Richard A. Schieber, M.D., M.P.H., Christine M. Branche-Dorsey, Ph.D., George W. Ryan, Ph.D., George W. Rutherford, Jr., M.S., Judy A. Stevens, M.S., M.P.H., and Joann O’Neil, B.S.
N Engl J Med 1996; 335:1630-1635November 28, 1996DOI: 10.1056/NEJM199611283352202Abstract


In-line skating is a fast-growing recreational sport in the United States. In-line skates, with three, four, or five low-friction wheels set in a single row, afford greater maneuverability and speed than traditional quad skates, which have four wheels arranged in two rows. An estimated 22.5 million people participated in 1995, reflecting a 79 percent increase over the 1993 figure.1 During the same two-year period, the estimated number of in-line skaters injured badly enough to require emergency department care increased by 169 percent to an annual level of 99,500 (in 1995).2 The most common site of injury is the wrist, accounting for 37 percent of all skating injuries, and two thirds of wrist injuries are fractures.3,4 Wrist guards are designed to prevent sudden extreme hyperextension, to absorb shock and dissipate kinetic forces by enabling the skater to slide forward on the guard’s hard volar plate, and to prevent gravel burns. Helmets, elbow pads, and knee pads are designed to absorb shock during a fall. The use of all four of these items of safety gear has been recommended,5-7 but their effectiveness, although suggested,3-5,8 had been untested. We surveyed a representative sample of injured in-line skaters who received emergency department treatment in order to identify behavioral and environmental risk factors for injury, the mechanism and nature of injury, and the degree of protection afforded by safety gear.


Sample Design and Data Acquisition

Subjects were drawn from a sampling frame consisting of all patients injured during the study period who were wearing in-line skates and sought treatment at 1 of the 91 hospital emergency departments participating in the National Electronic Injury Surveillance System (NEISS).9 NEISS is a national probability sample, stratified into four levels according to the annual number of emergency visits, of randomly selected U.S. hospitals with 24-hour emergency departments. The study period had two parts: December 25, 1992, through April 7, 1993, and July 1 to July 31, 1993.

We developed a list of possible risk factors for injury by conducting focus-group discussions with skaters10 and by interviewing novice and experienced skaters. We constructed a 20-minute questionnaire to gather information on the characteristics of injured skaters, the circumstances surrounding the injury, and the use of safety gear.

Injured skaters were interviewed by telephone by Dr. Schieber during August 1993 after providing informed consent. If the skater was a minor, consent was obtained from the participant and a parent. Children were interviewed directly (although a parent was present in a few instances); the parent was then interviewed privately to verify the child’s reported level of experience, the account of the injury, and the use of safety gear at the time.

The study was conducted with approval from the Office of Management and Budget, which authorized the Consumer Product Safety Commission to conduct confidential, selective telephone surveys of persons with product-related injuries to learn of the interactions among the product, the victims, and the environment.

Definitions and Analytic Design

Subjects were assigned to case and control groups for our analyses of the relative benefits of the use of wrist guards, elbow pads, knee pads, and helmets. To enhance precision, we defined four risk groups according to the four anatomical sites shielded by the various types of safety gear: the wrist, elbow, knee, and head (or face). Subjects were classified in a particular risk group if the given anatomical region had been exposed to injury in the subject’s fall. For example, subjects were included in the wrist risk group if they fell on one or both outstretched arms or otherwise struck the wrist or lower arm during the fall. Within each risk group, we defined as case patients those skaters who injured the specified anatomical site seriously enough to require medical attention. The controls were skaters in the risk group who did not injure the specified site, although they did injure another part of their body. The main exposure variable was the self-reported use (or nonuse) of the appropriate safety gear for the specified anatomical site. The outcome variable was the presence or absence of an injury to that anatomical site, as reported by the skater and confirmed by the NEISS record.

Selected characteristics of the skaters were considered as covariates. The subjects were classified according to age into groups: elementary-school children (6 to 12 years), junior or senior high-school youth (13 to 18 years), and adults (19 years or older). Subjects comfortable standing up on skates were classified as novices; beginners were those who could make a simple turn on two skates; intermediates could perform a crossover turn (made with a sequence of short, alternating cross-steps); and experts could do two or more stunts. In-line skating experience was classified according to the number of times the skater had participated in the sport. Ice hockey, ice skating, alpine skiing, roller skating, bicycle racing, and mountain bicycling were defined as cross-training sports. Lessons were defined as prior training with an instructor or accompaniment by a more skilled skater. The levels of perceived exertion at the time of injury, as assessed by the skaters themselves, were warming-up, cruising, and fatigued. Other covariates in the analysis were sex, handedness, and whether the skater performed tricks (as a covariate, considered a proxy for risk-taking behavior). Characteristics of the fall that were considered as covariates included the subject’s reported inability to stop just before injury and the severity of injury. A fracture, a dislocation, or an internal injury constituted a major injury; a laceration, abrasion, contusion, sprain, or strain was a minor injury.

We evaluated the possibility that selection bias might result from the use of controls who also were injured while skating. For example, many subjects treated as case patients in the analysis of wrist injury were used as controls for the analysis of elbow injury. If nonuse of wrist guards was highly correlated with nonuse of elbow pads, then these subjects might be less likely than the general population to wear gear and have protection. Therefore, separate covariates were defined for each type of gear used as well as for a composite variable indicating the use or nonuse of any additional safety gear not related to the injury site.

Statistical Analysis

All data reported have been weighted according to the probability of hospital selection from the four levels based on hospital size in the total NEISS sample. All analyses were conducted with SUDAAN statistical software to account for complex sampling.11 Log-linear chi-square tests were used to test for association between groups and for significant differences in proportions.11,12 Association was assessed with unadjusted odds ratios and 95 percent confidence intervals. Logistic regression was used to compute adjusted odds ratios with control for confounding factors. Population-attributable risks were calculated with established methods.13 A P value of less than 0.05 (two-tailed) was considered to indicate statistical significance, with the use of variances derived from the weighted sample.


Demographic Characteristics of the Skaters and Circumstances of the Injuries

We interviewed 161 of 206 eligible subjects (for a 78 percent response rate14); their data were weighted, on the basis of hospital selection, so as to represent the distribution of characteristics of the estimated 6331 in-line skating injuries treated in emergency departments nationally during the study period. The skaters ranged from 6 to 59 years of age (mean, 20.8; median, 15). Forty-eight percent were male, and 89 percent were right-handed (other characteristics of the skaters are provided in Table 1TABLE 1
Selected Characteristics of Injured In-Line Skaters in the National Sample.
; data on their falls and consequent trauma are provided in Table 2TABLE 2
Characteristics of Falls and of Injuries Sustained by In-Line Skaters.
). The most typical fall involved young novice or beginner skaters wearing little or no safety gear, who either spontaneously lost their balance while skating outdoors or fell after striking a road defect or debris. The falls typically occurred on outstretched arms without any attempt to stop. The wrist was the most common site of primary injury (32 percent); 25 percent of all injuries were wrist fractures. Seven percent of the injured skaters wore all four types of safety gear; 46 percent wore none. Of the total, 45 percent wore knee pads, 33 percent wrist guards, 28 percent elbow pads, and 20 percent helmets.

The Case–Control Study

Case patients and controls in the wrist risk group and in the elbow risk group are compared in Table 3TABLE 3
Selected Characteristics of the In-Line Skaters with Wrist or Elbow Injuries and Their Controls.
. The projected national total in the wrist risk group (5404 persons) was estimated on the basis of weighted data on 141 interviewed subjects; the projected total in the elbow risk group (4708) was estimated on the basis of 122 subjects. Case patients and controls in the wrist group differed significantly in age distribution (case patients tended to be younger) and in their use of wrist guards and elbow pads. Case patients and controls in the elbow group differed only in their use of elbow pads, the number of lessons taken (case patients tended to have taken more lessons), and history of trick skating (controls had more frequently performed tricks).
Analysis of crude odds ratios indicated that among those at risk for wrist injury, skaters who did not wear wrist guards had a likelihood of actually sustaining a wrist injury that was 12.9 times that among those who did wear wrist guards (95 percent confidence interval, 4.5 to 37.1) (Table 4TABLE 4
Odds Ratios for Injuries to In-Line Skaters Not Wearing Appropriate Safety Gear.
). Among skaters at risk for injury to the elbow, those not wearing elbow pads had a risk of sustaining an elbow injury that was 8.0 times that among skaters who did wear such gear (95 percent confidence interval, 2.1 to 30.1). Nonuse of knee pads was associated with an increase, but not a significant increase, in the risk of knee injury (crude odds ratio, 2.2; 95 percent confidence interval, 0.7 to 7.2). The association between not using a helmet and sustaining a head or facial injury was also not significant (odds ratio, 0.9; 95 percent confidence interval, 0.1 to 6.8).
Bivariate analysis showed a significant association between the likelihood of wrist injury and both the use of wrist guards and age and an association between the likelihood of elbow injury and the use of elbow pads, the taking of lessons, and a history of doing tricks (for all associations, P<0.05). The relation between wrist injury and the use of wrist guards was confounded by age and sex; the relation between elbow injury and the use of elbow pads was confounded by having taken lessons and a history of performing tricks. When we controlled for these confounders, the multivariate analysis indicated that the nonuse of wrist guards was associated with a risk of sustaining a wrist injury that was 10.4 times that among skaters who did wear wrist guards (95 percent confidence interval, 2.9 to 36.9), and nonuse of elbow pads was associated with an odds ratio of 9.5 for sustaining an elbow injury (95 percent confidence interval, 2.6 to 34.4) (Table 4). The use of wrist guards was not significantly associated with injury to the elbow (data not shown).
In calculations of population-attributable risk, the nonuse of wrist guards accounted for 87 percent (95 percent confidence interval, 84 to 96 percent) of all wrist injuries. Failure to use elbow pads accounted for 82 percent (95 percent confidence interval, 78 to 95 percent) of all elbow injuries. Failure to use knee pads accounted for 32 percent (95 percent confidence interval, 27 to 79 percent) of all knee injuries.


Our analysis of several environmental and behavioral risk factors for in-line skating injuries in a nationally representative sample of skaters found that wrist guards and elbow pads afforded skaters a high level of protection, but that their rate of use among injured patients was relatively low. The degree of protection afforded by wrist guards and elbow pads is similar to the degree of protection afforded the head by bicycle helmets (85 percent) in a study of bicycle crashes.15
The NEISS data set permits the estimation of the national incidence of an injury with known variance.9 Even using unweighted data (thereby converting the probability sample to a sample of convenience), we found that the adjusted odds ratio for wrist injuries associated with the failure to use wrist guards was 5.9 (95 percent confidence interval, 1.8 to 18.8), and that for elbow injuries associated with the failure to use elbow pads, 7.9 (95 percent confidence interval, 2.2 to 27.5). To bolster the strength of our analysis, we used a single interviewer and multiple lines of inquiry, sought parental confirmation for important data, controlled the analysis for many potential confounders, sampled a wide age range, sought (and achieved) a high response rate, and increased precision by restricting the selection of case and control patients to those exposed to the risk of injury to a specific anatomical site.

Our study, however, has several potential limitations. Selection bias may have resulted from our use of emergency department patients as controls, rather than skaters drawn from the general population. Controls who have sustained a skating injury may be more similar to case patients than population-based controls would be, which would have biased our results toward the null hypothesis. However, any potential bias of this type would not have altered the public health consequences of our findings because the point estimates of the odds ratios are so large.

On the other hand, had a significant correlation existed among the use of different kinds of safety gear, our results might have been biased away from the null hypothesis. The use of more than one type of gear was not a significant confounder, either when use of a given type was examined individually or when we assessed a composite indicator (data not shown). Forcing these variables for use of more than one type of gear into the final model decreased the point estimate of the odds ratio for wrist injury minimally and increased the odds ratio for elbow injury substantially, but it did not change the significance of these ratios. However, the level of precision was reduced considerably (data not shown). To recruit geographically matched, population-based controls would have been prohibitively costly; controls from a single area only might not have had attributes (e.g., skill level) that properly matched those of the case patients; and seeking to select controls from a random national telephone survey might not have, in 1993, yielded many in-line skaters.

Information bias may be present because these data were reported by the skaters themselves. The recall period ranged from several weeks to eight months, and some respondents may not have remembered details well. Some variables, such as skating frequency, were difficult for the youngest children to quantify. Lifetime person-hours of skating experience, the ideal measure of exposure to possible injury, was crudely approximated by the number of times a subject remembered having skated. This number may not have been recalled accurately. The study lacked sufficient power to detect a significant association between knee injury and the use of knee pads (the point estimate, however, was elevated) as well as between head injury and the use of a helmet. Some potential confounders could not be assessed, including skating speed, crash forces on impact, cross-training experience with cross-country skiing, the use of poorly fitting safety gear, and household income.
The nature of in-line skating has some practical implications for the use of safety gear. The widespread availability of gear in child and adult sizes and its relatively low cost should facilitate use. Automatic braking devices, which slow the wheels or apply a heel brake without dorsiflexion of the foot, may reduce the injury rate; such devices became available after the conclusion of the study. Currently, wrist guards may not be practical for participants in roller hockey because the wrist guards interfere with the player’s ability to hold the hockey stick firmly or make a wrist shot. This limitation warrants further attention on the part of product designers, given the popularity of roller hockey and the large proportion of the skaters in our study who were injured while playing hockey.

In conclusion, we recommend that in-line skaters wear wrist guards, elbow pads, knee pads, and helmets. Although the small number of skaters who sustained a head injury did not allow us to determine the degree of protection afforded by helmets, other studies15,16 indicate that helmets that meet existing standards for bicycle helmets17 are strongly protective against head injuries in physical environments quite similar to that of skaters.

The views expressed by the authors do not necessarily represent those of the Consumer Product Safety Commission.
We are indebted to the NEISS coding nosologists from each hospital, who retrieved the information needed to contact injured persons; to the adult skaters, parents, and children who permitted interviews; and to Marcie-jo Kresnow, M.S. (National Center for Injury Prevention and Control) and instructors Tammy Kesting and Liz Miller for their assistance in developing the questionnaire.

From the National Center for Injury Prevention and Control, Centers for Disease Control and Prevention, Atlanta (R.A.S., C.M.B.-D., G.W. Ryan, J.A.S., J.O.); and the Directorate for Epidemiology and Health Sciences, Consumer Product Safety Commission, Washington, D.C. (G.W. Rutherford).
Address reprint requests to Dr. Schieber at the National Center for Injury Prevention and Control, Centers for Disease Control and Prevention, Mailstop K-63, 4770 Buford Hwy., NE, Atlanta, GA 30341.