Unintentional Injury in Preschool Boys With and Without Early Onset of Disruptive Behavior

Abstract

Objective: To determine subsequent risk of unintentional injury among preschool boys diagnosed with ODD, boys with comorbid ODD and ADHD, and boys matched demographically to the clinical sample; to test predictive validity of a measure of injury proneness; and to examine factors that might predict injury beyond clinic status.

Methods: Seventy-nine consecutive clinic-referred preschool-age boys and 76 demographically matched boys without disruptive behavior participated in a 2-year prospective longitudinal design. Time 1 assessment included clinical diagnosis, parent-reported injury proneness, attachment, and verbal abilities. Injury history was measured 1 and 2 years later.

Results: Clinic-referred children had more injuries than the comparison group. Children with comorbid ODD and ADHD had approximately the same injury rate as those with ODD but not ADHD. Parent-reported injury proneness was unrelated to subsequent injuries. Neither attachment nor verbal ability predicted injury significantly beyond clinic status.

Conclusions: Children with early disruptive behavior are at increased risk of unintentional injury and therefore should be considered prime candidates for injury prevention campaigns.

Unintentional injury is the leading cause of mortality for children ages 1-18 (National Safety Council, 2001) and is annually responsible for the loss of nearly one and a half million years of children's productive life (Finney et al., 1993; Routh, 1997). Children with disruptive behavior disorders and attention deficits are believed to be particularly susceptible to unintentional injuries (Davidson, 1987; Manheimer & Mellinger, 1967). Indeed, DSM-IV (American Psychiatric Association [APA], 1994) includes unintentional injury as an associated feature of attention-deficit/hyperactivity disorder (ADHD) and conduct disorder (CD). However, the empirical basis for this assertion is unclear, and data regarding the relation between psychopathology and unintentional injury are generally limited.

In 1987, Davidson published the most comprehensive review to date concerning unintentional injuries among children with externalizing behavior problems. His review concluded that children with antisocial and oppositional behavior patterns have increased risk of injury. Data for children with hyperactive and impulsive behavior problems, he believed, were inconsistent and not conclusive. In his review, Davidson emphasized that most research suffers from deficits in research design and that most findings show only small increases in risk of injury for children with disruptive behavior patterns. Most of the studies Davidson reviewed were conducted from an epidemiological perspective with large samples and imprecise measures of children's behavior patterns and injury history. No studies were available that used clinically diagnosed samples of children to compare injury rate in clinical and comparison samples. Wazana (1997) conducted an abbreviated review 10 years later and reached similar conclusions.

Since Davidson's review, empirical efforts to clarify relations between behavior disorders and unintentional injuries have been few and far between. Davidson, Hughes, and O'Connor (1988) studied a birth cohort of nearly 1,000 children in South Wales in one of the few relevant studies that considered both oppositional and hyperactive behavior problems. A behavioral interview was conducted with each child's mother at about the time the child turned 5 years old to screen for behavioral problems, including high activity, poor concentration, discipline problems, and temper tantrums. Injury history was deduced through an examination of hospital emergency rooms serving the geographic area of the study. Results replicated those described in Davidson's (1987) literature review: there was slightly increased risk of injury in those children reported in parental interview to be discipline problems, but overactivity was not a predictor of increased injury risk.

Other studies have focused primarily on hyperactivity symptoms. Davidson, Taylor, Sandberg, and Thorley (1992) prospectively studied a cohort of 1,740 6- to 8-year-old boys in London, England. Both parents and teachers completed a brief questionnaire that assessed children's hyperactivity at the beginning of the study and again 16 months later; injury history was assessed through systematic review of medical records in area hospital emergency rooms. Results matched previous work: hyperactivity was unrelated to injury both prospectively and retrospectively and in both parent and teacher reports.

More recent studies have studied clinically diagnosed children with ADHD. DiScala, Lescohier, Barthel, and Li (1998) conducted a retrospective review of hospital charts to examine patterns of injuries to children with ADHD versus those without ADHD. Results suggested that children with ADHD were at increased risk for pedestrian and bicycling accidents and had more severe injuries than other children (DiScala et al., 1998). Barkley and his colleagues found in two studies that adolescents and young adults with ADHD are at increased risk for automobile crashes (Barkley, Guevremont, Anastopoulos, DuPaul, & Shelton, 1993; Barkley, Murphy, & Kwasnik, 1996). One problem with this research utilizing clinic samples is a failure to either control for or precisely identify patterns of comorbidity, known to occur at high rates in children with ADHD, especially diagnoses of oppositional defiant disorder (ODD) and conduct disorder (Biederman, Faraone, & Lapey, 1992). In the DiScala et al. (1998) study, for example, nearly a third of the sample had comorbid conditions that could influence the occurrence of injury.

The available evidence provides preliminary support for a modest link between disruptive behaviors (e.g., disobedience and aggression) and unintentional injury history. Results are mixed for the relation between injury history and behavioral patterns of hyperactivity, impulsivity, and inattention, perhaps due to the complication of unresolved comorbidity issues. To unravel the relations among behavior problems, comorbid illnesses, and injury, one should consider the process by which children with behavior disorders might develop increased risk for injury. Although the process is likely complex and multifaceted, preliminary evidence suggests that two factors in particular might contribute disproportionately: cognitive/neuropsychological factors and social factors. Children with behavior disorders typically have deficits in neuropsychological functioning, most frequently presented as poor attentional capacity, poor cognitive judgment, and impulsive behavior patterns (Moffitt & Silva, 1988). Behavior disorders characterized primarily by oppositional/defiant and conduct problems (with or without attention deficits) are associated with striking deficits in verbal processes. Research in this area is well summarized by Moffitt and Lynam (1995), who concluded that children with conduct problems generally perform more poorly than their peers on tasks that are administered orally, require language mediation, or call for a verbal response.

Many of the temperamental and cognitive factors linked to children's injuries in nonclinical samples are consistent with deficits in these same areas of neuropsychological functioning, including the inability to engage in “ cause-and-effect” reasoning about the potential dangers of environmental hazards (Coppens, 1985, 1986), poor self-awareness of physical capabilities and limitations (Plumert, 1995; Plumert & Schwebel, 1997), poor problem-solving strategies to prevent injury (Farmer & Peterson; 1995), and temperamental impulsivity, aggression, and undercontrol (Bijur, Golding, & Haslum, 1988; Schwebel & Plumert, 1999). These findings with community samples are supported by one study with a clinical sample, in which boys with ADHD were able to identify risky situations as well as comparison boys without ADHD but anticipated less severe consequences from risky behaviors and generated fewer means to avoid injury than the comparison group (Farmer & Peterson, 1995). Again, comorbidity of behavior patterns, such as oppositionality, limits interpretation of the results of this study.

A second factor that may help explain the process by which children with disruptive behavior are at increased injury risk is the quality of the parent-child relationship. In highly conflicted or distant parent-child relationships—often found in clinic samples of children with behavior disorders (Johnston & Mash, 2001)—there may be poor parental monitoring of the child, which may result in increased child exposure to environmental hazards and other unsafe conditions (Peterson, Bartelstone, Kern, & Gillies, 1995). Clinical observations suggest that toddlers or preschool children with chronically unresponsive parents may try to regulate the parent's attention and proximity by engaging in risky behaviors that increase the probability of parental intervention and protective action (Lieberman & Pawl, 1990). This hypothesis is reflected in at least one proposed diagnostic system for very young children, which includes an “injury prone” subclassification of parent-child “attachment problems” (Zeanah, 1996).

Primary Questions

This study used a prospective longitudinal design to examine injury risk of clinic-referred preschool boys with diagnosed ODD or ODD and ADHD over a 2-year period. A comparison group of boys without clinical-level behavior problems was matched to the clinic sample by age, race, family constellation, and socioeconomic status (SES); these matching variables have been associated with injury liability in previous research (Matheny, 1988). The comparison group was also followed for the 2-year study period. Three questions were of primary interest:

1. Are preschool boys with clinically diagnosed behavior disorders more likely to show subsequent injury than matched comparison boys? As discussed, previous research suggests that children with diagnosed behavioral disorders will be more prone to injury than matched controls (APA, 1994; Davidson, 1987).

2. Among boys with ODD, are those with comorbid ADHD more prone to injury than those with QDD alone? As noted, there has been little previous study of the relations between specific dimensions of externalizing problem behavior and injury occurrence. Given how poorly these processes are currently understood, no hypothesis was tendered for these analyses.

3. After controlling for clinical diagnosis, do other child characteristics contribute to the prediction of subsequent injury? What might these characteristics suggest about the linkage between disruptive behavior and injury risk? As discussed, child variables other than problem behavior are likely to influence the probability of subsequent injury and may help explain the process by which children are placed at increased risk for injury. We tested the role of verbal/cognitive ability and of attachment as possible predictors of injury beyond clinical status, and therefore as factors that might help explain the process by which children with behavior disorders place themselves at increased risk for injury.

Secondary Question

Beyond these hypotheses, the opportunity to test a second, peripheral hypothesis was available. The distinction between “injury proneness' and “ injury occurrence” is frequently blurred by injury researchers. Injury proneness refers to a person's tendency to engage in dangerous, potentially injury-causing, and therefore “risky” behavior, a construct distinct from the actual occurrence of injury, which is potentially influenced by factors other than the individual's behavior (e.g., extent to which environmental hazards are present). Previous findings with children are mixed on how strongly injury proneness, as measured by instruments such as the Injury Behavior Checklist (IBC; Speltz, Gonzales, Sulzbacher, & Quan, 1990), is related to children's actual rate of injury. Some studies have reported close correspondence between the two (Potts, Martinez, & Dedmon, 1995; Potts et al., 1997; Speltz et al., 1990), while other research has found that injury proneness is more closely related to parent reports of impulsivity, undercontrol, and activity level than to injury itself (Schwebel, 2000). In these studies, the IBC has been administered either concurrently or following the injury; there has been no examination of the predictive relation between the IBC and subsequent injury. Thus, we examined in both the clinic and comparison groups the extent to which injury proneness, as reported by parents, predicted boys' subsequent injuries.

Method

Participants

Parents and children were drawn from a longitudinal study of 172 preschool boys with and without clinically diagnosed oppositional defiant disorder and other disorders (see Coy, Speltz, DeKlyen, & Jones, 2001, for details). This research was primarily designed to test hypotheses about the influence of attachment and neuropsychological variables on the etiology of early-onset disruptive behavior disorders in boys. However, the inclusion of injury measures at the outset of the study allowed us to achieve the secondary goal of investigating relations among early psychopathology and injury risk in young children.

Participants in this study were 155 preschool-age boys and their families. The preschool years are among the most vulnerable for children's unintentional injury (Matheny, 1988). Seventy-nine of the boys were consecutive referrals to an early childhood behavior problem clinic, and 76 were normally developing boys without clinically significant disruptive behavior. This group of 155 was selected from the larger sample of 172 on the basis of having complete injury data for the entire 2 years of this project (i.e., excluding 17 families for whom complete injury data were missing, either because of dropout or failure to obtain all relevant information; analyses comparing these 17 with the 155 found no differences on the measures reported here). Boys' ages ranged from 47 to 68 months (M = 57 months) when they were first assessed (Time 1). Participants were evaluated again 1 year (Time 2) and 2 years (Time 3) following the initial evaluation. Mean ages were 69 and 82 months for Time 2 and Time 3, respectively. Family socioeconomic level was 2.3 on the 5-point Hollingshead scale (1 = professional, 5 = unemployed; Hollingshead, 1975). Mothers identified 82% of the sample as European-American; the remainder was African American (3%) or children of mixed ethnicity (15%). Children of mixed ethnicity typically had one European American parent with an Asian American, African American, or Native American partner. Most boys in the sample (67%) lived with both biological parents; the remainder lived with single mothers. Table I shows demographic variables by group (clinic or comparison).

Recruitment

Clinic Families. The Preschool Families Clinic was established as part of an outpatient child psychiatry clinic within an urban children's medical center. Among the clinic boys, 67% were referred by a health care professional (physician, psychologist, nurse), 23% by an educator (preschool teacher or daycare staff), and 10% were self (parent) referrals. This group differed from boys seeking services from similar clinics in that families received $50 reimbursement for completion of the assessment, agreed to be available for follow-up assessments 1 and 2 years later, and were asked questions about their sons' behavior, out-of-home care, and home environment in order to determine eligibility. Criteria for clinic group membership in the larger study included (1) unsolicited referral for disruptive, oppositional, and/or aggressive behavior; (2) age between approximately 4 and 5.5 years; (3) a T score of 65 or above on the parent version of the Child Behavior Checklist Aggression scale (CBCL; Achenbach, 1991); (4) a primary diagnosis of oppositional defiant disorder (ODD), with or without other Axis I disorders (e.g. anxiety, ADHD, encopresis/enuresis); (5) participation in out-of-home daycare or preschool (in order to obtain the reports of a nonfamilial, adult caregiver); (6) in cases of single parent families, that the mother was the primary caregiver in residence; and (7) child was free of an identified language, cognitive, motor, major health, or neurological disorder. Only boys were included in the sample because ODD occurs primarily among boys, girls with early-onset disruptive behavior are less common than boys and therefore difficult to recruit, and the factors associated with girls' disruptive behavior problems may differ from those for boys (Greenberg, Speltz, & DeKlyen, 1993). Boys also have moderately higher risk of injury than girls (Morrongiello, 1997). Only children with ODD were included in the sample because this best met the goals of the primary hypotheses of the original study (see Coy et al., 2001); no boys with ADHD but not ODD were in the sample.

Comparison Families. Comparison families were recruited through posters placed in pediatricians' offices, daycare centers, public libraries, and social service facilities. After telephone screening, parents completed child behavior checklists and family information forms to confirm the absence of significant behavior problems and to permit case matching to enrolled clinic families. Eligible comparison families were then called and a visit was scheduled. Comparison families were also reimbursed for their participation in the study.

Families of comparison boys were invited to participate in at least the first evaluation session if they met the criteria outlined for clinic boys, with three exceptions: (1) unlike clinic children, comparison boys were invited to participate in the study if they received a T score below 65 on the CBCL Aggression scale; (2) they were characterized by parents and teachers as typical children; and (3) they could be case-matched to a clinic subject with respect to age, ethnicity, family structure, and SES. There were 12 eligible clinic families for whom exact matches could not be found, due to statistically infrequent combinations of matching variables (e.g., it was difficult to find exact matches on all four matching variables when remarried parents had differing ethnic backgrounds). To maximize power, we included these clinic families in these analyses. These “ unmatched” clinical cases did not differ from matched clinic cases on any of the case-matching variables (i.e., age, ethnicity, family structure, and family SES). Comparison boys were retained in the study if they met the further criterion of having no diagnosable psychiatric condition (determined by a diagnostic interview given during the first evaluation session).

Procedure

Once eligible participants were identified, questionnaire packets were sent home and the family participated in clinic visits that included a diagnostic interview with parents, testing, observation of parent-child interactions, and teacher behavior checklists. Families were seen in the clinic at Times 1, 2, and 3. Only findings from measures relevant to injury status are included in this report (see Greenberg, Speltz, DeKlyen, & Jones, 2001; Speltz, DeKlyen, & Greenberg, 1999, for findings regarding other measures).

Time 1 Measures

DSM Diagnosis. DSM-III-R diagnoses were determined following a diagnostic interview, review of behavior checklist information from parents and teachers, and observations of boys during neuropsychological testing and parent-child interactions (DSM-IV was not yet available). A consensus diagnosis based on all available information was assigned during a team meeting including the interviewing clinician and at least one licensed psychologist or psychiatrist on the research team. The diagnostic interview incorporated most of the parent version of the Diagnostic Interview Schedule for Children (DISC; Fisher, Wicks, Schaffer, Piacentini, & Lapkin, 1992). As the DISC was designed for children ages 6 years and older, the Disruptive Behavior Disorder portion of the DISC was modified in several ways to better accommodate children slightly younger than that cutoff: (1) questions were modified to eliminate references to activities or events specific to older children (e.g., homework); (2) parents were prompted to consider whether behaviors such as activity level and short attention span were excessive for the child's age; and (3) examples of child behaviors and context were revised to reflect age-appropriate possibilities (e.g., noncompliance when asked “to put blocks away”). Team decision rules regarding diagnosis also considered the young age of the patients. Evidence for the functional impairment of symptoms was required (as in DSM-IV) and distinctions between normal development and maladaptation were emphasized (e.g., ADHD was evaluated in light of the expected high rate of motor activity and distractibility in 4-year-old boys). In assessing for CD, the symptom “often initiates physical fights” required a pattern of deliberate confrontation and initiation of physical fighting; occasional shoving or hitting that occurs commonly in preschools was not counted, nor was hitting of a parent or sibling during a tantrum. All other disorders were assessed by first asking parents screening questions taken from the DISC (parent version) and then, when indicated, with a semi-structured clinical interview (reviewing all relevant DSM-III-R criteria).

A random selection of 30 cases at intake and 24 cases at follow-up were coded by a second clinician using videotaped DISC interviews. At intake, agreement for ODD diagnosis was 94% (κ = .88), 97% for ADHD (κ = .91), and 97% for CD (κ = .78). At follow-up, agreement for ODD was 96% (κ = .78), 96% for ADHD (κ = .83), and 96% for anxiety disorders (κ = .65). Agreement for CD and mood disorders at follow-up was 100% (κ = 1.00).

Injury Proneness. Mothers completed the Injury Behavior Checklist (IBC; Speltz et al., 1990), a 24-item scale designed to rate children's display of specific injury-prone behaviors over the previous 6 months. Each item is answered on a 5-point scale. Sample items include “Runs out into the street,” “ Falls down,” and “Jumps down stairs.” Internal reliability is good (alpha =.87; Speltz et al., 1990). Convergent validity with related measures such as self- and teacher-report of risk-taking and sensation-seeking has also been demonstrated (Potts et al., 1995).

Verbal-Cognitive Ability. In a previous study (Speltz, DeKlyen, Calderon, Greenberg, & Fisher, 1999), all of the boys in this sample were administered a battery of cognitive and language measures, including a short form of the Wechsler Preschool and Primary Scale of Intelligence-Revised (WPSSI-R; Wechsler, 1989); the Expressive One-Word Picture Vocabulary Test—Revised (EOWPVT-R; Gardner, 1990); the Peabody Picture Vocabulary Test (PPVT; Dunn & Dunn, 1981); the Test of Early Reading Ability-2 (TERA-2; Reid, Hresko, & Hammill, 1989) and the Developmental Test of Visual Motor Integration (VMI; Beery, 1989). Speltz, DeKlyen, Calderon, et al. (1999) used confirmatory factor analysis to test a model of neuropsychological development in this data set that included verbal, visual-motor, and executive function domains. Goodness of fit indices indicated that this model provided an acceptable fit to the data (e.g., Bentler-Bonnett Normed Fit Index = .94, Comparative Fit Index = .94). The verbal ability factor was used in this study to examine the hypothesized relation between verbal skills and injury. This factor included the WPPSI-R Arithmetic and Comprehension subtests, EOWPVT-R, PPVT, and TERA-2. In univariate tests conducted by Speltz, DeKlyen, Calderon, et al. (1999) with this sample, the clinic group was found to perform more poorly than the comparison group on this factor, F(1, 155) = 37.8, p < .001.

Quality of Child's Attachment to the Mother. The Preschool Attachment Assessment System (PAAS; Cassidy & Marvin, 1992) was used to assess the quality of the child's relationship, or attachment, to the mother. The methodology and analyses of these data in this sample of participants are fully reported by Speltz, DeKlyen, and Greenberg (1999), who also include a full review of the concurrent and predictive validity of the PAAS and similar coding methods. Briefly, the PAAS was applied to videotaped clinic observations of parent-child separations and reunions, resulting in a 5-category attachment classification and two ratings of the child's verbal and nonverbal interactions with the mother during the reunion episode: (1) “secure” attachment, as indicated by comfortable, relaxed, and positive approach to and interaction with the mother and (2) “ avoidant” attachment, as indicated by remote neutrality with or avoidance of the mother. As these ratings were significantly (inversely) correlated, we used only the avoidance ratings in this study (classification data were not used due to small attachment subgroups within the comparison group). Higher ratings indicate greater avoidance. Each mother-child pair was classified by two coders who were unaware of children's clinic or injury status. Four coders were involved in the coding; all were certified as reliable on standard reliability tapes by Cassidy and Marvin (1992). Interrater reliability for avoidance ratings was r = .71, p < .001. Speltz, DeKlyen, and Greenberg (1999) found that the clinic group showed significantly greater avoidance during reunions with their mothers than the comparison group (clinic boys also showed lower security ratings and higher rates of insecure attachment classifications).

Measures at Time 2 and 3

When families returned to the clinic 1 and 2 years after intake (Times 2 and 3), parents were asked to provide detailed information about any injuries their children experienced since the last clinic visit. Injury reports were included only if the injury required direct examination by a medical professional (i.e., a physician, dentist, or nurse). Children's scores from both years were summed to form a single index of injury occurrence over the 2-year period following the intake.

Results

Intercorrelations Among Variables

Table II shows the intercorrelations among injury and predictor variables for each group separately (clinic and comparison). In neither group was there a significant relation between children's injury proneness (as reported by parents on the IBC) and subsequent injury occurrence; indeed, these bivariate correlations were very near zero in both groups. In the clinic group only, there was an inverse relation between verbal ability and injury occurrence (r = -.32, p < .01) and a positive relation between avoidant attachment ratings and injury occurrence (r = .22, p = .05). No significant bivariate relations were found in the comparison group.

Injury Occurrence and Injury Proneness by Group and Diagnosis

Table III shows the number of injuries by group and by diagnosis within the clinic group (ODD with or without ADHD). A total of 33 boys reported injuries over the 2-year study period. Nearly all injured boys experienced a single injury (30, or 19.4% of the entire sample). Only three boys experienced two injuries (1.9% of the entire sample; all were clinic boys), and none experienced more than two injuries. Given the highly skewed distribution of injury frequency, we consolidated these data to form a two-category injury variable for all subsequent analyses (no injuries vs. one or two).

Most of the injuries reported were lacerations (14 in clinic group, 2 in control group). Other types of injuries reported included bruises/bumps (two in clinic group), broken limbs (two in control group), dental injuries (three in clinic group, one in control group), objects in the nose or ear (two in clinic group), dislocations (one in clinic group), animal bites (one in clinic group), concussions (one in control group), sprains (one in control group), and eye injuries (one in control group). For five injuries (three clinic group, two control group), insufficient details were available to identify the type of injury incurred.

As shown in Table III, the rate of injury in the clinic group (29%) was over twice the rate in the comparison group (13%), a statistically significant difference (x² [1] = 5.89, p < .01). The clinic group also had higher mean IBC scores (M = 33.9, SD = 13.3) than the comparison group (M = 17.0, SD = 9.0, t [151] = 9.12, p < .01).

Within the clinic group, boys with ODD alone had a slightly higher rate of injuries (33%) than those with ODD and ADHD (25%); however, this difference was not significant (x² [1] = .54, p = .46).

Contribution of Verbal Ability and Attachment to Prediction of Injury

Logistic regression was used to examine the contribution of verbal ability and attachment to the prediction of injury, after controlling for clinical status (i.e., clinic or comparison group membership). The dependent variable was injury occurrence (none vs. one or two). A two-step model was tested: in the first step, only clinic status was entered; in the second, verbal ability and attachment were entered together. A classification cutoff score of.20 was used, based on the rate of injury obtained in this sample. As anticipated by the univariate analysis reported, clinic status was a significant predictor of injury (x² [1] = 6.03, p < .05), correctly predicting 70% of those with injuries, but only 54% of those without (overall prediction accuracy was 57%). When avoidant attachment and verbal cognitive ability were entered on the second step, the model was also significant (x² [3] = 12.43, p < .01). Verbal ability was the only significant univariate predictor in the final model (R² = -.15, Wald [1] = 5.43, p < .05). This model correctly predicted 67% of boys with injuries and 62% of those without (63% overall).

Discussion

Results demonstrate that boys with clinically diagnosed behavior problems have an increased rate of injury compared to those without clinical diagnoses. Boys with ODD had twice the number of injuries over the two years following diagnosis than the matched comparison group; the presence of comorbid ADHD did not appear to increase or decrease injury risk.

Injury Rate in Clinically Diagnosed and Nonpathological Boys

Clinicians have long suspected that clinic-referred children with disruptive behavior are at increased risk of injury (APA, 1994; Dunbar, Wolfe, & Rioch, 1939; Krall, 1953). This study confirms that belief empirically: clinically diagnosed children had a significantly higher rate of injury than matched controls.

We did not find a significant difference between injury rate of boys diagnosed with ODD and those diagnosed with comorbid ODD and ADHD. One might suppose comorbid ADHD would increase injury risk because of the increased rate of hyperactive and impulsive behavior. However, previous findings from literature reviews (Davidson, 1987; Wazana, 1997), empirical studies with nonclinical samples (e.g., Bijur et al., 1988; Davidson et al., 1988, 1992; Manheimer & Mellinger, 1967; Petridou et al., 1998), recent reports in the interntional (i.e., abusive) injury literature (Ford et al., 1999, 2000), and a study of driving safety in adolescents and young adults with behavior disorders (Barkley et al., 1993) support the possibility suggested by these data that hyperactivity and ADHD symptoms may be less important than oppositionality and ODD symptoms as predictors of injury. Studies that contradict this possibility have reported elevated injury rates in children with ADHD, but those studies generally focus on injuries that are the result of poor decision making, poor motor control, and poor attentional capacity—most frequently pedestrian and bicycling injuries in children (DiScala et al., 1998) and motor vehicle crashes in adolescents and young adults (Barkley et al., 1996). They are also limited by high rates of comorbid illness in the samples.

Other Characteristics That May Contribute to Injury Risk Beyond Clinic Status

We tested whether verbal ability and parental attachment might predict injuries beyond clinic status. Although we felt it reasonable to hypothesize poor verbal ability or poor attachment to the parent might predict injury proneness beyond clinic status, regressions found that clinic status alone predicted injury as well as a model that also included verbal ability and attachment. In both cases, the model predicted about 70% of the boys with injuries over the subsequent 2 years. Verbal ability was the strongest predictor when it was included in the model, but its power of prediction was not significantly greater than the contribution of clinic status alone.

From an applied perspective, one might interpret these results as an indication that the predictive power of clinic status and verbal ability overlap greatly, enough so that one “substitutes” for the other in a predictive model. Thus, one could use either clinic status or verbal ability to predict children at risk for injury with reasonable accuracy. Since identification of a child's clinic status is less costly than individualized psychometric testing, we suggest a focus on the former as a means of identifying candidates for injury prevention campaigns; such identification will generally use liberal criteria for program inclusion.

From a theoretical viewpoint, one might interpret these results as an indication that poor verbal abilities help explain the process by which children with disruptive behavior disorders have increased risk for injury. Although further research is needed, it is reasonable to hypothesize that the cognitive and neuropsychological deficits known to be present in children with disruptive behavior disorders (Moffitt & Silva, 1988), and that also have been linked to increased injury rate in nonclinical samples (Coppens, 1986; Plumert, 1995), might mediate the relation between disruptive behavior patterns and increased rate of unintentional injury (see also Farmer & Peterson, 1995).

Measures of “Injury Proneness” as Predictors of Injury Rate

Injury proneness, as measured by the parent-report IBC, had near-zero correlations with subsequent injury history, as measured by the number of visits children had to a medical professional over the subsequent 2 years. Previous reports of concurrent validity found the IBC was related to previous injury history (Potts et al., 1995, 1997; Speltz et al., 1990), but this was the first test of the IBC's prospective validity with future injuries.

One possible explanation for this finding is that our measure of injury history was invalid. Although parent report of major injuries is among the most frequently used techniques to measure injuries, it is subject to reporting biases. A more likely explanation for the weak correlation between IBC scores and injury reports is that the parent completing the IBC is biased by the child's previous injury history when completing the measure. In other words, previous injuries may influence the parent's reporting of the child's injury-prone behaviors. Of further import is the fact that the IBC is designed to measure injury proneness rather than actual injuries. Some children might be injured more frequently because of environmental hazards, parenting issues, or other external factors even if they do not engage in injury-prone behaviors, as assessed by the IBC. Other children might be injured less frequently because of close parental supervision and monitoring and relatively safe environments, even though those children engage in dangerous behaviors frequently as measured by the IBC.

Limitations

Several limitations of this study deserve comment. First, because the data were collected with the primary purpose of examining etiological factors for early-onset conduct problems, boys with ADHD but not ODD were not recruited. Attempts to replicate these findings should clearly include this group of clinically diagnosed children to further clarify the contribution of comorbidity to injury occurrence in clinical samples. Second, the sample included only boys. Although boys are at increased risk for both disruptive behavior problems and for unintentional injuries, it is unknown whether results might generalize to girls. Third, injury history was assessed only through a measure of injuries requiring professional medical attention. Future research would benefit from inclusion of an assessment of minor injuries.

Finally, this was an opportunistic study, taking advantage of injury-relevant measures in research primarily designed to address other issues. Many of the findings are therefore tentative, particularly our examination of verbal-cognitive ability and attachment as factors that mediate the relation between behavior disorders and injury. Further investigation of these factors is needed, as is investigation of other neuropsychological, cognitive, social, and temperamental factors that may elucidate the process by which children—those with behavior disorders and ultimately all children—experience unintentional injury.

Implications for Injury Prevention and Clinical Practice

Unintentional injuries are caused by countless independent factors, and no single intervention will prevent all of them. But injury remains the leading cause of death for children (National Safety Council, 2001), and any intervention that might prevent a substantial number of injuries is worth pursuing. These data indicate that children with ODD, with or without ADHD, are at a greatly increased risk of injury (nearly a third of children with ODD experienced at least one injury requiring professional medical attention over the 2-year study period) and therefore should be targeted for active intervention. We recommend that children with clinic-referred early-onset behavior problems be routinely screened for injury history and liability. Just as pediatricians routinely enact primary prevention toward children at increased risk of disease, child clinicians should enact primary prevention with behavior-disordered children who are at increased risk for injury. Prevention techniques might include education of the child and parent on safety issues, investigation and removal of hazards in the home and community, and alerting school officials to the child's increased risk of injury.

This research was supported by NIMH grant R01-MH4537 to Dr. Speltz. We thank Sharman Conner for her help with data management and manuscript preparation.

References