Copyright 2006 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2006
To assess risk compensation and risk homeostasis theory in children.
We used a case-control study design in children aged 8 to 18 years who had an injury while participating in an activity that did or could entail the use of protective equipment (PE).
Montreal Children's Hospital emergency department from December 1, 2001, to November 30, 2002.
We interviewed consenting children and compared the reports of risk-taking behaviors in those who did and those who did not report using PE.
Main Outcome Measures
Indicators of risk-taking behavior and injury severity.
A total of 674 children presented with injuries during the study, and 394 were interviewed (235 PE users and 159 nonusers). There was no evidence of an association between indicators of risk-taking behavior and PE use after adjusting for age, sex, personality, and type of activity and no relationship between injury severity and PE use.
Results of this study provide no support for hypotheses about risk homeostasis theory among children using PE. The validity of the theory appears highly doubtful for children in this age range.
Risk homeostasis theory (RHT) has perplexed the injury-prevention field for the past quarter century. The theory can be understood by analogy with a thermostat that adjusts the heat depending on the temperature of the surrounding area, which results in homeostasis—“resistance to change or maintaining an equilibrium.” With respect to injuries, RHT asserts that everyone has an acceptable preset level of risk taking, such that when a safety measure is applied, thus decreasing risk, the preventive benefit of the measure will be offset by more dangerous or risky behavior.
The theory was popularized by Wilde1 in 1984 and has been the topic of much heated debate that continues to the present.2- 7 The debate is fueled in part by the many important implications of the theory for several aspects of injury prevention.
These aspects range from the formulation of public health policies or programs to research strategies. For example, if RHT were proved and as widely applicable as its adherents contend, preventing injuries would require fostering changes in the risk tolerance of entire populations—a daunting, if not impossible, task. For researchers determined to find new strategies for prevention, it would suggest that most such efforts are futile.
Before struggling with these issues, we need first to assess the strength of the evidence in support of RHT. Several investigators have attempted to do so and have identified a number of critical deficiencies8,9 to which Wilde and his colleagues have offered rebuttals.6,10,11 The main criticism is that convincing empirical evidence is lacking. Most supposedly supportive studies are based on secondary analyses, almost exclusively in the domain of driving and motor vehicle crashes. These analyses have been dismissed as flawed or superseded by stronger evidence.
Several related issues also have been largely overlooked. One such is whether the theory, if true, applies equally to all subgroups in the population. In particular, it is not known whether children are affected in the same way as adults. There is virtually no mention of children in Wilde's text, and only 2 possibly pertinent articles have been published. Morrongiello and Major12 found that parents appear to permit their children to engage in riskier behaviors when safety devices, such as bicycle helmets, are used. This premise, however, is an example of adult behavior and does not address the child question directly. In another study, Mok et al13 provided pilot study data from interviews with 6- to 14-year-old children that offer modest support for RHT.
This article presents the results of an extension of that pilot study. It differs with respect to the larger number of children involved and the manner in which the data are analyzed. As in the pilot study, it is based on self-reported risk-taking behaviors of injured children who did and did not use protective equipment (PE) during sports and recreational activities. The sample is large enough to permit us to detect a 2-fold or greater difference in risk-compensatory behavior between the groups.13
In Canada, the parents of all children attending a children's hospital emergency department (ED) for an injury are asked to provide details of the event. This information is recorded on a form used by the Canadian Hospitals Injury Reporting and Prevention Program (CHIRPP).14 From these reports at the Montreal Children's Hospital from December 1, 2001, through November 30, 2002, we selected children aged 8 to 18 years who had an injury while participating in an activity that did or could entail the use of PE. We then interviewed these children and compared reports of behaviors from those who did and those who did not report using any such equipment. The research ethics board of the Montreal Children's Hospital approved this study.
An interviewer (H.M.) blinded to PE use status contacted families by telephone shortly after the ED visit, explained the purpose of the study, and sought verbal consent from parents. She then asked the child a series of questions about PE (eg, helmet) use, actual behavior (eg, whether, at the time of injury, the child judged that he or she was cycling or skating faster than usual), and severity (eg, equipment damage). Type of activity was taken from the CHIRPP form and confirmed during the telephone interview. Ages and categories were obtained from the CHIRPP form and confirmed by telephone interviews (Table 1 and Table 2). For the regression models, we classified age as younger than 11 years, 11 to 14 years, and 15 to 18 years; personality as careful or one who takes chances (on the basis of the Zuckerman scale15), and activity into 5 groups: hockey, ski/snowboard, football, bicycle, or scooter.
The questions used focused on actual behavior and were intended to provide more compelling information than the hypothetical questions used in the pilot study. Because the previous form of the question could lead to bias, the new questions were based on a specified event (eg, when you were injured bicycling, were you taking more chances than usual). The interview included administration of the Zuckerman thrill- and adventure-seeking subscale.15 An individual could obtain a score between 0 and 10, with 10 indicating high thrill and adventure seeking, a proxy for risk taking. These scores were divided into 3 risk categories: 0-4, low; 5-8, medium; and 9-10, high.
We calculated that with a sample of 400 children there would be 80% power to detect a 50% reported change in behavior, assuming a PE use rate of about 50% (ie, to 25% or 75%). The number of respondents we sought was based on estimates of anticipated nonresponses, refusals, and those who could not be traced.
Using CHIRPP reports for the 10 months from December 2001 through November 2002, we identified 674 children meeting the selection criteria and successfully interviewed 394 (58.5%). The average time between the ED visit and initial follow-up was 27 days, with a maximum of 49 days. Messages were left but not returned by 169 families; 15 were wrong numbers, and 90 did not answer after 3 attempts at different times of day. Only 4 were outright refusals, and 2 were missed for other reasons. Table 1 shows that those who completed the interview differed importantly from those who did not only by where the injury occurred, activity at the time of injury, and PE use.
Initial information obtained from the CHIRPP form revealed that 325 children used PE. However, at the interview, only 234 claimed to have done so (55% used a helmet, 23% used knee pads, <1% used wrist guards, and <1% used other PE). All analyses that follow are based on the number interviewed. As shown in Table 1, 70.3% were boys, and 86.5% were 11 years or older. Nearly one third were injured while skating or playing hockey, 30.5% while skiing or snowboarding, and 15.7% while bicycling. Only 5.6% were admitted to the hospital or kept for observation in the ED; 40.1% had a fracture or dislocation, and 16.5% involved a head-neck injury. The most commonly injured body parts were the upper and lower extremities (52.0% and 23.4%, respectively). About one third of the injuries occurred on a ski hill (30.7%) or in an arena (23.4%), 19.0% on the street, and 10.9% in a park.
Table 2 provides a comparison of sociodemographic and clinical characteristics between PE users and nonusers on the basis of the completed interview sample. A significantly higher percentage of PE users were boys; they were injured in an arena or while skiing, snowboarding, skating, or playing hockey. The PE users were also more likely to be younger children. Use of PE was not, however, related to type of injury or severity (using treatment provided in the ED as a proxy for severity).
In Table 3, we compare PE users and nonusers with respect to measures of risk taking, past injuries, and proxies for injury severity. Significantly more of the PE nonusers described themselves as risk takers, but there were no differences in the Zuckerman thrill-seeking scores. The PE users were more likely to report having used a helmet previously and to have been injured previously in the same activity, although in neither case did the odds ratio exclude the null.
In Table 4, we directly test the RHT by asking PE users and nonusers specific questions about risky behavior. Both the unadjusted and adjusted odds ratios indicate no evidence of a relationship between these indicators of risk-taking behavior and PE use after adjusting for age, sex, personality, and type of activity.
In Table 5 we examined the relationship between non–PE-related injury severity and PE use. The RHT would predict that, for example, as arm injury severity in an individual wearing a helmet increases, the proportion wearing a helmet should increase (ie, with RHT we would expect more severe arm injuries as the result of helmeted riders taking more chances). Using disposition in the ED as a proxy for severity, we found no evidence to suggest such a relationship (P = .56).
Taken together, our findings provide little support for the existence of risk compensation among children in this age group. This study is one of the first attempts to test the RHT directly in children. To our knowledge, there are no other studies in the literature in which this question is addressed in a similar manner. Thus, until such time as other study results are reported, these findings are the best indication of how RHT affects children.
With a much larger sample, we may have been able to search for, and possibly find, what might be regarded as an example of effect modification. This possibility arises because we previously postulated that there might be important subgroups in any population, including children. These subgroups might consist of inherent risk takers or thrill seekers at one extreme or the exceptionally cautious at the other. If RHT were to operate, it would be most evident among a middle group, but it is uncertain how large this group might be. However, as some data in Table 3 suggest, PE users may be different from non-PE users with respect to risk-taking behavior.
We acknowledge that these findings depend heavily on the ability of children in this age group to provide reasoned responses. Conclusions also depend on the assumption that those responses are not biased. Both assumptions seem reasonable but cannot be proved with the data at hand. The question of bias is important in this context only if it operates in what is, in effect, the null direction (ie, so that children who actually take more risks when using PE report not doing so perhaps because this response is more socially acceptable).
Despite the number of respondents involved, whether the number is sufficient to fully test the hypothesis is uncertain. We estimated that the sample was large enough to detect a 25% difference in responses, assuming a base rate (ie, for non-PE users) of 50%. That estimate was remarkably accurate, and although none of the differences approached 25%, the order of magnitude was so small that we believe the conclusion is well supported.
Among the limitations that also must be considered are the statistical requirements for the logistic regression on which the results rest heavily. To the extent that we were able to do so, these requirements appear to have been met. Another statistical issue is that of multiple testing. We did not deem it necessary to adjust for repeated tests using the same data, and not doing so made it more likely for us to find a difference. Hence, we are being conservative in our interpretation of a null result.
The use of CHIRPP data poses other limitations, especially with respect to the generalizability of the results. If there were solid reasons for believing that children seen at this hospital differ in some important ways from others and that those differences might influence their responses, this must be taken into account. However, we think it highly unlikely that the children in our sample were so unusual, but it is also possible that having chosen to test the hypothesis in children who have been injured might constitute a bias. Nevertheless, it is difficult to imagine that asking these questions of children who have not been injured would result in anything more than speculation because they would not have a memory cueing/anchoring event on which to base their responses.
There were large differences in PE use when we compared CHIRPP forms with follow-up interviews. On the basis of the CHIRPP form, those missed were less likely to wear PE. If they were also less likely to take additional risks when they were injured, their inclusion would have increased the association between PE use and risk taking. On the other hand, if these individuals were more likely to take additional risks when they were injured, their inclusion would have reduced the association between PE use and risk taking.
Our 58.5% response rate raises the possibility of a selection bias. However, we conducted a sensitivity analysis by including nonresponders in the examination of the relation between PE use and injury severity and found almost identical results (ie, no statistically significant relationships).
On the basis of these results, there appears to be little, if any, basis for believing that risk compensation is a phenomenon that influences the behavior of children using PE. Although this finding does not necessarily relegate RHT to the realm of myth, at least for children it does not seem to be a significant reality. The practical and important public health implication of this conclusion is that efforts must continue to persuade children to protect themselves by whatever means possible. Our results indicate that doing so will not result in greater risk taking and more injuries.16
Correspondence: I. Barry Pless, MD, Montreal Children's Hospital, 2300 Tupper St, Room F-259, Montreal, Quebec, Canada H3H 1P3 (email@example.com).
Accepted for Publication: December 21, 2005.
Author Contributions: Barry Pless had full access to all the data in the study and takes full responsibility for the integrity of the data and the accuracy of the data analysis.
Funding/Support: Dr Hagel is supported through a Professorship in Child Health and Wellness, funded by an anonymous donor to the Alberta Children's Hospital Foundation.
Acknowledgment: We thank Sofia Bamboulas, Sébastien Dubé, MSc, and Xun Zhang, PhD, for invaluable assistance.
Pless IB, Magdalinos H, Hagel B. Risk-Compensation Behavior in ChildrenMyth or Reality?. Arch Pediatr Adolesc Med. 2006;160(6):610-614. doi:10.1001/archpedi.160.6.610