Bicycle-Riding Circumstances and Injuries

in School-aged Children
A Case-Control Study
Yvonne D. Senturia, MD; Tonja Morehead; Susan LeBailly, PhD; Elaine Horwitz, MD;
Morris Kharasch, MD; Joel Fisher, MD; Katherine Kaufer Christoffel, MD, MPH

Objective: To identify bicycle-riding circumstances as-
sociated with bicycle-related injury among school-aged
children.
Design: Case-control.
Setting: One metropolitan emergency department and
3 suburban emergency departments.
Subjects: Consecutive sample of children aged 7 through
18 years who experienced bicycle-related trauma and con-
trol children seen for non\p=m-\bicycle-relatedtrauma
(matched for age within 1 year, sex, and area of resi-
dence [urban vs suburban]).
Methods: Parents and case children were interviewed
by telephone about the bicycle ride resulting in their visit
to the emergency department. Parents and control chil-
dren were interviewed about their most recent bicycle
ride. The survey instrument addressed the following po-
tential risk factors: helmet use, bicycle speed, road con-
ditions, riding location, bicycle condition, an adult pres-
ence, riding destination, bicycle style, and stunt riding.
Results: Interviews were completed with 47 (73%) of 64
eligible case children and 42 (69%) of 61 control children
with the following age distribution: 27 (30%) of the inter-
views were completed with children aged 7 to 9 years, 40
(45%) of the interviews were completed with children aged
10 to 14 years, and 22 (25%) of the interviews were com-
pleted with children aged 15 to 18 years. Fourteen children
(16%) were wearing helmets. There was a high degree of
agreement between parent and child responses, higher for
case children than for control children. In univariate analy-
ses, injury wasassociated with riding with other children
(vs riding alone or with adults), riding fast or slow (vs nor-
mal speed),riding a BMX-style (motocross) bicycle (vs an-
other standard or multispeed style bicycle), playing on the
bicycle (vs going to school or other purposeful or nonpur-
poseful trip), and riding only on the sidewalk (vs in the street).
More case children than control children were farther than
\m=3/4\mile (>1.2km) from home (38%vs 19%,P=.05). Mul-
tiple logistic regression identified slow riding speed (odds
ratio, 10.3;95% confidence interval, 1.6-66.8), distance from
home farther than \m=3/4\mile (> 1.2km) (odds ratio, 3.7; 95%
confidence interval, 1.1-12.5), and riding on the sidewalk
(odds ratio, 6.1; 95% confidence interval, 1.8-20.5) as in-
dependent risk factors for injury.
Conclusions: This study identifies 3 counterintuitive but
apparently strong behavioral risk factors for bicycle inju-
ries treated in an emergency department in children aged
7 through 18 years in the Chicago (Ill) area. These find-
ings will need to be confirmed in larger samples from a
wider range of locales. In addition to stressing the impor-
tance of wearing a helmet when riding a bicycle, it may
be desirable to include the findings of this study in antici-
patory guidance discussions with school-aged children.
Arch Pediatr Adolesc Med. 1997;151:485-489

15 years.4 Despite growing interest in

Editor's Note: The results of this study are not only counterintui¬ preventing bicycle-related injuries, the
tive—they are like a diagnostic test that does not make sense. Can circumstances under which nonfatal
injuries occur have not been well re¬
we have this repeated, please preferably by many investigators.
.

Some children and parents think

. .

Catherine D. DeAngeüs, MD ported.

bicycle-related injuries only happen
when children are riding far from home
Departments of Pediatrics, or doing tricks in the street. Several stud¬

Children's Memorial Medical
Center, Northwestern University,
Chicago (Drs Senturia, Kaufer
Christoffel, LeBailly and
Ms Morehead); Highland Park
Hospital, Highland Park
(Dr Horwitz); Evanston Hospital,
Evanston (Dr Kharasch); and
Northwest Community Hospital,
Arlington Heights
(Dr Fischer), Ill.
Bicycle
injuries are the single
most important cause of
head injury in childhood
and account for more than
580 000 emergency depart¬
ment visits and more than 900 deaths per
year.1 Bicycle-related injuries are a lead¬
ing cause of pediatrie head injuries requir¬
ing hospitalization.2·3 Bicycle injury rates
are highest for children aged 5 through
ies have described bicycle-riding behav¬
iors of injured children seen in emer¬
gency departments,5"7 hospitalized,4·8 or
killed,9 with bicycle-related injuries im¬
plicating "faulty riding,"3 falling,8 or rid¬
ing in the street.6 But there have been no
published case-control studies of bicycle-
riding habits of school-aged children to
allow assessment of the risk associated
with specific behaviors.

Downloaded From: http://archpedi.jamanetwork.com/ by a University of Iowa User on 06/12/2015


PARTICIPANTS AND METHODS
The study protocol was approved by the Institutional Re¬
view Board of all participating hospitals.
REFERENCE POPULATION
The reference population included all children aged 7
through 18 years who were seen in participating hospital
emergency departments between June 5,1995, and July 21,
1995, and who met inclusion criteria. These were the fol¬
lowing: case children were those seen as a result of bicycle-
related trauma in which case children were the bicycle rid¬
ers; control children were seen on the same day in the same
emergency department for non-bicycle-related trauma; and
control children were matched for age (within 1 year), sex,
and area of residence (urban or suburban as defined by a
Chicago or suburban ZIP code). Exclusion criteria in¬
cluded the following: child or primary caretaker whose pri¬
mary language was not English and family with no access
to a telephone. Participating Illinois hospitals included 1
that was urban (Children's Memorial Hospital, Chicago)
and 3 that were suburban (Evanston Hospital, Evanston,
Highland Park Hospital, Highland Park, and Northwest
Community Hospital, Arlington Heights).
CASE ASCERTAINMENT
Case ascertainment began at each participating hospital af¬
ter Institutional Review Board approval and continued
through July 25, 1995. Every week during the case ascer¬
tainment period, the research assistant (T.M.) visited the
emergency department of each hospital to review medical
records to identify age-eligible children who had been seen
with bicycle-related trauma. A medical record abstraction
document was used to collect the following information:
child's name, address, birthdate, and telephone number;
name of parent or guardian; date seen in the emergency de¬
partment; date of injury; diagnosis at the time of hospital
emergency department discharge; and disposition (whether
the child was discharged from the emergency department
or admitted to the hospital). Records from the same day in
the same emergency department were reviewed to iden¬
tify the first child with non-bicycle-related trauma who ful¬
filled the matching criteria. If no suitable match was found,
the search was expanded to the day before and the day af¬
ter the case child's visit to the emergency department. The
same medical record abstraction document used for case
children was used to collect information from emergency
department medical records for the control children.
We hypothesized that situational factors increase
a child's risk of injury while riding a bicycle, and that
identification of riding behaviors that differentiate
children who are and are not injured while riding
bicycles might help focus intervention efforts in
various ways, including increasing the ability
of anticipatory guidance to reduce injury risk. Our
study objective was to identify bicycle-riding circum¬
stances associated with bicycle-related injury among
school-aged children by comparing case and control
children.
Downloaded From: http://archpedi.jamanetwork.com/ by a University of Iowa User on 06/12/2015
RECRUITMENT AND CONSENT
A letter was sent to the parents of each potentially eligible
case and control child, introducing the study and the re¬
search assistant and explaining that the research assistant
would be calling to complete a short telephone interview
with parent and child. The letter included the telephone
number of the principal investigator (Y.D.S.) to answer any
questions parents might have, and it also gave the number
of a telephone answering machine for those who declined
to participate in the study.
DATA COLLECTION
After allowing 7 to 10 days for the family to receive the in¬
troductory letter and respond, the research assistant called
the family to confirm the family's consent for study partici¬
pation, conduct the interview with the child, and conduct the
interview with the parent. In some cases, initial calls re¬
sulted in an appointment to call back to complete the inter¬
views. After confirming consent from the parent and child,
case children were asked a series of questions pertaining to
the bicycle ride immediately preceding their visit to the emer¬
gency department. To answer the same series of questions,
control children were asked to recall "the last time [they] rode
[their] bicycle." Some additional questions pertained only to
children who were injured during the ride.
SURVEY INSTRUMENT
The survey instrument was adapted from questions used
by previous investigators in emergency department sur¬
veys of children with bicycle-related injuries.67 Variables
included the following potential risk factors: helmet use,
bicycle speed, road conditions, riding locations, bicycle
condition, an adult presence, riding destination, bicycle
style, and stunt riding.
DATA ANALYSES
Frequency distributions were performed for case and con¬
trol responses separately for children and parents, 2 Analy¬
ses were carried out to assess study sample homogeneity of
each riding variable. To determine if variables were indepen¬
dent risk factors for bicycle-related injury, riding variables
suggested by the 2 tests were entered into multiple logistic
regression, with bicycle injury as the dependent variable.
Separate cross tabulations for case and control chil¬
dren, matching parent and child responses to individual
questions, were used to assess agreement between parent
and child.
RESULTS
CASE AND CONTROL SAMPLES
A total of 73 potential case children were ascertained from
emergency department medical records. Nine were ex¬
cluded (8 with no telephone and 1 non-English-
speaking). Interviews were completed with 47 (73%) of
64 eligible case children: 3 parents refused family par¬
ticipation, 13 were unable to be reached after numerous
attempts, and in the family of 1 case child, only the par-
 ent was able to be contacted. After the exclusion of 5 con¬
trol children who were not bicycle riders, 2 who came Table 1. Univariate Analyses: Child Interviews4
from non-English-speaking families, and 1 with no tele¬
Mo. (%) No. (%)
phone, 42 (69%) of 61 eligible control children com¬ of Case of Control
pleted interviews: 6 parents refused family participa¬ Children Children Pf
tion, 12 were unable to be reached after numerous Riding companion
attempts, and in the family of 1 control child, only the Alone 16 (34) 25 (60)
parent was able to be contacted. All children ap¬ >1 Adult 4 (9) 4(9) <.05
proached for interviews agreed to participate. 2=1 Child, no adult 27 (57) 13(31)
For combined case and control children complet¬ Riding speed
Slow 11 (23) 2(5)
ing the interview, the age distribution is described in Normal 20 (43) 32 (76) <.01
categories that are most consistent with standard US Fast 16(34) 8(19)
groupings10:30% of interviews were completed with chil¬ Purpose of ride
dren aged 7 to 9 years, 45% of interviews were com¬ Playing 7(15) 1(2) <.05
pleted with children aged 10 to 14 years, and 25% of Purposeful or nonpurposeful trip 40 (85) 41 (98) J
interviews were completed with children aged 15 to 18 Riding location
Sidewalk only 31 (66) 17 (41)
years. There was no significant difference in age distri¬ Street with or without sidewalk 16 (34) 25 (59) J
<.05
bution between children interviewed and those eligible Distance from home
but not interviewed ( 2=4.36; P>.05). Same block 7(15) 13(31)
Fourteen children were wearing helmets (16%). Four =s% Mile (s1.2 km) 22(47) 21 (50) <.05
case children were involved in collisions with motor ve¬ >% Mile (>1.2 km) 18(38) 8(19)
hicles. All case children experienced blunt injuries, with Style of bicycle
Standard or multispeed 27 (59) 33 (79)
most of the injuries occurring to the upper (n=14) or <.05
lower extremities (n=16). Three of the 9 case children 19(41) 9(21) J
Helmet used
with head or face injuries had closed head injuries not No 39 (83) 36 (86) " >.05
requiring hospitalization. No case children were admit¬ Yes 8(17) 6(14) J
ted to the hospital. Twenty-nine (66%) of the 44 case chil¬ Weather
dren with adequate information had injury severity Clear 45 (96) 37 (88)
Rainy 1 (2) 3(7) >.05
scores11 of 1, indicating mild injury. Only 9 case chil¬ 1
Windy (2) 2(5)
dren had injury severity scores of 4, and none had Lighting .

injury severity scores greater than 5. Daylight 40 (85) 34(81)

Dusk 4(9) 3(7) >.05
PARENT-CHILD AGREEMENT Dark 3 (6) 5(12)
Reflective clothing
Interview responses showed a high degree of agreement be¬ No 36(77) 34(81) >.05
Yes 11 (23) 8(19)
tween parent and child; agreement was higher for case chil¬ Stunt riding
dren than for control children. The highest degree of agree¬ No 35(74) 35 (83) >.05
ment between parent and child occurred on responses to Yes 12(26) 7(17)
the following questions: wearing a helmet (94% case chil¬ Age of bicycle, y
dren; 98% control children), riding alone (94% case chil¬ <1 14(30) 9(21)
1-5 27(57) 29 (69) >.05
dren; 83% control children), weather conditions (94% case >5 6(13) 4(10)
children; 83% control children), lighting conditions (89% Bicycle needs repairs
case children; 83% control children), and medical problems No 39 (83) 32(76) >.05
(89% case children; 83% control children). The lowest de¬ Yes 8(17) 10(24) J
gree of agreement occurred on responses related to distance Problem with surface
from home (74% case children; 52% control children), kind No 28 (60) 23 (55) >.05
of bicycle (77% case children; 79% control children), and Yes 19(40) 19 (45) J
Child owns the bicycle
riding speed (83% case children; 69% control children). No 7(15) 2(5) >.05
Yes 40 (85) 40 (95) J
VARIATIONS WITH AGE Time riding this bicycle, mo
£6 21 (45) 12(29) >.05
Reported helmet use was much more likely in children >6 26 (55) 30(71) J
aged 7 to 9 years than in older children (37% vs 15%;
P<.01). Distance traveled from home, in contrast, in¬ *n equals 47 for case children and equals 42 for control children.
creased with increasing age group (P<.01). fP for the 1 test or the Fisher exact test.
\BMX is a type of motocross bicycle.
CASE CHILDREN VS CONTROL CHILDREN
analyses, injury was related to riding with other children
Univariate Analyses (vs riding alone or with adults), riding fast or slow (vs
normal speed), riding a BMX-style (motorcross) bicycle
Table 1 compares the responses of case children with those (vs another standard or multispeed-style bicycle), playing
of control children for each possible risk factor. In these on the bicycle (vs going to school or other purposeful or

Downloaded From: http://archpedi.jamanetwork.com/ by a University of Iowa User on 06/12/2015

 Table 2. Odds Ratio of Risk Factors Associated
by events around them and so lose control of their bi¬
With Bicycle-Riding Injuries*
cycles. Children riding on the sidewalk may also be dis¬
tracted by pedestrians or experience problems dealing with
uneven surfaces. Perhaps children are less careful when
Odds 95% Confidence
Factor Ratio Interval riding on the sidewalk, because the sidewalk is consid¬
ered a safer place to ride. Although our data suggest that
Speed
Normal 1.0 riding on the sidewalk decreases with age, we did not col¬
Fast 3.2 0.9-11.8 >.05 lect information to directly address whether less expe¬
Slow 10.3 1.6-66.8 .02 rienced riders use sidewalks and that it is their lack of
Riding location!
6.1 1.8-20.5 <.01
riding proficiency that increases their risk of injury. The
Sidewalk only
Distance from hornet
relationship to distance from home could be related di¬
>% Mile (>1.2 km) vs rectly to distance or time of exposure; the available in¬
£% Mile (£1.2 km) 3.7 1.1-12.5 .04 formation suggests that it is the time, not the distance
Style of bicycle per se, that is more important.12·13
BMX§ 2.4 0.07-8.4 >.05
Riding companion COMPARISON WITH EARLIER REPORTS
With adults 1.0
Alone 0.44 0.1-1.0 >.05
With other children, no adults 0.91 0.1-1.8 >.05
Data for bicycle-related injuries seen in emergency de¬
Purpose of ride|| partments from 1987 through 1989 compiled by the Na¬
Playing 4.0 0.3-47.3 >.05 tional Electronic Injury Surveillance System14 reported
Stunt riding! the place of injury as "at home" (rather than on roads)
No stunt riding 2.6 0.5-10.5 >.05 for many children younger than 10 years, referring prob¬
Obtained from multiple
ably to sidewalk and driveway injuries. Perhaps consis¬
*
logistic regression analysis, likelihood 2 equal tent with this, our study also found that riding on the
to 32.7 with 9 \, ^.001.
tRiding sometimes/always in the street is the reference group. sidewalk was a risk factor for injury.
\Three quarter mile or less is the reference group. An exposure survey from the same source found that
§BMX is a type of motocross bicycle. Standard or multispeed-style 15% of children younger than age 15 years wear hel¬
bicycle is the reference group. mets all or most of the time.14 Our study found a simi¬
\\Purposeful or nonpurposeful trip is the reference group.
istunt riding is the reference group. larly low proportion of both case and control children
reporting they were wearing bicycle helmets. This rate
nonpurposeful trip), riding only on the sidewalk (vs in is far lower than in Seattle, Wash, where a community
the street), and riding farther than 3A mile (>1.2 km) bicycle helmet campaign has been implemented.15
from home. A population-based case-control study of 109 child
bicycle injuries in Melbourne, Australia, found chifdren
Multivariate Analyses riding on the sidewalk for more than 5 km had an in¬
creased risk of injury (odds ratio, 3.1).13 Adjusting for
A multiple logistic regression model was constructed, in¬ age and sex did not change the magnitude of the asso¬
cluding variables identified above and additional vari¬
the ciation with riding on the sidewalk. Similar to our study,
ables (eg, stunt riding) identified as empirically important they found increased risk among children spending any
based on previously published work. Several new vari¬ amount of riding time in ptay rather than a purposeful
ables were constructed by combining selected nonsignifi¬ trip (odds ratio, 2.0-2.5), and only a few injuries related
cant variables, but none remained in the final model. Lo¬ to contact with motor vehicles (17%).
gistic regression identified 3 independent riding risk factors A survey of 520 children and their caretakers seen
(Table 2): slow riding speed, distance from home farther in an emergency department with bicycle-related injury
than 3A mile (> 1.2 km), and riding only on the sidewalk. found most injuries (84%) occurred less than 5 blocks from
home and that 49% occurred in the street.6 These results
COMMENT are compatible with the findings of our study. In the same

survey, 36% of case children admitted to stunt riding or

MAIN RESULTS
When riding behaviors of those injured were compared
with riding behaviors of a matched control group, a
different picture emerged than in previous studies that
looked only at injured children. Our age-, sex-, and area-
matched case and control children engaged in stunts and
had bicycles in disrepair with equal frequency. Three fac¬
tors were found to be significant: distance from home,
riding on the sidewalk, and slow riding speed.
Riding slowly was the strongest independent risk fac¬
tor for bicycle injury identified in this study. Perhaps slow
riding identifies less proficient riders who are more apt
to fall off their bicycles, or children who are distracted
going too fast, not meaningfully different from the 31%
found in our sample of case and control patients. Thus,
"risky riding" seems to be common in a significant mi¬
nority of all school-aged bicyclists, and most injured chil¬
dren were not doing stunts or going fast when injured.
METHODOLOGICAL ISSUES
Logistics and Sample Characteristics
This study's timing was based on medical student sum¬
mer research availability and the desire to capture the peak
bicycle-riding season in the Midwestern climate. How¬
ever, completing interviews in a timely manner during

Downloaded From: http://archpedi.jamanetwork.com/ by a University of Iowa User on 06/12/2015

 summer vacation proved problematic, because families
left town on holiday with increasing frequency as the sum¬
mer progressed. Any interviews not completed by the sec¬
ond week of August were classified as unreachable, be¬
cause of potential problems with recall bias if some families
were contacted within 2 weeks of their emergency room
visits and other families were contacted after a consid¬
erably longer interval.13
The case children in this series did not require hos¬
pitalization after their bicycle-related trauma, and only
3 (6%) of 47 were diagnosed with closed head injuries.
Our results are, therefore, not necessarily generalizable
to a population of children with injuries requiring hos¬
pitalization.
We suspect that BMX riding in our sample may have
been a proxy for riskier riding. A larger sample of BMX
riders would be needed to explore this further.
Parent-Child Agreement
Parents were surprisingly well informed about their chil¬
dren's bicycle riding habits. This is useful information for
future researchers, because the time and effort involved in
obtaining telephone interviews with child respondents is
often greater than that required to contact parent proxies.
Agreement between parent and child, however, does not
prove that the report of either respondent is an accurate
representation of the events, and this needs to be ex¬
plored further. There are several reasons why agreement
between parent-child pairs might have been higher in case
children than in control children. The bicycle-injury event
and associated emergency department visit were likely to
be stimuli for accurate recall and also for discussions be¬
tween parent and child regarding the circumstances of the
injury-associated bicycle ride. Further, it clearly identi¬
fied a specific ride for description, while control children
and their parents may have been remembering different
rides. Other researchers have also noted more recall bias
in control children than in case children.16
It is encouraging that there is such excellent agreement
between parent and child on whether a helmet was worn,
but discouraging that the proportion of helmet users was
so low. It is not surprising that parents might not know how
fast their children were riding; it was surprising that almost
25% did not know what kind of bicycle their children were
riding. Future studies should continue to use child respon¬
dents for the variables with lower concordance: distance
from home, kind of bicycle, and riding speed.
Study Implications
This case-control study emphasizes the need to confirm
apparent risk factors gleaned from descriptions of case
patients only. This complements earlier work concern¬
ing the importance of product exposure in assessing prod¬
uct-related injury risk.121517 This study identified 3 be¬
havioral risk factors for bicycle injuries treated in an
emergency department in a sample of 7- through 18-
year-olds in the Chicago area. These findings will need
to be confirmed in studies in a wider range of locales,
using larger samples that include more severely injured
cases. Until then, pediatricians may wish to reinforce rou-
Downloaded From: http://archpedi.jamanetwork.com/ by a University of Iowa User on 06/12/2015
tine counseling on the need for constant bicycle helmet
use via discussion of the counterintuitive findings of this
preliminary study: that risk is apparently related to side¬
walk and distant riding location and slow riding speed.
If more studies confirm speed and distance from home
as risk factors for injury, it may become important to en¬
courage parents to become more aware of where their
children are riding, as well as their riding speed.
Accepted for publication October 15, 1996.
This study was supported in part by Northwestern
University Medical School Summer Student Research
Program.
Presented in part at the Ambulatory Pediatrie
Association Annual Meeting, Washington, DC, May 10,
1996.
We thank Steven M. Selbst, MD, and Phyllis Agran,
MD, for permission to adapt their survey instruments6·7
for use in this project; the staff in the participating emer¬
gency departments for facilitating data collection; Edwin
H. Chen, PhD, for statistical consultation; Martha J.
Barthel, RN, MSN, for injury severity score coding; Jean
Gagliardi for secretarial assistance; and Kevin Christoffel
(then aged 16 years) for the question that led to the study.
Reprints: Yvonne D. Senturia, MD, 2300 Children's Plaza
Box208, Chicago, IL 60614 (e-mail:y-senturia@juno.com).
REFERENCES
1. Fleming A, ed. Fatality Facts 1994. Arlington, Va: Insurance Institute for High-
way Safety; July 1994.
2. O'Rourke NA, Costello F, Yelland JD, Stuart GG. Head injuries to children rid-
ing bicycles. Med J Aust. 1987;146:619-621.
3. Ivan LP, Chao SH, Ventureyra EC. Head injuries in childhood: a 2-year survey.
Can Med Assoc J. 1983;128:281-284.
4. Baker SP, Li G, Fowler C, Dannenberg AL. Injuries to Bicyclists: A National
Perspective. Baltimore, Md: Johns Hopkins Injury Prevention Center; 1993:13,
14, 64.
5. Acton CHC, Thomas S, Nixon JW, Clark R, Pitt WR, Battistutta D. Children and
bicycles: what is really happening? studies of fatal and non-fatal bicycle in-
jury. Injury Prev. 1995;1:86-91.
6. Selbst SM, Alexander D, Ruddy R. Bicycle-related injuries. AJDC. 1987;141:
140-144.
7. Agran PF, Winn DG. The bicycle: a developmental toy versus a vehicle. Pedi-
atrics. 1993;91:752-755.
8. Ashbaugh SJ, Macknin ML, VanderBrug Medendorp S. The Ohio Bicycle In-
jury Study. Clinical Pediatr (Phila). 1995;34:256-260.
9. Rowe BH, Rowe AM, Bota GW. Bicyclist and environmental factors associated
with fatal bicycle-related trauma in Ontario. Can Med Assoc J. 1995;152:45\x=req-\
53.
10. Christoffel KK, Scheidt PC, Agran PF, Kraus JF, McLoughlin E, Paulson JA.
Standard Definitions for Childhood Injury Research. Bethesda, Md: National
Institute of Child Health and Human Development; 1992. US Dept of Health
and Human Services publication NIH 92-1586.
11. The Abbreviated Injury Scale 1990 Revision. Des Plaines, III: Association for
the Advancement of Automotive Medicine; 1990.
12. Hu X, Wesson DE, Chipman ML, Parkin PC. Bicycling exposure and severe
injuries in school-aged children: a population-based study. Arch Pediatr Ado-
lesc Med. 1995;149:437-441.
13. Carlin JB, Taylor P, Nolen T. A case-control study of child bicycle injuries:
relationship of risk to exposure. Accid Anal Prev. 1995;27:839-844.
14. Bicycle Use and Hazard Patterns in the United States. Washington, DC: Con-
sumer Product Safety Commission; June 1994.
15. Rivara FP, Thompson DC, Thompson RS, et al. The Seattle children's bicycle
helmet campaign: changes in helmet use and head injury admissions. Pedi-
atrics. 1994;93:567-569.
16. Drews C, Greenland S. The impact of differential recall on the results of case\x=req-\
control studies. IntJ Epidemiol. 1990;19:1107-1112.
17. Senturia YD, Binns HJ, Christoffel KK, Tanz RR. Exposure corrected risk es-
timates for childhood product-related injuries. Accid Anal Prev. 1993;25:473\x=req-\
477.