any medical conditions” No Yes

3023 1071

73.8% 26.2%

152348 54758

73.6% 26.4%

621 262

70.3% 29.7%

31707 14060

69.3% 30.7%

Pearson ( ²=4.567, df(1), N(4977),p.033)

Relative Risk = 1.15 (95% c.i.<1.01, 1.32>) Parent 1 “Post natal depression after birth of study child” No Yes

2634 449

85.4% 14.6%

131652 22764

85.3% 14.7%

533 119

81.7% 18.3%

27605 6215

81.6% 18.4%

Pearson ( ²=5.676, df(1), N(3735),p.017)

Relative Risk = 1.24 (95% c.i.<1.04, 1.49>)

32

Comparative Analysis~ Child characteristics, injured vs. non-injured children Non-Injured Children -Injured Children Child Variables (continuous).

Significance calculated using unweighted data Unweighted sample Weighted sample Unweighted sample Weighted sample N Mean S.D N Mean S.D N Mean S.D N Mean S.D “Age of child in months” Significance: F(1,4976) =3.869, p=.049.

4094 56.94 2.670 207105 57.06 2.683 883 46.75 2.509 45766 56.87 2.527

“Child Hyperactivity (mean SDQ scale score)” Significance: F(1, 4963) =25.632, p=.<.001

4082 3.43 2.270 206335 3.50 2.274 882 3.86 2.336 45734 3.97 2.320

“Child Persistence (mean SDQ scale score)” Significance: F(1, 4194) =6.455, p=.011

3472 3.937 .945 174588 3.92 .962 723 3.839 .965 37545 3.892 .982

Child Prosociality (mean SDQ scale score)” Significance: F(1, 4963) =4.790, p=.029

4082 7.76 1.780 206335 7.75 1.788 882 7.62 1.824 45734 7.62 1.856

“Child Conduct (mean SDQ scale score)” Significance: Kruskal Wallis ( ²=20.093, df(1), N(4964),p001)

4082 7.57 1.979 206335 7.53 2.000 882 7.21 2.127 45734 7.14 2.133

Non-Injured Children Injured Children Unweighted Weighted Unweighted Weighted

Child variables (categorical)

N % N % N % N %

Significance

(calculated using unweighted data) “Child’s emotional health is a worry?” None A little bit Somewhat Quite a lot A lot

1880 1338 517 263 96

45.9% 32.7% 12.6% 6.4% 2.3%

96395 66393 25549 13675 5094

46.5% 32.1% 12.3% 6.6% 2.5%

365 272 127 85 34

41.3% 30.8% 14.4% 9.6% 3.9%

18641 13959 6468 4744 1954

40.7% 30.5% 14.1% 10.4% 4.3%

Pearson ( ²=22.860, df(4), N(4977),p<.001)

Relative Risk using recoded variable (No concerns vs.Cconcerns)

= 1.17, (95% c.i <1.03, 1.32>)

“Child’s sleeping patterns a problem?” Large problem Moderate problem Small problem No problem

174 356 815 2747

4.3% 8.7% 19.9% 67.1%

9142 17878 40943 139057

4.4% 8.6% 19.8% 67.2%

52 90 185 556

5.9% 10.2% 21.0% 63.0%

2743 4988 9120 28915

6.0% 10.9% 19.9% 63.2%

Pearson ( ²=8.372, df(3), N(4975),p..039)

Relative Risk using recoded variable (No problems vs. Problems)

= 1.16, (95% c.i. <1.03, 1.31>) “Child gender” Female Male

2065 2029

50.4% 49.6%

103593 103512

50.0% 50.0.%

378 505

42.8% 57.2%

19773 25994

43.2% 56.8%

Pearson ( ²=16.925, df(1), N(4977), p<.001) Relative Risk = 1.29 (95% c.i.<1.14, 1.45>)

33

Non-Injured Children Injured Children

Unweighted Weighted Unweighted Weighted Child variables (categorical) cont.

N % N % N % N %

Significance

(calculated using unweighted data) “Child has ongoing medical conditions No Yes

3294 800

80.5% 19.5%

165719 41386

80.0% 20.0%

668 215

75.7% 24.3%

34298 11468

74.9% 25.1%

Pearson ( ²=10.343, df(1), N(4977),p.001) Relative Risk = 1.26 (95% c.i.<1.10, 1.44>)

“Child has ADD or ADHD” No Yes

4058 36

99.1% 0.9%

205104 2001

99.0% 1.0%

865 18

98.0% 2.0%

44729 1037

37.7% 2.3%

Pearson ( ²=9.094, df(1), N(4977),p.003)

Relative Risk = 1.90 (95% c.i.<1.29, 2.78>) “Child’s choice to spend free time” Inactive pursuits Either inactive or active pursuits Active pursuits

1030 1898 1162

25.2% 46.4% 28.4%

54764 92771 59362

26.5% 44.8% 28.7%

186 413 282

21.1% 46.9% 32.0%

10666 20594 14372

23.4% 45.1% 31.5%

Pearson ( ²=8.177, df(2), N(4971),p..017) Relative Risk using recoded variable

(Inactive vs. Active pursuits) = 1.28, (95% c.i. <1.08, 1.51>)

Tamara Blakemore Research Projects ~ Research and Data Management

34

Multivariate Logistic Regression Models.

Model 2. All significantly differing parent / family variables Parent/family variables: model sig. (Wald χ²= 33.485, df(9), N(2509), p<.001), Nagelkerke R Square =.022 Variables Entered B SE Prob. Odds Ratio 95% C.I Parent 1 age -.024 .011 .028 .976 .956, .997 Parent 1 legal / registered marital status .045 .040 .257 1.046 .967, 1.132 Parent 1 BMI .119 .046 .010 1.126 1.029, 1.233 Parent 1 education .084 .053 .112 1.088 .980, 1.207 Number of stressful life events in past year .051 .029 .075 1.053 .995, 1.214 Parent 1 report of parenting skills .074 .061 .226 1.077 .955, 1.114 Parent 1 medical conditions -.125 .123 .311 .883 .694, 1.124 Parent 1 post natal depression with study child .142 .143 .324 1.152 .870, 1.526 Parent 1 sleep problems .118 .114 .302 1.125 .899, 1.408

Model 3, Significant parent / family variables plus significant interaction terms. Parent/family variables: model sig. (Wald χ²= 43.554, df(9), N(2805), p<.001), Nagelkerke R Square =.025 Variables Entered B SE Prob. Odds Ratio 95% C.I Parent 1 age -.030 .010 .003 .970 .951, .990 Parent 1 legal / registered marital status .032 .038 .406 1.032 .958, 1.112 Parent 1 BMI -.216 .120 .071 .806 .637, 1.019 Parent 1 sleep problems .112 .108 .298 1.119 .906, 1.381 Parent 1 education .059 .050 .237 1.061 .962, 1.170 Parent 1 report of parenting skills -.249 .152 .102 .779 .578, 1.050 Number of stressful life events in past year (parent) .170 .066 .010 1.185 1.041, 1.348 Parent 1 BMI x Self report of parenting skills .148 .051 .003 1.160 1.051, 1.281 Number of life events x parenting skills -.055 .028 .053 .947 .895, 1.001

Model 1. All significantly differing context variables Contextual variables: model sig. (Wald χ²= 31.414, df(9), N(1036), p<.001), Nagelkerke R Square =.048 Variables Entered B SE Prob. Odds Ratio 95% C.I Number of homes child has lived in .082 .108 .449 1.085 .878, 1.342 Neighbourhood liveability .003 .182 .988 1.003 .702, 1.432 Economic hardship .184 .050 .000 1.202 1.089, 1.326 Badly deteriorated or poor housing .055 .280 .845 1.056 .610, 1.827 Neighbourhood poor or very poor to raise children. .231 .197 .240 1.260 .857, 1.854 Heavy traffic in the street .425 .163 .009 1.530 1.111, 2.107 Clutter in the home .065 .250 .794 1.067 .654, 1.742 Loud or moderate background noise -.027 .176 .879 .974 .690, 1.374 Public housing tenant .149 .202 .463 1.160 .780, 1.725

Tamara Blakemore Research Projects ~ Research and Data Management

35

Model 3. All significantly differing child variables Child variables: model sig. (Wald χ=35.227, df(11), N(2208), p<.001), Nagelkerke R Square =.027 Variables Entered B SE Prob. Odds Ratio 95% C.I Male gender (child) .261 .119 .028 1.298 1.029, 1.639 Child’s emotional health a worry to parent 1 .048 .121 .690 1.049 .828, 1.329 Child’s sleep patterns problematic .217 .123 .077 1.242 .977, 1.580 Child has a medical condition -.038 .139 .783 .962 .733, 1.264 Child has ADD / ADHD .668 .411 .104 .1951 .872, 4.362 Child’s age in months -.09 .022 .681 .991 .950, 1.034 Child’s mean prosociality score .062 .036 .084 1.063 .992, 1.140 Child’s mean hyperactivity score .082 .031 .009 1.085 1.020, 1.154 Child’s mean conduct score -.022 .033 .513 .979 .917, 1.044 Child’s mean persistence score .024 .067 .727 1.024 .897, 1.169 Child chooses active ways to spend free time. .194 .117 .096 1.214 .966, 1.527

Model 4. Significant child variables plus significant interaction terms Child variables plus interactions: model sig. (Wald χ²= 47.290, df(8), N(2650), p<.001), Nagelkerke R Square =.029 Variables Entered B SE Prob. Odds Ratio 95% C.I Child’s emotional health a worry to parent 1 .100 .108 .357 1.105 .893, 1.367 Child’s sleep patterns problematic -.228 .220 .300 .796 .517, 1.226 Male gender (child) .358 .108 .001 1.430 1.158, 1.766 Child chooses active ways to spend free time. .251 .105 .017 1.285 1.046, 1.580 Child’s mean hyperactivity score .069 .023 .003 1.071 1.024, 1.120 Hyperactivity x Sleep problem .101 .045 .023 1.106 1.014, 1.207 Emotional health a worry x male gender -.457 .212 .031 .633 .418, .959

Model 5. Significant variables across domains including significant interaction terms from each domain. Integrated model sig. (Wald χ²= 63.072, df(17), N(1443), p<.001), Nagelkerke R Square =.070 Variables Entered B SE Prob. Odds Ratio 95% C.I Economic hardship -.029 .066 .658 .971 .853, 1.106 Heavy traffic in the street -.558/ .303 .066 .572 .316, 1.037 Parent 1 Age -.035 .014 .012 .966 .940, .993 Parent 1 BMI -.263 .168 .118 .769 .552, 1.070 Number of stressful life events in past year (parent) .161 .087 .066 1.174 .990, 1.394 Parent report of parenting skills -.391 .214 .068 .676 .445, 1.029 Child’s mean hyperactivity score .061 .033 .066 1.063 .996, 1.135 Male gender (child) .161 .148 .279 1.174 .878, 1.570 Child chooses active ways to spend free time. .255 .146 .081 1.290 .969, 1.719 Child’s emotional health a worry to parent 1 -.061 .155 .693 .941 .695, 1.374 Child’s sleep patterns problematic .419 .190 .028 1.520 1.047, 2.207 Parent 1 BMI x Self report of parenting skills .189 .071 .008 1.208 1.051, 1.389 Number of life events x parenting skills -.053 .039 .169 .948 .879, 1.023 Child’s emotional health a worry x heavy traffic -.676 .303 .026 .509 .281, .920 Economic hardship x child sleep problems -.298 .120 .013 .742 .587, .938 Child’s mean hyperactivity score x heavy traffic .187 .064 .003 1.206 1.064, 1.366 Child chooses active past time x male gender .518 .292 .076 1.679 .947, 2.977

Tamara Blakemore Research Projects ~ Research and Data Management

36

Investigation of third variable effects ~ contextual domain.

Finding: Economic hardship confounds the relation between the number of homes lived in and child injury Testing steps in third variable effect model B Std Error Prob. Odds Ratio 95% C.I Testing Step 1 (Path c) Outcome: Injury status (non-injured vs. injured). Predictor: Number of homes lived in ª .126 .077 .103 1.134 .975, 1.319 Testing Step 2 (Path a) Third variable: Economic hardshipª Predictor: Number of homes lived inª .416 .059 <.001 1.516 1.350, 1.703 Testing Step 3 (Path b and ć) Outcome: Injury status (non-injured vs. injured). Third variable: Economic hardship .268 .075 <.001 1.308 1.130, 1.514 Predictor: Number of homes lived inª .099 .078 .201 1.104 .949, 1.286 Tests of significance: Sobel’s Test = 3.19, p<.001 Note : C.I = Confidence interval ; Number of homes lived in ª :[ 0= one home only, 1= two or more homes] Economic hardshipª [0= below mean number of hardship items endorsed, 1= above mean number of items endorsed.]

Finding: Neighbourhood liveability mediates the relation between economic hardship and child injury . Testing steps in third variable effect model B Std Error Prob. Odds Ratio 95% C.I Testing Step 1 (Path c) Outcome: Injury status (non-injured vs. injured). Predictor: Economic hardshipª .277 .074 <.001 1.319 1.140, 1.526 Testing Step 2 (Path a) Third variable: :Neighbourhood liveability Predictor: Economic hardshipª .599 .060 <.001 1.820 1.616, 2.048 Testing Step 3 (Path b and ć) Outcome: Injury status (non-injured vs. injured). Third variable: Neighbourhood liveabilityª .161 .077 .036 1.174 1.010, 1.365 Predictor: Economic hardshipª .255 .075 .001 1.290 1.114, 1.495 Tests of significance: Sobel’s Test = 2.05, p=.004 Note : C.I = Confidence interval ; Economic hardshipª [0= below mean number of hardship items endorsed. 1= above mean number of items endorsed.] Neighbourhood liveabilityª [mean neighbourhood liveability score]

Finding: Economic hardship confounds the relation between public housing and child injury . Testing steps in third variable effect model B Std Error Prob. Odds Ratio 95% C.I Testing Step 1 (Path c) Outcome: Injury status (non-injured vs. injured). Predictor: Public housing -.312 .156 .045 .732 .539, .994 Testing Step 2 (Path a) Third variable: Economic hardshipª Predictor: Public housingª -.668 .147 <. 001 .513 .384, .684 Testing Step 3 (Path b and ć) Outcome: Injury status (non-injured vs. injured). Third variable: Economic hardshipª .458 .149 .002 1.582 1.181, 2.118 Predictor: Public housingª .251 .158 .111 1.286 .944, 1.751 Tests of significance: Sobel’s Test = -2.54, p=.001 Note : C.I = Confidence interval ; Public housingª [0= renting but not in public housing. 1 = renting from a government authority] Economic hardshipª [ 0=below mean number of hardship items endorsed, 1= above mean number of items endorsed.].

Tamara Blakemore Research Projects ~ Research and Data Management

37

Finding: Economic hardship confounds the relation between ‘how you feel about your neighbourhood’ and injury. Testing steps in third variable effect model B Std Error Prob. Odds Ratio 95% C.I Testing Step 1 (Path c) Outcome: Injury status (non-injured vs. injured). Predictor: How you feel about your neighbourhoodª -.235 .107 .028 .791 .641, .975 Testing Step 2 (Path a) Third variable: Economic hardshipª Predictor: How you feel about your neighbourhoodª -.601 .086 <.001 .548 .463, .649 Testing Step 3 (Path b and ć) Outcome: Injury status (non-injured vs. injured). Third variable: Economic hardshipª .277 .082 .001 1.319 1.122, 1.550 Predictor: How you feel about your neighbourhoodª .197 .108 .067 1.218 .986, 1.505 Tests of significance: Sobel’s Test = -2.48 p=.001 Note : C.I = Confidence interval ; How you feel about your neighbourhood as a place to raise childrenª [ 0= very good or good, 1= fair, poor or very poor] Economic hardshipª [ 0=below mean number of hardship items endorsed, 1= above mean number of items endorsed.]

Findings: Clutter in the home mediates the relationship between economic hardship and child injury. Testing steps in third variable effect model B Std Error Prob. Odds Ratio 95% C.I Testing Step 1 (Path c) Outcome: Injury status (non-injured vs. injured). Predictor: Economic hardshipª .277 .074 <. 001 1.319 1.140, 1.526 Testing Step 2 (Path a) Third variable: Clutterª Predictor: Economic hardshipª .705 .104 <.001 2.025 1.653, 2.481 Testing Step 3 (Path b and ć) Outcome: Injury status (non-injured vs. injured). Third variable: Clutterª -.326 .123 .008 .722 .568, .918 Predictor: Economic hardshipª .253 .076 .001 1.288 1.110, 1.494 Tests of significance: Sobel’s Test = -.247 p=.001 Note : C.I = Confidence interval ; Economic hardshipª [ 0= below mean number of hardship items endorsed, 1= above mean number of items endorsed] Clutterª (home is cluttered) [0 = no, 1= yes.]

Tamara Blakemore Research Projects ~ Research and Data Management

38

Finding: Housing quality mediates the relation between economic hardship and child injury. Testing steps in third variable effect model B Std Error Prob. Odds Ratio 95% C.I Testing Step 1 (Path c) Outcome: Injury status (non-injured vs. injured). Predictor: Economic hardshipª .277 .074 <. 001 1.319 1.140, 1.526 Testing Step 2 (Path a) Outcome: Housing qualityª Predictor: Economic hardshipª 1.016 .136 <. 001 2.762 2.118, 3.602 Testing Step 3 (Path b and ć) Outcome: Injury status (non-injured vs. injured). Third variable: Housing qualityª .345 .151 .022 1.412 1.051, 1.897 Predictor: Economic hardshipª .249 .075 .001 1.282 1.106, 1.486 Tests of significance: Sobel’s Test = 2.18, p=.020 Note : C.I = Confidence interval ; Economicª = 0= below mean number of hardship items endorsed 1= above mean number of items endorsed] Housing qualityª [0= fair condition or well kept dwelling, 1= badly deteriorated or poor condition dwelling]

Findings: Heavy traffic in the street mediates the relation between economic hardship and child injury. Testing steps in third variable effect model B Std Error Prob. Odds Ratio 95% C.I Testing Step 1 (Path c) Outcome: Injury status (non-injured vs. injured). Predictor: Economic hardshipª .277 .074 <. 001 1.319 1.140, 1.526 Testing Step 2 (Path a) Third variable: Trafficª Predictor: Economic hardshipª .367 .061 <. 001 1.444 1.281, 1.627 Testing Step 3 (Path b and ć) Outcome: Injury status (non-injured vs. injured). Third variable: Trafficª .226 .078 .004 1.253 1.076, 1.460 Predictor: Economic hardshipª .257 .075 .001 1.293 1.117, 1.497

Tests of significance: Sobel’s Test = 2.967 p=.003 Note : C.I = Confidence interval ; Economic hardshipª [ 0= below mean number of hardship items endorsed ,1= above mean number of items endorsed.] Trafficª (There is heavy traffic on our street.)[ 0 = strongly disagree or disagree, 1= strongly agree or agree]

Finding: Economic hardship confounds the relation between background noise in the home and child injury. Testing steps in third variable effect model B Std Error Prob. Odds Ratio 95% C.I Testing Step 1 (Path c) Outcome: Injury status (non-injured vs. injured). Predictor: Noiseª .167 .082 .042 1.182 1.006, 1.389 Testing Step 2 (Path a) Third variable: Economic hardshipª Predictor: Noiseª .420 .065 <.001 1.521 1.340, 1.727 Testing Step 3 (Path b and ć) Outcome: Injury status (non-injured vs. injured). Third variable: Economic hardshipª .265 .075 <.001 1.303 1.125, 1.509 Predictor: Noiseª .142 .083 .087 1.152 .980, 1.356 Tests of significance: Sobel’s Test = 3.10, p=.002 Note : C.I = Confidence interval ; Noiseª(amount of background noise in the home) [0 =limited or no noise, 1= moderate or loud noise.] Economicª [ 0= below mean number of hardship items endorsed 1= above mean number of items endorsed.]

Tamara Blakemore Research Projects ~ Research and Data Management

39

Investigation of third variable effects ~ parent/family domain.

Finding: Marital status mediates the relation between number of stressful life events in past year and child injury Testing steps in third variable effect model B Std Error Prob. Odds Ratio 95% C.I Testing Step 1 (Path c) Outcome: Injury status (non-injured vs. injured). Predictor: Life eventsª .213 .082 .009 1.238 1.055, 1.453 Testing Step 2 (Path a) Third variable: Marital statusª Predictor: Life eventsª -.808 .075 <.001 .446 .385, .517 Testing Step 3 (Path b and ć) Outcome: Injury status (non-injured vs. injured). Third variable: Life eventsª .168 .083 .043 1.183 1.005, 1.393 Predictor: Marital statusª -.317 .094 .001 .728 .605, .875 Tests of significance: Sobel’s Test = 3.22, p=.001 Note : C.I = Confidence interval ; Life eventsª (number of stressful life events endorsed) [0= below mean number of events, 1= above mean number of events] Marital statusª (Legal/registered marital status)[ 0= married, 1= separated, divorced, widowed or never married.]

Finding: Number of stressful life events in past year suppresses the relation between parent 1 BMI and child injury. Testing steps in third variable effect model B Std Error Prob. Odds Ratio 95% C.I Testing Step 1 (Path c) Outcome: Injury status (non-injured vs. injured). Predictor: BMIª -.275 .099 .006 .760 .626, .923 Testing Step 2 (Path a) Third variable: BMIª Predictor: Life eventsª -.279 .077 <.001 .757 .650, .881 Testing Step 3 (Path b and ć) Outcome: Injury status (non-injured vs. injured). Third variable: Life eventsª .232 .100 .020 1.262 1.067, 1.575 Predictor: BMIª .260 .099 .020 1.262 1.037, 1.535 Tests of significance: Sobel’s Test = 1.96, p=.05. Note : C.I = Confidence interval ; Parent 1 BMIª [0 = BMI within the underweight or normal weight ranges. 1= BMI within the overweight or obese weight ranges] Life eventsª (number of stressful life events endorsed)[ 0= below mean number of events, 1= above mean number of events]

Finding: Number of stressful life events in past year confounds the relation between parent sleep quality and child injury. Testing steps in third variable effect model B Std Error Prob. Odds Ratio 95% C.I Testing Step 1 (Path c) Outcome: Injury status (non-injured vs. injured). Predictor: Parent sleep qualityª .226 .078 .004 1.254 1.077, 1.461 Testing Step 2 (Path a) Third variable: Life eventsª Predictor: Parent sleep qualityª .566 .067 <.001 1.761 1.545, 2.008 Testing Step 3 (Path b and ć) Outcome: Injury status (non-injured vs. injured). Third variable: Life eventsª .197 .082 .017 1.218 1.036, 1.432 Predictor: Parent sleep qualityª .131 .087 .131 1.140 .962, 1.351 Tests of significance: Sobel’s Test = 2.31, p=.02 Note : C.I = Confidence interval ; Parent sleep qualityª [0= sleep quality is very good or fairly good, 1= sleep quality is fairly bad or very bad] Life eventsª (number of stressful life events endorsed) [0= below mean number of events, 1= above mean number of events]

Tamara Blakemore Research Projects ~ Research and Data Management

40

Finding: Parent 1 education confounds the relation between number of stressful life events in the past year and child injury Testing steps in third variable effect model B Std Error Prob. Odds Ratio 95% C.I Testing Step 1 (Path c) Outcome: Injury status (non-injured vs. injured). Predictor: Life eventsª .213 .082 .009 1.238 1.055, 1.453 Testing Step 2 (Path a) Third variable: Educationª Predictor: Life eventsª .247 .064 <.001 1.334 1.731, 1450 Testing Step 3 (Path b and ć) Outcome: Injury status (non-injured vs. injured). Third variable: Educationª .272 .082 .001 1.312 1.117, 1.541 Predictor: Life eventsª .200 .082 .015 1.221 1.040, 1.434 Tests of significance: 2.51, p=.01 Note : C.I = Confidence interval ; Life eventsª (number of stressful life events endorsed)[ 0= below mean number of events, 1= above mean number of events] Educationª (Parent 1) [0= completed year 12, 1= did not complete year 12] Finding: Number of stressful life events in the past year confound the relation between parenting skill and child injury. Testing steps in third variable effect model B Std Error Prob. Odds Ratio 95% C.I Testing Step 1 (Path c) Outcome: Injury status (non-injured vs. injured). Predictor: Parenting skillª .084 .042 .044 1.088 1.002, 1.181 Testing Step 2 (Path a) Third variable: Life eventsª Predictor: Parenting skillª .125 .036 <.001 1.134 1.057, 1.216 Testing Step 3 (Path b and ć) Outcome: Injury status (non-injured vs. injured). Third variable: Life eventsª .204 .082 .013 1.226 1.044, 1.440 Predictor: Parenting skillª .065 .046 .163 1.067 .974, 1.168 Tests of significance: 2.02, p=.002 Note : C.I = Confidence interval ; Parenting skillª [self report of parenting skill (very good , above average, average, some trouble & not very good)] Life eventsª (number of stressful life events endorsed) [0= below mean number of events, 1= above mean number of events] Investigation of third variable effects ~ child domain. Finding: Hyperactivity confounds the relation between child’s emotional health and child injury Testing steps in third variable effect model B Std Error Prob. Odds Ratio 95% C.I Testing Step 1 (Path c) Outcome: Injury status (non-injured vs. injured). Predictor: Emotional health .187 .075 .013 1.205 1.040, 1.396 Testing Step 2 (Path a) Third variable: Hyperactivity ª Predictor: Emotional healthª .555 .058 <.001 1.742 1.556, 1.951 Testing Step 3 (Path b and ć) Outcome: Injury status (non-injured vs. injured). Third variable Hyperactivity ª .316 .075 <.001 1.372 1.184,1.590 Predictor: Emotional healthª .140 .076 .065 1.151 .991, 1.336 Tests of significance: 3.86, p<.001 Note : C.I = Confidence interval ; Child’s emotional healthª (Worry over child’s emotional health?)[ 0 = none, 1= a little bit, somewhat, quite a lot or a lot] Hyperactivityª (mean hyperactivity scale score) [0 = below mean , 1= above mean]

Tamara Blakemore Research Projects ~ Research and Data Management

41

Investigation of third variable effects ~ across domains.

Finding: Parent 1 BMI mediates the relation between economic hardship and child injury. Testing steps in third variable effect model B Std Error Prob. Odds Ratio 95% C.I Testing Step 1 (Path c) Outcome: Injury status (non-injured vs. injured). Predictor: Economic hardshipª .277 .074 .<.001 1.319 1.140, 1.526 Testing Step 2 (Path a) Third variable: BMIª Predictor: Economic hardshipª .384 .077 <.001 1.468 1.262, 1.708 Testing Step 3 (Path b and ć) Outcome: Injury status (non-injured vs. injured). Third variable: BMIª .256 .100 .010 1.292 1.063, 1.570 Predictor: Economic hardshipª .201 .100 .045 1.222 1.055, 1.487 Tests of significance: Sobel’s Test = 2.28, p=.02 Note : C.I = Confidence interval ; Economic hardshipª [0= below mean number of hardship items endorsed 1= above mean number of items endorsed] Parent 1 BMIª [0 = BMI within the underweight or normal weight ranges. 1= BMI within the overweight or obese weight ranges]

Finding: Number of stressful life events in past year confounds the relation between child sleep problems and child injury. Testing steps in third variable effect model B Std Error Prob. Odds Ratio 95% C.I Testing Step 1 (Path c) Outcome: Injury status (non-injured vs. injured). Predictor: Child sleep problemsª .183 .077 .018 1.201 1.032, 1.397 Testing Step 2 (Path a) Third variable: Life eventsª Predictor: Child sleep problemsª .389 .066 <.001 1.475 1.296, 1.678 Testing Step 3 (Path b and ć) Outcome: Injury status (non-injured vs. injured). Third variable: Life eventsª .200 .082 .015 1.221 1.040, 1.435 Predictor: Child sleep problemsª .149 .085 .081 1.160 .982, 1.371 Tests of significance: Sobel’s Test = 2.25, p=.02 Note : C.I = Confidence interval ; Child sleep problemsª [0= sleeping patterns are not problematic, 1= Sleeping patterns are a small, moderate or large problem]. Life eventsª (number of stressful life events endorsed)[ 0= below mean number of events, 1= above mean number of events]

Finding: Economic hardship suppresses the relation between child sleep problems and child injury Testing steps in third variable effect model B Std Error Prob. Odds Ratio 95% C.I Testing Step 1 (Path c) Outcome: Injury status (non-injured vs. injured). Predictor: Child sleep problemsª .183 .077 .018 1.201 1.032, 1.397 Testing Step 2 (Path a) Third variable: Economic hardshipª Predictor: Child sleep problemsª .317 .060 <.001 1.372 1.219, 1.544 Testing Step 3 (Path b and ć) Outcome: Injury status (non-injured vs. injured). Third variable: Economic hardshipª .265 .075 <.001 .1.304 1.126, 1.509 Predictor: Child sleep problemsª .163 .078 .036 1.177 1011, 1.370 Tests of significance: Sobel’s Test = 2.94, p=.003 Note : C.I = Confidence interval ; Child sleep problemsª [0= sleeping patterns are not problematic, 1= sleeping patterns are a small, moderate or large problem]. Economic hardshipª [0= below mean number of hardship items endorsed, 1= above mean number of items endorsed].

Tamara Blakemore Research Projects ~ Research and Data Management

42

Finding: Parent 1 sleep quality mediates the relation between economic hardship and child injury Testing steps in third variable effect model B Std Error Prob. Odds Ratio 95% C.I Testing Step 1 (Path c) Outcome: Injury status (non-injured vs. injured). Predictor: Economic hardshipª .277 .074 .<.001 1.319 1.140, 1.526 Testing Step 2 (Path a) Third variable: Parent sleep qualityª Predictor: Economic hardshipª .550 .061 <.001 1.733 1.537, 1.955 Testing Step 3 (Path b and ć) Outcome: Injury status (non-injured vs. injured). Third variable: Sleep qualityª .193 .079 .014 1.213 1.040, 1.414 Predictor: Economic hardshipª .251 .075 .001 1.285 1.110, 1.489 Tests of significance: 2.36, p=.01 Note : C.I = Confidence interval ; Economic hardshipª [0= below mean number of hardship items endorsed, 1= above mean number of items endorsed.] Parent sleep qualityª[ 0= sleep quality is very good or fairly good, 1= sleep quality is fairly bad or very bad]

Finding: Economic hardship mediates the relationship between parent 1 education and child injury. Testing steps in third variable effect model B Std Error Prob. Odds Ratio 95% C.I Testing Step 1 (Path c) Outcome: Injury status (non-injured vs. injured). Predictor: Educationª .288 .075 <.001 1.334 1.153, 1.544 Testing Step 2 (Path a) Third variable: Economic hardshipª Predictor: Educationª .715 .058 <.001 2.043 1.822, 2.292 Testing Step 3 (Path b and ć) Outcome: Injury status (non-injured vs. injured). Third variable: Economic hardshipª .236 .076 .002 .1.266 1.092, 1.469 Predictor: Educationª .248 .076 .001 1.282 1.105, 1.487 Tests of significance: 3.01, p=.003 Note : C.I = Confidence interval ; Educationª (Parent 1)[ 0= completed year 12, 1= did not complete year 12] Economic hardshipª [0= below mean number of hardship items endorsed, 1= above mean number of items endorsed]

Finding: Economic hardship confounds the relation between parenting skill and child injury. Testing steps in third variable effect model B Std Error Prob. Odds Ratio 95% C.I Testing Step 1 (Path c) Outcome: Injury status (non-injured vs. injured). Predictor: Parenting skillª .084 .042 .044 1.088 1.002, 1.181 Testing Step 2 (Path a) Third variable: Economic hardshipª Predictor: Parenting skillª .150 .032 <.001 1.162 1.091, 1.238 Testing Step 3 (Path b and ć) Outcome: Injury status (non-injured vs. injured). Third variable: Economic hardshipª .265 .075 <.001 1.303 1.126, 1.508 Predictor: Parenting skillª .073 .042 .082 1.075 .991, 1.167 Tests of significance: 2.82, p=.004 Note : C.I = Confidence interval ; Parenting skillª [self report of parenting skill (very good , above average, average, some trouble & not very good)] Economic hardshipª[ 0= below mean number of hardship items endorsed, 1= above mean number of items endorsed]

Tamara Blakemore Research Projects ~ Research and Data Management

43

Finding: Economic hardship mediates the relation between number of stressful life events in past year and child injury** Testing steps in third variable effect model B Std Error Prob. Odds Ratio 95% C.I Testing Step 1 (Path c) Outcome: Injury status (non-injured vs. injured). Predictor: Life eventsª .213 .082 .009 1.238 1.055, 1.453 Testing Step 2 (Path a) Third variable: Economic hardshipª Predictor: Life eventsª 1.025 .065 <.001 2.786 2.455, 3.161 Testing Step 3 (Path b and ć) Outcome: Injury status (non-injured vs. injured). Third variable: Economic hardshipª .246 .084 .003 1.279 1.085, 1.508 Predictor: Life eventsª .151 .084 .073 1.164 .986, 1.373 Tests of significance: 2.88, p=.004 Note : C.I = Confidence interval ; Life eventsª (number of stressful life events endorsed)[ 0= below mean number of events, 1= above mean number of events] Economic hardshipª [0= below mean number of hardship items endorsed, 1= above mean number of items endorsed] **(This relationship may alternately be termed a confounding one).

Finding: Hyperactivity mediates the relation between economic hardship and child injury. Testing steps in third variable effect model B Std Error Prob. Odds Ratio 95% C.I Testing Step 1 (Path c) Outcome: Injury status (non-injured vs. injured). Predictor: Economic hardshipª .277 .074 .<.001 1.319 1.140, 1.526 Testing Step 2 (Path a) Third variable: Hyperactivity ª Predictor: Economic hardship ª .432 .057 <.001 1.540 1.376, 1.724 Testing Step 3 (Path b and ć) Outcome: Injury status (non-injured vs. injured). Third variable: Hyperactivity ª .309 .075 <.001 1.362 1.176, 1.578 Predictor: Economic hardship ª .244 .075 .001 1.277 1.117, 1.479 Tests of significance: 3.62 p<.001 Note : C.I = Confidence interval ; Economic hardshipª [0= Below mean number of hardship items endorsed, 1= above mean number of items endorsed] Hyperactivityª (mean hyperactivity scale score) [0 = below mean, 1= above mean)].

Finding: Number of stressful life events in the past year confound the relation between child’s emotional health and child injury Testing steps in third variable effect model B Std Error Prob. Odds Ratio 95% C.I Testing Step 1 (Path c) Outcome: Injury status (non-injured vs. injured). Predictor: Child’s emotional health ª .187 .075 .013 1.205 1.040, 1.396 Testing Step 2 (Path a) Third variable: Life events ª Predictor: Child’s emotional health ª .626 .064 <.001 1.871 1.650, 2.121 Testing Step 3 (Path b and ć) Outcome: Injury status (non-injured vs. injured). Third variable: Live events ª .195 .083 .018 1.216 1.034, 1.429 Predictor: Child’s emotional health ª .121 .083 .147 1.129 .959, 1.329 Tests of significance: 2.28, p=.02 Note : C.I = Confidence interval ; Child’s emotional healthª (Worry over child’s emotional health?) [0 =none, 1= a little bit, somewhat, quite a lot or a lot] Life eventsª (number of stressful life events endorsed) [0= below mean number of events, 1= above mean number of events].

Tamara Blakemore Research Projects ~ Research and Data Management

44

Finding: Economic hardship confounds the relation between child’s emotional health and child injury Testing steps in third variable effect model B Std Error Prob. Odds Ratio 95% C.I Testing Step 1 (Path c) Outcome: Injury status (non-injured vs. injured). Predictor: Child’s emotional health ª .187 .075 .013 1.205 1.040, 1.396 Testing Step 2 (Path a) Third variable: Economic hardship ª Predictor: Child’s emotional health ª .363 .058 <.001 1.437 1.284, 1.609 Testing Step 3 (Path b and ć) Outcome: Injury status (non-injured vs. injured). Third variable: Economic hardship ª .263 .075 <.001 1.301 1.124, 1.506 Predictor: Child’s emotional health ª .158 .076 .036 1.171 1.010, 1.359 Tests of significance: 3.06 p=.02 Note : C.I = Confidence interval ; Child’s emotional healthª (Worry over child’s emotional health?) [0=–none, 1= a little bit, somewhat, quite a lot or a lot] Economic hardshipª[ 0= below mean number of hardship items endorsed, 1= above mean number of items endorsed]