Childhood Psychiatric Disorder and Unintentional Injury: Findings from a National Cohort Study

Transcription

1 Childhood Psychiatric Disorder and Unintentional Injury: Findings from a National Cohort Study Richard Rowe, 1 PHD, Barbara Maughan, 1 PHD, and Robert Goodman, 2 PhD 1 MRC Social, Genetic, and Developmental Psychiatry Center, Institute of Psychiatry, King s College London, UK, and 2 Department of Child and Adolescent Psychiatry, Institute of Psychiatry, King s College London, UK Objective We set out to examine the relationship between unintentional injury and common child psychiatric disorders, including both internalizing and externalizing diagnoses. Methods The 1999 British Child and Adolescent Mental Health Survey provided a nationally representative sample of over 10,000 children aged 5 15 years. Measures included assessment of diagnoses of psychiatric disorder from the Diagnostic and Statistical Manual of Mental Disorders, fourth edition, unintentional injury, and a range of potentially related psychosocial factors. Results Children with psychiatric disorders had higher rates of unintentional injury. After accounting for psychosocial risk factors and comorbid psychopathology, oppositional defiant disorder (ODD) was independently associated with burns and poisoning. Attention deficit hyperactivity disorder (ADHD) was related to fractures, and depression and anxiety also showed independent links to some injury types. Conclusions ODD and ADHD, rather than conduct disorder, appear to be the externalizing disorders associated with unintentional injury. We discuss possible models of the relationships between internalizing disorders and unintentional injury. Key words accident; psychopathology; children; adolescents. Unintentional injury is a major cause of childhood suffering and mortality. For children aged 5 14 years in the United Kingdom, 29% of deaths in boys (41.7 deaths per million) and 19% of deaths in girls (21.4 deaths per million) were caused by injuries in 1999 (National Statistics, 2001). Epidemiological studies of the relationship between child psychopathology and unintentional injury have been conducted using several large-scale datasets, with particular attention paid to associations with antisocial behavior and hyperactivity. There is much less evidence regarding possible links with other common types of psychopathology, such as anxiety and depression. There is some agreement that antisocial behavior is related to unintentional injury (Bijur, Golding, Haslum, & Kurzon, 1988; Bijur, Stewart-Brown, & Butler, 1986; Davidson, Taylor, Sandberg, & Thorley, 1992; Langley, McGee, Silva, & Williams, 1983). Although the Diagnostic and Statistical Manual of Mental Disorders, fourth edition (DSM-IV) (APA, 1994), mentions increased accident liability as a characteristic of conduct disorder (CD), a case-control study has shown that unintentional injury is also increased in children with oppositional defiant disorder (ODD) (Schwebel, Speltz, Jones & Bardina, 2002). The DSM-IV also mentions physical injury as an associated feature of attention deficit hyperactivity disorder (ADHD), but the empirical evidence regarding hyperactivity and related behaviors has been mixed; a relationship has been reported in some studies (Bijur et al., 1986; Langley et al., 1983) but not others. Davidson et al. (1992) found no evidence of a simple relationship between hyperactivity and later unintentional injury in boys, while Bijur et al. (1988) found that there was a simple relationship between overactivity and later unintentional injury, but that this was not indepen- All correspondence concerning this article should be addressed to Dr Richard Rowe, MRC Social, Genetic, and Developmental Psychiatry Research Center, Box P046, Institute of Psychiatry, De Crespigny Park, Denmark Hill, London SE5 8AF, UK. r.rowe@iop.kcl.ac.uk. Journal of Pediatric Psychology 29(2) pp , 2004 DOI: /jpepsy/jsh015 Journal of Pediatric Psychology vol. 29 no. 2 Q Society of Pediatric Psychology 2004; all rights reserved

2 120 Rowe, Maughan, and Goodman dent of the relationship between unintentional injury and antisocial behavior. In general, the association has been reported to be stronger in cross-sectional studies than in prospective designs. Davidson et al. (1992) noted that this may suggest that cross-sectional findings have been influenced by biases in retrospective injury reporting. A recent study has suggested that ADHD may be related to unintentional injury in a design free from reporting biases. Analysis of population health records from British Columbia have shown that children who had been prescribed methylphenidate were at increased risk of sustaining a variety of injuries (Brehaut, Miller, Raina, & McGrail, 2003). Apart from psychopathology, a number of other demographic, family, and environmental risk factors for unintentional injury have been demonstrated. For example, unintentional injury has consistently been found to be more common in boys than girls. Other risk factors identified for unintentional injury include nontraditional family structure and parental history of teenage pregnancy (O Connor, Davies, Dunn, & Golding, 2000). Because such risk factors are also well-documented correlates of psychopathology (e.g., Rutter, Giller, & Hagell, 1998), it is important to take these confounds into account. Bijur et al. (1986) addressed this issue, including in their analyses a range of family and environmental correlates. They found that the relationships between unintentional injury and psychopathology were reduced after control for environmental confounds but that they remained substantial. Viewed from the alternative perspective, it is also possible that demographic and psychosocial associations with unintentional injury are mediated by psychopathology. For example, the preponderance of antisocial behavior in boys may contribute to the gender difference in rate of injury. This study examines the cross-sectional relationships between childhood injury and psychiatric disorder in a large-scale nationally representative sample of British children aged 5 15 years as collected in the 1999 British Child and Adolescent Mental Health Survey (BCAMHS- 99). Psychiatric disorder was assessed using the Development and Well-Being Assessment interview (DAWBA) (Goodman, Ford, Richards, Gatward, & Meltzer, 2000), which generates clinically assigned DSM-IV diagnoses. A range of psychosocial risk factors was also assessed during the interview. Methods Population Full details of the survey design and measures are available elsewhere (Meltzer, Gatward, Goodman, & Ford, 2000). Briefly, the Child Benefit Register was used as a sampling frame to identify children aged 5 15 years in England, Scotland, and Wales. The Register should provide a relatively complete basis for sampling the child population because the Child Benefit program is available to all families with children in the UK. This frame was stratified by Regional Health Authority, and within that by sociodemographic groupings. Information from up to three informants was collected on 10,438 children. Ethical approval for the BCAMHS-99 was obtained from the appropriate multicenter research ethics committee in the United Kingdom. Informed consent was obtained from both child and parent. Signed parental consent was also required before the teacher was invited to provide information on the index child. Measures Demographic data (including ethnicity, family type, and parental employment status/socioeconomic group) were collected in the parent interview. Psychiatric disorder was assessed using the DAWBA (Goodman et al., 2000), which consists of a parental interview (available for 99.7% of cooperating families), a child interview for children aged 11 years and over (available for 95.3% of age-appropriate children), and a teacher questionnaire (available for 80.3% of children). Data on symptoms and related psychosocial impairment were probed through a structured, computer-assisted interview. In addition, interviewers recorded participants verbatim descriptions of symptomatic behaviors in response to openended questions. Preliminary DSM-IV diagnoses were made by computer algorithm on the basis of symptom and impairment data, using evidence from each available informant. Experienced clinicians then reviewed a computer-generated case summary including both the symptom and impairment data and the verbatim transcripts and either accepted or amended the computergenerated diagnoses. A reliability study (see Heyman et al., 2001) in which two clinicians rated disorder in 500 children showed that the chance-corrected reliability for the presence of any diagnosis was.86 (95% CI 5.78 to.95). The DAWBA assessed all the information required to make DSM-IV diagnoses of a range of psychiatric disorders, including duration criteria and measures of functional impairment. In our analyses, we diagnosed ODD independently of CD, rather than applying the DSM-IV rule excluding ODD diagnoses for children with CD. Information on unintentional injury was also collected during the parental interview. Parents were

3 Psychopathology and Injury 121 asked whether their child had ever suffered a head injury with loss of consciousness, a broken bone other than head injury (hereafter referred to as fractures), a burn requiring hospital admission, or a poisoning requiring hospital admission. Where injuries were reported, the parent was asked to categorize the timing of the incident as either less than 1 month ago, at least 1 month ago but less than 6 months ago, at least 6 months ago but less than a year ago, or more than a year ago. Severe injuries of this kind were relatively uncommon, so we analyzed binary variables indicating whether each type of injury had taken place at any prior time unless otherwise stated. A range of child and family/environmental factors, which are known to correlate with psychiatric disorder (Meltzer et al., 2000), were also collected during the assessment procedures. Many of these factors are also potential correlates of unintentional injury. In order to account for these factors in our analyses, we formed a number of binary indicators of risk. The socioeconomic characteristics of the neighborhood were assessed using the ACORN [A Classification Of Residential Neighbourhoods] scheme (see Meltzer et al., 2000). We formed a binary variable to index whether or not children lived in a striving neighborhood (the most disadvantaged of the broad categorizations made by ACORN), which identified 25% of the sample. Ethnic origin was coded as a binary variable to indicate majority (white, 91%) or minority status (black, Indian, Pakistani, or other, 9%), reflecting the ethnic mix of the child population of Great Britain at the time of the study. Family structure was categorized as traditional (66%), single parent (22%) or stepparent (11%). Binary variables were coded to indicate large family (four or more children; 11%), low income (household income less than 200 [approx $300] per week; 24%), and low social class (22%). Using the Standard Occupation Classification (OPCS, 1991), low social class (22%) was coded if the highest classification of the occupation of both parents was social class IV (partly skilled) or V(unskilled) or if neither parent had ever worked. Family functioning was assessed using the General Functioning Scale of the McMaster Family Assessment Device (Epstein, Baldwin, & Bishop, 1983). A binary variable was coded to index the most unhealthy functioning 19% of families. A binary variable was also formed to identify parents who used harsh physical discipline as indexed by the use of an implement with which to hit their child (2%). A number of measures of the characteristics of the interviewed parent (usually the mother) were also assessed. Current symptoms of anxiety and depression were assessed using the 12-item version of Goldberg and Williams (1988) General Health Questionnaire (GHQ) which has been the subject of a number of validation studies. For example, it was shown to have good sensitivity (76.3%) and specificity (83.4%) in identifying psychiatric diagnoses in a validation study conducted across 15 clinical settings located around the world (Goldberg et al., 1997). We formed a binary variable to identify the most symptomatic 25% of parents on the GHQ which corresponded to the positive endorsement of three items. Education level was also assessed and we coded a binary variable to identify parents who had not passed any public examinations usually taken on leaving school at age 16 (26%). The parent s age at the index child s birth was also recorded and we coded a binary variable to index teenage parenthood (6%). All children were invited to participate in an assessment of intellectual ability. Of particular interest here was global cognitive ability, which was measured using the British Picture Vocabulary Scale (Dunn, Dunn, Whetton, & Burley, 1997), and reading ability, which was measured using the Word Reading Scale of the British Ability Scale, second edition (BAS II) (Elliot, Smith, & McCulloch, 1996), on an age-standardized scale. For both measures we formed binary variables to identify children who performed in the lowest quartile of the distribution. IQ data were available for 9,884 children (94.7%), and reading data were available for 9,337 (89.4%). Data Analyses The survey models of STATA 7 (StataCorp, 2001) were used for all analyses. These models allowed for the use of weights to account for the unequal probability of postal sector selection and to correct for nonresponse bias associated with region, age, and gender (see Meltzer et al., 2000). The models also accounted for the clustering of observations taken from the same postal sector and for stratification by Regional Health Authority and socioeconomic group. We analyzed each type of injury separately, and the first step in our strategy was to identify significant univariate predictors. We employed logistic regression models for this purpose, which calculate odds ratios (ORs) as a measure of association. With a binary predictor, an OR may be interpreted as the difference between the odds of having suffered the particular injury by people who have and have not endorsed the predictor variable. For example, if the OR for ADHD predicting ever having had a burn is 2.3, this shows that the odds of

4 122 Rowe, Maughan, and Goodman Figure 1 The rate of ever having suffered each injury across age in boys. having ever been burned are 2.3-fold greater for children with ADHD than for children who do not have ADHD. Ninety-five percent CIs are also provided with all ORs, and these show the range of values which are consistent with the data. Where the CIs do not include 1, effects are significant at the p,.05 level, and we used this conventional significance level to guide variable selection throughout. Within the psychiatric and psychosocial domains, we built joint predictor models from significant univariate predictors. Significant predictors in these joint models are shown to have effects that are independent of the other predictors in the model. Nonsignificant effects in these joint models imply that the significant univariate relationships were observed only because of relationships with other predictors included in the model. In the case of the psychiatric models, such interrelation of predictors is usually referred to as comorbidity. Significant predictors in these psychiatric and psychosocial models were then combined into a final model of each injury type. We confirmed that the psychiatric correlates of unintentional injury identified in this way did not differ from models in which all predictors were entered simultaneously (details available on request). In view of the wide developmental range included in the study, we checked whether the correlates included in the final models significantly differed in subgroups of children aged 11 years and younger and 12 years and older. Data on all variables of interest in this paper were available for 8,337 (80%) of the 10,438 children included in the study. Although data on psychiatric diagnoses were available for all observations and missing data on unintentional injury were rare (1.2 to 1.3%), information on a number of the risk measures was unavailable for a slightly larger proportion of cases. In Figure 2 The rate of ever having suffered each injury across age in girls. addition to child IQ and reading levels (noted above), other variables with limited availability included income level (6.4%) and teenage parenthood (3.9%). Unintentional injury was reported at similar levels in those with complete and incomplete observations, but those with incomplete data did show elevated rates of CD (1.3% vs. 2.1%; OR 5 1.7; 95% CI ; p 5.004), anxiety (3.5% vs. 4.8%; OR 5 1.4; 95% CI , p 5.003), and ADHD (2.1% vs. 2.8%; OR 5 1.4; 95% CI ; p 5.04). Incomplete observations were omitted from all analyses. This approach should not lead to substantial bias of the results, although the power to detect relationships between unintentional injury and psychopathology may be reduced. Results Injury Prevalence and Age/Sex Trends Twenty percent of girls and 24% of boys had suffered at least one unintentional injury during their lifetime. Fractures were most common (15.8%), followed by head injury (4.3%), burns (2.3%), and poisoning (2.1%). Age trends in unintentional injury at any time prior to interview are shown in Figures 1 (boys) and 2 (girls). As shown in Table I, boys were at greater risk for all types of injury. Across gender, rate of injury increased with age, although not significantly so regarding poisoning. The overall age trend was also nonsignificant for burns, but there was a hint of interaction between age and sex (p 5.053) in that the age trend was flat for boys (OR [for a 5-year increase in age] 5 0.9; 95% CI ; p 5.6) but increased significantly in girls (OR 5 1.5; 95% CI ; p 5.04). The increase in fractures across age was significantly steeper for boys than

5 Psychopathology and Injury 123 Table I. Age and Gender Effects in Unintentional Injury at Any Time Prior to Interview Gender a Age 3 Gender Injury Type Rate in Boys Rate in Girls OR (95% CI) OR b (95% CI) Boys c Age b Trend Girls c Age b Trend Burn ( )** 0.6 ( ) 1.1 (.9 1.4) Poisoning ( )* 1.0 ( ) 1.2 (.9 1.5) Head injury ( )*** 1.1 ( ) 1.6 ( )*** Fracture ( )** 1.5 ( )*** 2.7 ( )*** 1.8 ( )*** OR 5 odds ratio; coefficients in boldface indicate p,.05. a Gender tested in single predictor model. b OR given for a 5-year increase in age. c Single age coefficient given when Age 3 Gender interaction was not significant. * p,.05; ** p,.01; *** p,.001. for girls, and the age trends in poisoning and head injury were similar across gender. Burns Table II shows that burns were more common among children with all psychiatric diagnoses than in those without disorder but that these contrasts reached significance for only ADHD, ODD, and anxiety. Only ODD remained significant when these three disorders were modeled jointly. A number of psychosocial factors predicted burns in simple models, but the multiple predictor modeling indicated that only poor reading ability and stepfamily status were independently significant. In the final model, only ODD and poor reading ability retained significant links with increased risk for burns. Poisoning As shown in Table III, poisoning was significantly more common in children with all types of psychiatric disorder. The effects of only ODD and anxiety remained significant when all disorders were considered jointly, however. Of the psychosocial risk factors, poor child reading, large family size, and high parental GHQ score had effects that were independent from the other psychosocial measures. In the final model, ODD, anxiety, poor reading, and large family size had significant independent effects. Head Injury Rates of head injury were elevated in all types of psychiatric disorder except depression (Table IV), but only anxiety remained significant in the multiple psychiatric predictor model; ADHD was of borderline significance (p 5.055). Of the psychosocial factors, high parental GHQ score and poor child reading were independently significant risk factors, while rates of head injury were lower in children from ethnic minorities. We explored this unexpected finding in more detail and found that white children had similar rates of head injury (4.4%) as Pakistani children (6.5%; p 5.3) but higher rates than blacks (1.7%; p 5.08) and Indians (0.8%; p 5.06). Anxiety remained significant in the final model, as did the psychosocial predictors. Additionally, we found some evidence of interaction between age and high parental GHQ score (p 5.057). There was a strong association with head injury in children aged 11 and younger (OR 5 1.9; 95% CI ; p,.001) but not in children aged 12 and over (OR 5 1.3; 95% CI ; p 5.2). Fractures Of the psychiatric disorders, only ADHD and depression were significantly related to fractures, and these relationships were independent of each other. Ethnic minority status was once again a significant protective factor; white children (16.4%) had a similar rate of fractures to children of Pakistani origin (11.2%; p 5.2) but significantly higher rates than blacks (8.8%; p 5.01) and Indians (5.7%; p 5.002). None of the other psychosocial risk factors was significant. ADHD, depression, and ethnic minority status remained significant predictors in the final model (see Table V). The higher frequency of fractures relative to other injury types meant that fractures during the most recent 12 months (3.0%) could be analyzed. We investigated whether recall bias was involved in the relationship identified between ADHD and fracture. We found that ADHD was related to more recent fractures (OR 5 1.9; 95% CI ; p 5.07) in a very similar way to less recent fractures (OR 5 1.6; 95% CI ; p 5.017); the comparison between these coefficients did not approach significance (OR 5 1.1; 95% CI ; p 5.8).

6 124 Rowe, Maughan, and Goodman Table II. Predictors of Burns % Burned If % Burned If Joint OR a Predictor Predictor Absent Predictor Present Simple OR a Psychiatric Model Psychosocial Model Final Model Psychiatric disorder ADHD ( )* 1.5 ( ) Conduct disorder ( ) ODD ( )*** 2.7 ( )*** 2.7 ( )*** Anxiety ( )* 1.6 ( ) Depression ( ) Intellectual functioning Low child IQ ( )** 1.0 ( ) Poor child reading ( )*** 1.5 ( )* 1.6 ( )** Socioeconomic status Striving neighborhood ( )* 1.0 ( ) Low income ( )*** 1.3 ( ) Low social class ( )*** 1.1 ( ) Family characteristics Ethnic minority ( )* 1.5 ( ) Single-parent family ( )*** 1.4 ( ) Stepparent family ( )* 1.6 ( )* 1.4 ( ) Large family ( ) Unhealthy family functioning ( ) Harsh physical discipline ( ) Parent characteristics Limited educational qualifications ( )*** 1.3 ( ) High GHQ score ( )* 1.2 ( ) Teen parenthood ( )* 1.2 ( ) OR 5 odds ratio; ADHD 5 attention deficit hyperactivity disorder; ODD5 oppositional defiant disorder; GHQ 5 General Health Questionnaire. Coefficients in boldface indicate p,.05. a Adjusted for age and gender. * p,.05; ** p,.01; *** p,.001. Sex Differences in Unintentional Injury Many of the factors identified as correlates of unintentional injury such as externalising disorder are more common in boys than girls. It is therefore possible that the correlates identified above mediate the relationship between gender and unintentional injury. However, the effects of gender in the final models for each type of injury (Table VI) were almost unchanged from the simple effects presented in Table I. Discussion The analyses reported here contribute to the growing body of evidence demonstrating a relationship between psychopathology and unintentional injury during childhood. The BCAMHS-99 has a number of strengths for this purpose, including a large, nationally representative sample and assessment of a range of psychosocial factors known to correlate with psychiatric disorder and unintentional injury. The use of clinically assigned DSM-IV diagnostic measures and the assessment of different types of injury within the same study provide a unique opportunity to assess the form and specificity of relationships between psychiatric disorder and unintentional injury. The existing literature provided evidence that disruptive behaviors were a risk factor for injury, but there was some controversy over the forms of disruptive behavior that were particularly salient. Our study provided clinical assessment of the three most common types of externalizing disorder: ADHD, ODD, and CD. The multivariate analysis techniques allowed us to assess whether the simple relationships identified with injury were independent of both the high levels of comorbidity known to exist between these externalizing disorders (Maughan, Rowe, Messer, Goodman, & Meltzer, 2004) and shared psychosocial risk factors that

7 Psychopathology and Injury 125 Table III. Prediction of Poisoning % Poisoned If % Poisoned If Joint OR a Predictor Predictor Absent Predictor Present Simple OR a Psychiatric Model Psychosocial Model Final Model Psychiatric disorder ADHD ( )* 1.2 ( ) Conduct disorder ( )** 1.0 ( ) ODD ( )*** 3.9 ( )*** 3.4 ( )*** Anxiety ( )*** 2.3 ( )*** 2.2 ( )** Depression ( )* 1.5 ( ) Intellectual functioning Low child IQ ( )* 1.0 ( ) Poor child reading ( )*** 1.7 ( )** 1.6 ( )* Socioeconomic status Striving neighborhood ( ) Low income ( )* 1.2 (.8 1.9) Low social class ( ) Family characteristics Ethnic minority ( ) Single parent family ( )* 1.2 (.8 1.8) Stepparent family ( )* 1.5 ( ) Large family ( )** 1.6 ( )* 1.6 ( )* Unhealthy family functioning ( ) Harsh physical discipline ( ) Parent characteristics Limited educational qualifications ( ) High GHQ score ( ) Teen parenthood ( ) OR 5 odds ratio; ADHD 5 attention deficit hyperactivity disorder; ODD 5 oppositional defiant disorder; GHQ 5 General Health Questionnaire. Coefficients in boldface indicate p,.05. a Adjusted for age and gender. * p,.05; ** p,.01; *** p,.001. the study measured. Unusually for a study of this scale, we were additionally able to assess the relationships between anxiety/depression and injury during childhood, a topic which has received much less attention in the existing literature. Despite the many strengths of this dataset for studying the relationship between injury and psychopathology, at least two weaknesses should be noted. First, unintentional injuries were retrospectively reported. It is possible that parents forgot injuries, particularly if they had taken place in the less recent past, and that other retrospective reporting biases (Wazana, 1997) were involved. More serious injuries are more memorable, however (Harel et al., 1994; Peterson, Ewigman, & Kivlahan, 1993), so it is likely that the severe nature of the injuries studied here has minimized any biases introduced in this way. Second, although we were able to include a wider range of psychosocial correlates than has typically been included in studies of unintentional injury, one feature that may be especially salient the quality of parents supervision of the child s activities was not directly assessed. It is therefore possible that inadequate parental supervision may contribute to some of the associations between psychopathology and injury reported in this study. Results for the four types of injury studied here showed an interesting pattern of similarities and differences. As expected, all four were more common in boys than girls, but only head injuries and fractures showed the strong age trends that might be expected in measures of injury across the lifetime. Poisoning showed no significant increase across age, while there was some evidence that the frequency of burns increased with age for girls but not boys. This suggests that many of the burns and poisonings reported here took place very early in childhood, before the study entry age of 5 years. While

8 126 Rowe, Maughan, and Goodman Table IV. Prediction of Head Injury % Injured If % Injured If Joint OR a Predictor Predictor Absent Predictor Present Simple OR a Psychiatric Model Psychosocial Model Final Model Psychiatric disorder ADHD ( )** 1.8 ( ) Conduct disorder ( )* 1.8 ( ) ODD ( )* 1.1 ( ) Anxiety ( )** 1.8 ( )** 1.7 ( )* Depression ( ) Intellectual functioning Low child IQ ( ) Poor child reading ( )* 1.2 ( ) Socioeconomic status Striving neighborhood ( ) Low income ( ) Low social class ( ) Family characteristics Ethnic minority ( )* 0.5 ( )** 0.5 ( )* Single parent family ( ) Stepparent family ( ) Large family ( ) Unhealthy family functioning ( )** 1.3 ( ) Harsh physical discipline ( ) Parent characteristics Limited educational qualifications ( ) High GHQ score ( )*** 1.5 ( )** 1.6 ( )*** Teen parenthood ( ) OR 5 odds ratio; ADHD 5 attention deficit hyperactivity disorder; ODD5 oppositional defiant disorder; GHQ 5 General Health Questionnaire. Coefficients in boldface indicate p,.05. a Adjusted for age and gender. * p,.05; ** p,.01; *** p,.001. Bold coefficients indicate p,.05. recall biases cannot be ruled out as contributing to what might seem a surprising result, it is possible that the requirement for hospital admission in the definition of these injuries is salient here. Burns and poisonings of a similar nature may continue to befall children at older ages, but hospital admission may more rarely be required when they happen at later stages of development. Within the sample there was very little evidence that the correlates of injury involvement differed for older and younger children. The only hint we found was that high parental GHQ score was more strongly associated with head injury in younger children (11 years and younger) than older children, although this effect fell below the conventional level of significance. The four types of injury also showed interesting variations in the pattern of associated psychiatric disorder and psychosocial risk factors, as discussed below. Previous studies have consistently reported a strong relationship between unintentional injury and antisocial behavior in childhood (Bijur et al., 1986; Langley et al., 1983; Schwebel et al., 2002). This was confirmed in our analyses, and we found that ODD was more strongly related than CD. CD showed simple relationships with poisoning and head injury, but these were not independent of the other psychiatric disorders measured. By contrast, ODD was independently related to both poisoning and burns. As discussed above, it is likely that many of the burns and poisonings studied here took place prior to age 5. The DAWBA addresses symptomatology in the present and recent past, but given the typically early onset of many ODD symptoms (Angold & Costello, 1996), it is plausible that substantial ODD behaviors were evident in many of the diagnosed children at much younger ages. One possible mechanism

9 Psychopathology and Injury 127 Table V. Prediction of Fractures % Injured If % Injured If Joint OR a Predictor Predictor Absent Predictor Present Simple OR a Psychiatric Model Final Model Psychiatric disorder ADHD ( )** 1.7 ( )** 1.6 ( )** Conduct disorder ( ) ODD ( ) Anxiety ( ) Depression ( )* 1.9 ( )* 1.9 ( )* Intellectual functioning Low child IQ ( ) Poor child reading ( ) Socioeconomic status Striving neighborhood ( ) Low income ( ) Low social class ( ) Family characteristics Ethnic minority ( )*** 0.5 ( )*** Single parent family ( ) Stepparent family ( ) Large family ( ) Unhealthy family functioning ( ) Harsh physical discipline ( ) Parent characteristics Limited educational qualifications ( ) High GHQ score ( ) Teen parenthood ( ) OR 5 odds ratio; ADHD 5 attention deficit hyperactivity disorder; ODD5 oppositional defiant disorder; GHQ 5 General Health Questionnaire. Coefficients in boldface indicate p,.05. a Adjusted for age and gender. * p,.05; ** p,.01; *** p,.001. Bold coefficients indicate p,.05. for the relationship with unintentional injury may be through the problematic parent-child relationships that characterize oppositionality. For example, children with ODD may not follow the safety instructions of supervising adults, putting them at greater risk for injury. This finding is consistent with the growing evidence that ODD may have important negative consequences that are independent of its relationships with CD (Greene et al., 2002; Maughan et al., 2004; Pickles et al., 2001). It is possible that the relationship between CD and unintentional injury noted in the DSM-IV might be at least partly due to the substantial comorbidity of CD and ODD. The picture regarding ADHD in the previous literature was less clear. We found that although ADHD had simple relationships with all types of injury measured in this study, this effect was independent of other psychiatric disorders only for fractures. An independent relationship with head injury fell just short of significance. As noted in the introduction, it has been suggested that biases in retrospective assessment of injuries may spuriously inflate the relationship between hyperactivity and unintentional injury in cross-sectional studies (Davidson et al., 1992). Injuries were reported retrospectively in this study, as discussed above. However, substantial bias due to differential rates of forgetting by parents of children with and without ADHD seems unlikely here. ADHD was related similarly to fractures reported to have occurred during the last 12 months and the less recent past. Given the recovery time required for fractures, many of the children injured within the last 12 months would still

10 128 Rowe, Maughan, and Goodman Table VI. Sex Differences in Unintentional Injury after Accounting for Independently Significant Predictors of Unintentional Injury Injury Type Sex Odds Ratio (95% CI) from Final Model Burns 1.4 ( )* Poisoning 1.3 ( ) Head injury 1.7 ( )*** Fractures 1.2 ( )** * p,.05; ** p,.01; *** p,.001. Coefficients in boldface indicate p,.05. have been suffering related difficulties or would only recently have returned to full health at the time of the interview, making it unlikely that recall of these recent fractures would have been compromised in any important ways. Associations between unintentional injury and internalizing disorder have received little attention in previous studies. We found that depression was independently related to fractures, while anxiety was independently related to poisoning and was the only independent psychiatric correlate of head injury. In multiple predictor analyses, we confirmed that these relationships were not products of the comorbidity between the internalizing and externalizing disorders and the relationships between externalizing disorders and unintentional injury. Given the cross-sectional nature of the dataset, it is impossible to determine the order of occurrence of the injury and the onset of emotional difficulties. As with externalizing disorders, it is possible that anxiety and depression put children at increased risk for some types of unintentional injury. One possible mechanism here may be that internalizing disorders disrupt attentional control during potentially hazardous activities, making an accident more likely. First, general anxious and depressive cognitions may be distracting. Second, anxiety about performing the activity itself may cause interference. Alternatively, injury could contribute to risk for psychopathology, and this pathway may be particularly plausible regarding internalizing disorders. A range of adverse life events have been documented as risk factors for anxiety and depression in children (Goodyer, 1994; Goodyer, Cooper, Vize, & Ashby, 1993; Silberg et al., 1999), and it is possible that severe accidents contribute to risk in a similar way. An alternative mechanism for the relationship between head injury and anxiety may be that brain damage suffered as a component of the injury has led to the disorder (Brown, Chadwick, Shaffer, Rutter, & Traub, 1981). Our findings suggest that these differing possibilities require further examination in longitudinal studies. Of the psychosocial risk factors studied, poor reading skill emerged as an independent correlate of both burns and poisoning. One possible interpretation of this association is that reading problems contribute in a direct way to unintentional injury, by making it difficult, for example, for children to read safety instructions. Because most burns and poisonings appear to have been experienced prior to age 5, however, this seems an unlikely explanation in the majority of instances. Instead, two other accounts seem more plausible. First, reading difficulties are known to be associated with increased rates of inattention and overactivity (Hinshaw, 1992), which, even if not severe enough to meet criteria for ADHD, might nonetheless put poor readers at increased risk for unintentional injuries. Second, reading problems may act as a marker for other family difficulties, including stressors that reduce the effectiveness of parents capacity to supervise young children; the association between large family size and risk for poisoning may also point in a similar direction. One unexpected finding was the lower rate of fractures and head injury within some ethnic minorities. Recent findings from the Health Survey for England (Nazroo, Becher, Kelly, & McMunn, 2001) have also, however, documented differences between minority groups in parental report of recent major accidents, and more strikingly in relation to minor accidents. As these authors note, this may suggest underreporting in some groups; alternatively, other social or family correlates may be implicated. Once again, additional research is clearly needed to examine these important issues more closely. Finally, we found robust gender differences in unintentional injury. Consistent with many other studies, boys were at greater risk for all the injuries studied here. We had anticipated that some of this effect might be mediated by gender differences in the correlates of unintentional injury, and externalizing disorders, which are more common in boys than in girls (Rutter et al., 1998), seemed to be the primary candidates here. Our analyses provided little support for this hypothesis, however, as the increased rates of injury in boys compared with girls were little changed when the important correlates of unintentional injury were accounted for. This suggests that other factors might mediate the sex differences; one possibility is that temperamental characteristics such as activity levels and impulsivity are important here (Plumert & Schwebel, 1997), indexing a broader spectrum of risk than that

11 Psychopathology and Injury 129 identified by diagnostic groups alone. Other overlapping possibilities include biological differences and differences in the socialization of boys and girls that make them more or less vulnerable to unintentional injuries. For example, gender role stereotypes for boys may include more injury-prone activities such as contact sports and rough-and-tumble play. Understanding the roots of these gender differences in unintentional injury must remain a challenge for further research. These analyses provide further evidence that child behavior is related to unintentional injury. Modification of behavior therefore offers a potential strategy to reduce the incidence of unintentional injury during childhood. Should it prove replicable, the pattern of results found here suggests that young children with ODD and children of all ages with ADHD may particularly benefit from such interventions in terms of reducing their injury liabilities. The strong associations between ODD and certain types of injury provide additional motivations for clinicians to focus on treatment of oppositional symptomatology, whether or not the disorder presents with other comorbid externalizing disorders. If, as discussed above, the parent/child relationship is of particular importance in mediating this contingency, then interventions targeting improvements in parenting (Scott, Spender, Doolan, Jacobs, & Aspland, 2001) may be of particular benefit for reducing injury risk. Longitudinal studies will be required to identify whether the internalizing disorders increase risk for injury rather than being consequences of it. If anxiety and depression are found to increase risk, however, then effective treatment of these disorders may provide an additional benefit in terms of reducing unintentional injury. As well as targeting psychiatric symptomatology directly, programs aimed at specifically reducing injury in these high-risk groups may also be useful. One important issue here is the level at which to target such interventions (Garbarino, 1988). Children suffering from disruptive disorders may be unreceptive to such interventions given the nature of their symptoms. Therefore programs targeting the primary caregiver, perhaps aiming to reduce the presence of household hazards in the child s environment, may be more appropriate here. Children with internalizing disorders may be more receptive to safety instructions and could therefore also be targeted for injury prevention education, perhaps as part of a familywide intervention. The relationships between injury and psychopathology in childhood mean that health care professionals should be on the lookout for treatable psychiatric disorder among children who present with injury in emergency room and primary care settings. Received December 2, 2002; revisions received April 2, 2003; accepted April 30, 2003 References APA [American Psychiatric Association] (1994). Diagnostic and statistical manual of mental disorders, 4th edition. Washington DC: Author. Angold, A., & Costello, E. J. (1996). Toward establishing an empirical basis for the diagnosis of oppositional defiant disorder. Journal of the American Academy of Child and Adolescent Psychiatry, 35, Bijur, P. E., Golding, J., Haslum, M., & Kurzon, M. (1988). Behavioral predictors of injury in school-age children. American Journal of Diseases of Children, 142, 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, Brehaut, J. C., Miller, A., Raina, P., & McGrail, K. M. (2003). Childhood behavior disorders and injuries among children and youth: A population-based study. Pediatrics, 111, Brown, G., Chadwick, O., Shaffer, D., Rutter, M., & Traub, M. (1981). A prospective study of children with head injuries: III. Psychiatric sequelae. Psychological Medicine, 11, Davidson, L. L., Taylor, E. A., Sandberg, S. T., & Thorley, G. (1992). Hyperactivity in school-age boys and subsequent risk of injury. Pediatrics, 90, Dunn, L., Dunn, L., Whetton, C., & Burley, I. (1997). The British Picture Vocabulary Scale second edition testbook. Windsor: NFER. Elliot, C., Smith, P., & McCulloch, K. (1996). British Ability Scales second edition: Administration and scoring manual. Windsor: NFER. Epstein, N. B., Baldwin, L. M., & Bishop, D. S. (1983). The McMaster Family Assessment Device. Journal of Marital and Family Therapy, 9, Garbarino, J. (1988). Preventing childhood injury: Developmental and child health issues. American Journal of Orthopsychiatry, 58, Goldberg, D. P., Gater, R., Sartorious, N., Uston, T. B., Picinelli, M., Gureje, O., et al. (1997). The validity of two versions of the GHQ in the WHO study of

12 130 Rowe, Maughan, and Goodman mental illness in general health care. Psychological Medicine, 27, Goldberg, D. P., & Williams, P. (1988). A user s guide to the GHQ. Windsor: NFER-Nelson. Goodman, R., Ford, T., Richards, H., Gatward, R., & Meltzer, H. (2000). The Development and Well-being Assessment: Description and initial validation of an integrated assessment of child and adolescent psychopathology. Journal of Child Psychology and Psychiatry, 41, Goodyer, I. M. (1994). Developmental psychopathology: The impact of recent life events in anxious and depressed school-age children. Journal of the Royal Society of Medicine, 87, Goodyer, I. M., Cooper, P. J., Vize, C. M., & Ashby, L. (1993). Depression in year-old girls: The role of past parental psychopathology and exposure to recent life-events. Journal of Child Psychology and Psychiatry, 34, Greene, R. W., Biederman, J., Zerwas, S., Monuteaux, M. C., Goring, J. C., & Faraone, S. V. (2002). Psychiatric comorbidity, family dysfunction, and social impairment in referred youth with oppositional defiant disorder. American Journal of Psychiatry, 159, Harel, Y., Overpeck, M. D., Jones, D. H., Scheidt, P. C., Bijur, P. E., Trumble, A. C., et al. (1994). The effects of recall on estimating annual nonfatal injury rates for children and adolescents. American Journal of Public Health, 84, Heyman, I., Fombonne, E., Simmons, H., Ford, T., Meltzer, H., & Goodman, R. (2001). Prevalence of obsessive-compulsive disorder in the British nationwide survey of child mental health. British Journal of Psychiatry, 179, Hinshaw, S. P. (1992). Academic underachievement, attention deficits, and aggression: Comorbidity and implications for intervention. Journal of Consulting and Clinical Psychology, 60, Langley, J., McGee, R., Silva, P., & Williams, S. (1983). Child behavior and accidents. Journal of Pediatric Psychology, 8, Maughan, B., Rowe, R., Messer, J., Goodman, R., & Meltzer, H. (2004). Conduct disorder and oppositional defiant disorder in a national sample: Developmental epidemiology. Journal of Child Psychology and Psychiatry, 45, Meltzer, H., Gatward, R., Goodman, R., & Ford, T. (2000). Mental health of children and adolescents in Great Britain. London: HMSO. National Statistics (2001). Mortality Statistics: Review of the Registrar General on deaths in England and Wales, 1999 (Series DH1 no. 32). London:HMSO. Nazroo, J., Becher, H., Kelly, Y., & McMunn, A. (2001). Children s health. In B. Erens, P. Primatesta, & G. Prior (Eds.), Health Survey for England the health of minority ethnic groups 99. London: HMSO. O Connor, T. G., Davies, L., Dunn, J., & Golding, J., and the ALSPAC [Avon Longitudinal Study of Parents and Children] Study Team (2000). Distribution of accidents, injuries, and illnesses by family type. Pediatrics, 106, e68. OPCS [Office of Populations Census and Surveys] (1991). Standard Occupation Classification. London: HMSO. Peterson, L., Ewigman, B., & Kivlahan, C. (1993). Judgments regarding appropriate child supervision to prevent injury the role of environmental risk and child age. Child Development, 64, Pickles, A., Rowe, R., Simonoff, E., Foley, D., Rutter, M., & Silberg, J. (2001). Child psychiatric symptoms and psychosocial impairment: Relationship and prognostic significance. British Journal of Psychiatry, 179, 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, Rutter, M., Giller, H., & Hagell, A. (1998). Antisocial behavior by young people. New York: Cambridge University Press. Schwebel, D. C., Speltz, M. L., Jones, K., & Bardina, P. (2002). Unintentional injury in preschool boys with and without early onset disruptive behavior. Journal of Pediatric Psychology, 27, Scott, S., Spender, Q., Doolan, M., Jacobs, B., & Aspland, H. (2001). Multicentre controlled trial of parenting groups for childhood antisocial behaviour in clinical practice. British Medical Journal, 323, Silberg, J., Pickles, A., Rutter, M., Hewitt, J., Simonoff, E., Maes, H., et al. (1999). The influence of genetic factors and life stress on depression among adolescent girls. Archives of General Psychiatry, 56, StataCorp (2001). Stata Statistical Software: Release 7. College Station, TX: Stata Corporation. Wazana, A. (1997). Are there injury-prone children? A critical review of the literature. Canadian Journal of Psychiatry, 42,