Intended for healthcare professionals


Effectiveness of bicycle helmets in preventing head injury in children: case-control study

BMJ 1994; 308 doi: (Published 15 January 1994) Cite this as: BMJ 1994;308:173
  1. S Thomas,
  2. C Acton,
  3. J Nixon,
  4. D Battistutta,
  5. W R Pitt,
  6. R Clark
  1. Epidemiology Unit, Queensland Institute of Medical Research, Herston, Queensland, 4029
  2. Australia Queensland Radium Institute, Herston, Queensland, 4029
  3. Department of Child Health, University of Queensland, Herston, Queensland, 4029
  4. Mater Children's Hospital, South Brisbane, Queensland, 4101
  5. Royal Children's Hospital, Herston, Queensland, 4029
  1. Correspondence to Dr. Acton
  • Accepted 28 September 1993


Objective : To examine the risk of injury to the head and the effect of wearing helmets in bicycle accidents among children.

Design : Case-control study by questionnaire completed by the children and their carers.

Setting : Two large children's hospitals in Brisbane, Australia.

Subjects : 445 children presenting with bicycle related injuries during 15 April 1991 to 30 June 1992. The cases comprised 102 children who had sustained injury to the upper head including the skull, forehead and scalp or loss of consciousness. The controls were 278 cyclists presenting with injuries other than to the head or face. A further 65 children with injuries to the face were considered as an extra comparison group.

Main outcome measures : Cause and type of injury, wearing of helmet.

Results : Most children (230) were injured after losing control and falling from their bicycle. Only 31 had contact with another moving vehicle. Children with head injury were significantly more likely to have made contact with a moving vehicle than control children (19 (19%) v 12 (4%), P<0.001). Head injuries were more likely to occur on paved surfaces than on grass, gravel, or dirt. Wearing a helmet reduced the risk of head injury by 63% (95% confidence interval 34% to 80%) and of loss of consciousness by 86% (62% to 95%).

Conclusions : The risk of head injury in bicycle accidents is reduced among children wearing a helmet. Current helmet design maximises protection in the type of accident most commonly occurring in this study. Legislation enforcing helmet use among children should be considered.

Public health implications

  • Public health implications

  • Bicycle accidents are a common cause of injury in children

  • In this study most injuries were caused by losing control of the bicycle and did not involve a moving vehicle

  • Bicycle helmets reduced the risk of upper head injury by 63% and loss of consciousness by 86%

  • Children should be encouraged to wear cycle helmets


Childhood bicycle injuries are one of the main reasons for presentation to paediatric emergency departments in Brisbane, Australia. Admission to hospital and death from bicycle related trauma are usually due to head injury.1 Several studies of bicycle safety helmets report an associated reduction of head injuries,2,3 and in the only case-control study the risk of head injury was significantly reduced if a helmet was worn.3 Wearing helmets became mandatory in Queensland, Australia, in July 1991. We examined the risk of upper head injury or loss of consciousness associated with helmet wearing among children.

Subjects and methods

The study was conducted between 15 April 1991 and 30 June 1992 at the Royal Children's and Mater Misericordiae Children's Hospital, the two main children's hospitals in Brisbane. Children attending the emergency departments were ascertained by specific flagging by triage staff. We also carried out daily checks of triage books, patient presentation lists, and hospital wards, as well as computer searches of hospital records. Four hundred and forty five children aged 14 years or less had bicycle related injuries during the study period. A search of death certificate files showed one death of a cyclist who had not presented to the reference hospitals and was not therefore included in the study. Data collection started two and a half months before wearing helmets became compulsory (54 children). The case group comprised the 102 children with injuries to the upper head area, the region potentially protected by a bicycle helmet, including injuries to the skull, forehead, and scalp or loss of consciousness. To determine the protective effect of helmets against loss of consciousness, a subgroup of 41 children who lost consciousness was considered separately.

The control group consisted of the 278 cyclists who were treated for injuries other than to the upper head or face. The 65 children with injuries to the face were used as an additional comparison group. Assuming a two sided hypothesis and available data on 102 cases and 278 controls the relative odds of a head injury of at least 2.0 among non-helmet wearers would be able to be detected at 95% significance with 80% power, assuming that 47% of controls wore helmets.

A self administered questionnaire was completed by the child and his or her carer. Information was recorded on the factors leading to the accident, including bicycle malfunction, riding incorrectly, poor road conditions (for example potholes), avoiding objects including pedestrians, and contact with other moving or stationary objects. The surface on to which the child fell was also recorded. The degree of damage to the bicycle was used as a proxy to assess severity of impact.3

Injuries were defined by the clinician using a standard Queensland injury surveillance prevention project form. Ownership and use of a bicycle helmet at the time of the accident were recorded. We grouped children according to the number of years of education of the most educated parent to reflect the levels of lower secondary (10 years), upper secondary (12 years), and tertiary training in Queensland.

Non-responders were followed up within three weeks of the injury. In a concurrent repeatability study, we reinterviewed a random sample of 30 subjects by telephone within three weeks of the initial self administered questionnaire. The data were found to be almost identical for all variables.

The relation of injuries to the upper head or loss of consciousness to helmet wearing and other variables was investigated by X2 contingency tests. To produce final risk estimates, unconditional logistic regression models of the log odds of injury to the upper head or loss of consciousness were adjusted for the potential confounding effects of sex, age, hospital, parental education, the main cause of the accident, contact with a moving vehicle or a stationary object, and the severity of the impact based on the repair needs of the bicycle.4

Controls could have been less likely to hit their heads in the accident.3 We therefore also compared the 102 children with injuries to the upper head (the case group) to a second control group of 65 children with injuries to the face but no concurrent injury to the upper head, as both groups had struck their head in the accident.

Brisbane is a subtropical city with a population of about 1.3 million. At the 1986 census, the workforce comprised 30% professional, paraprofessional, and managerial workers; 55% tradespeople, clerks, or sales related workers; and 14% manual labourers.5 There were about 950 000 vehicles registered on Brisbane roads in 1991. Results

Three quarters of those injured in bicycle accidents were boys. This proportion was also reflected among children with upper head injury and those who lost consciousness (table I). Age was not significantly associated with upper head injury. More than half of injuries were reported to be caused by faulty riding, and 46 resulted from a faulty bicycle (table II). Contact with another moving vehicle was reported by 31 children. Significantly more children with upper head injury (P<0.001) had accidents involving contact with another moving vehicle (table II). More injuries to the upper head occurred when the children fell on paved surfaces than on gravel, dirt, or grass (P=0.012). Bicycles belonging to children who had sustained an upper head injury were significantly more likely to require repair than those belonging to controls (P<0.001), and a larger proportion were beyond repair in the group who lost consciousness compared with the controls (P<0.001).


Demographic characteristics of 102 children with head injury (cases), the subset of 41 who lost consciousness, and 278 with other bicycle related injury (controls)

View this table:

Characteristics of bicycle accidents in Brisbane children*

View this table:

Wearing of helmets among children who had bicycle accidents*

View this table:

The crude protective effect of wearing a helmet against upper head injuries remained after adjustment for potential confounding effects. We did not adjust for the surface on which the child fell as the data merely reflected responses to contact with a stationary object. Risk of injuries to the upper head was 2.7-fold (95% confidence interval 1.5 to 4.9) higher among non-helmet wearers than among helmet wearers. For loss of consciousness the risk was 7.3-fold higher (2.6 to 20.4) among non-helmet wearers than among helmet wearers. This translates to a reduction in risk among helmet wearers of 63% (34% to 80%) for upper head injuries and 86% (62% to 95%) for loss of consciousness.

Thirty three (51%) of the 65 children who had an injury to the face but not concurrent injury to the upper head were wearing a helmet. When these children were used as a control group the reduction in risk of injury to the upper head among helmet wearers was 51% (-10% to 78%).


Helmet wearing was significantly associated with a reduced risk of upper head injury and loss of consciousness. The reduction in risk persisted after adjustment for the confounding effects of age, sex, the main cause of the accident, contact with other objects including motor vehicles, the road surface, and the severity of damage to the bicycle.

The controls used in this study were similar to the emergency room controls used by Thompson et al.3 They reported a significant increase in risk of upper head injury and loss of consciousness for those not wearing helmets compared with the population based on the emergency room control group. The point estimates in the present study were consistent with the findings for the emergency room control group of Thompson et al.3 An emergency department control group has some limitations. Although the legislation enforcing helmet wearing in Queensland did not include penalty provisions during the study, children not wearing helmets who had a minor head injury may have been less likely to present to hospital. This would underestimate the effectiveness of helmets. We believe such a bias is unlikely, particularly among children needing admission to hospital for loss of consciousness. Other sources of selection bias relating to the emergency department control group have been discussed by Thompson et al,3 but are unlikely to affect the direction of the result. For example, cyclists who had not attended hospital after striking but not injuring their head when wearing a helmet would strengthen the observed association between helmets and a reduced risk of upper head injury had they been included in the study. Similarly, reduction in the estimated risk of upper head injury would have been greater if cyclists wearing helmets who did not have head injuries were more likely than non-helmet wearers to attend hospital.

When the analysis was restricted to children who had hit their head in the accident, the reduction in risk associated with helmet wearing was similar to that found for the sample as a whole. This suggests the reduction in risk of upper head injury was not due to cases having accidents which were more likely to damage the head. We did not directly measure risk taking behaviour, an important potential confounding variable for bicycle related injuries. However, the study population was limited to children who had been injured in a bicycle accident. We surmise that this group, regardless of site of injury were more likely than the general population to be risk takers. Confounding was controlled at least in part by this restriction. In addition we adjusted for factors potentially related to risk taking including sex, type of accident, its severity, and the involvement of another moving vehicle.

Although some people claim that motorists cause most accidents involving cyclists,6 our study indicates that the oppopsite is the case for children in Brisbane. The British Standard 6863 for cycle helmets,7 (the Australian Standard8 is similar) states they are designed to protect “in the kind of accident in which the rider falls onto the road without other vehicles being involved.” These were the most commonly occurring accidents in this study.

The association between wearing helmets and reduced risk of head injury among child bicyclists is now compelling. The high level of compliance with recent helmet wearing legislation9 is likely to help reduce further the incidence of bicycle related head injury.10 The crucial question not answered by this study is whether there is a cause and effect association or whether influences leading to non-helmet wearing are associated with other risk taking behaviour. Prospective cohort studies in populations with high compliance with compulsory helmet wearing will give insight into these issues.

This study was funded by the Federal Office of Road Safety. We acknowledge the help of Joanne Lansbury and Susan Simons and advice from Professor John Pearn and Drs Thompson-R and Adelle Green.