child pedestrian injury prevention project: student results
TRANSCRIPT
Preventive Medicine 30, 179–187 (2000)doi:10.1006/pmed.1999.0622, available online at http://www.idealibrary.com on
Child Pedestrian Injury Prevention Project: Student ResultsDonna Cross, Ed.D.,*,1 Mark Stevenson, Ph.D.,† Margaret Hall, Grad. Dip. H.Sc.,* Sharyn Burns, M.P.H.,*
Denise Laughlin, M.P.H.,‡ Jill Officer, Grad. Dip. H.Sc.,* and Peter Howat, Ph.D.*
*Centre for Health Promotion Research and †Department of Epidemiology and Biostatistics, School of Public Health,Curtin University of Technology, Bentley, Western Australia 6102; and ‡Coastal and Wheatbelt Public Health Unit,
Health Department of Western Australia,
natural increase in children’s pedestrian-related risk
Dedicated to the Mem
Background. Few comprehensive pedestrian safetyinterventions for primary-school-age children havebeen developed and evaluated. This paper reports theimpact of the 3-year (1995–1997) Child Pedestrian In-jury Prevention Project (CPIPP) on a cohort of 1603children followed from age 6 to 9 years. This multicom-ponent project comprised an educational interventionfor students, their parents and teachers, and the localcommunity, as well as several environmental interven-tions. The primary aim of CPIPP was to improve chil-dren’s road-related behavior and to enhance the safetyof their road environment.
Methods. Three communities were assigned to thetreatment conditions: (1) high—education, community,and environmental interventions; (2) moderate—education intervention only; and (3) comparison (usualroad safety education). Children’s pedestrian knowl-edge and road crossing and playing behaviors wereassessed using a pre- and posttest self-report question-naire. Their self-reported road crossing behaviorswere validated using an observational schedule andbrief interview.
Results. Children in the high and moderate interven-tion groups were significantly more likely to cross theroad with adult supervision (P 5 0.013) and play awayfrom the road (P 5 0.000) than the comparison group.No differences were detected in children’s pedestriansafety knowledge between the intervention and com-
parison groups.Conclusions. While several methodological limita-tions may have influenced the study outcomes, thesedata nonetheless indicate that in the study sample theCPIPP educational intervention deaccelerated the
1 To whom reprint requests should be addressed at Centre forHealth Promotion Research, School of Public Health, Curtin Univer-sity, GPO Box U1987, Perth, Western Australia 6845. Fax: -61 892662958. E-mail: [email protected].
17
Northam, Western Australia, Australia
ory of Ernst Wynder
behavior. q 2000 American Health Foundation and Academic Press
Key Words: pedestrian injury; children; schools; par-ent education; health promotion; road safety.
INTRODUCTION
In Western Australia, pedestrian injury is the leadingcause of injury-specific death in 5- to 9-year-old children[1]. In 1995, the pedestrian fatality rate for WesternAustralian children aged 5 to 9 years (3.1/100,000 per-sons) was more than double that for Australia as awhole (1.3/100,000 persons) [2]. Of those severely in-jured, 80% experienced a head injury [3].
While injury prevention is recognized as a major pub-lic health priority in Australia and other countries [4,5],few comprehensive interventions designed to reducepedestrian injuries in children have been developed andevaluated. Of those currently available, most have notbeen subjected to rigorous evaluation, including themeasurement of long-term knowledge and behaviorchange [6,7].
The Child Pedestrian Injury Prevention Project(CPIPP) is one of the first major intervention researchprojects to evaluate the efficacy of a comprehensiveschool, home, and community education program in as-sociation with road environment changes on 6- to 9-year-old children’s pedestrian-related knowledge androad crossing and playing behavior.
The primary behavioral message of this program forstudents and parents was that children under the ageof 10 require adult accompaniment when crossing the
road. This age group lacks the cognitive and perceptualability to deal with complex traffic situations [8]. Theirphysical size and emotional maturity also limit theirroad-crossing ability.9 0091-7435/00 $35.00Copyright q 2000 by American Health Foundation and Academic Press
All rights of reproduction in any form reserved.
180 CROSS
Using the findings from an extensive formative evalu-ation, the school-based component of the CPIPP inter-vention comprised developmentally appropriate pedes-trian-related knowledge-, attitude-, and interpersonal-skills-based activities, as well as essential “real” roadpractice [9]. Parents were also actively engagedthroughout the program. These features are rarelyfound in previous school road safety programs[6,10–16].
The environmental component of the CPIPP inter-vention is based on a community development model.A community advisory committee of key road safetystakeholders and other community representatives wasformed to determine and provide road-safety-relatedcommunity education as well as recommend modifica-tions to improve the safety of the community’s roadcrossing environments, especially for children. Envi-ronmental changes implemented during the 3-yearproject (mostly in 1997) included new footpaths; saferparking/drop-off areas for parents; markings to indicatesafer crossing areas; and some traffic calming. Furtherdetails of the environmental intervention have beenreported elsewhere [17].
The Child Pedestrian Injury Prevention Project isone of the first school- and community-based programsto show that it may be possible to deaccelerate thenatural increase of pedestrian-related risk taking inchildren. The intervention is more comprehensive thanany previously published and it is one of the first studiesto rigorously validate self-report road crossing behav-
iors in 6 to 9 year olds via observation and follow-up interview.This paper describes the Project’s methods and re-sults collected from a cohort of 1603 children from Janu-ary 1995 to November 1997.
METHODS
Overview
The Child Pedestrian Injury Prevention Project wasa 3-year quasi-experimental trial developed to improvechildren’s pedestrian safety knowledge and their road-related behaviors—crossing and playing—and to re-duce their risk in and exposure to traffic. This reductionin risk was also achieved by the community advisorycommittee lobbying to modify hazards in the road envi-ronment. While not assessed, the long-term goal ofCPIPP was to reduce pedestrian injury in 6- to 9-year-old children. The Curtin University Human ResearchEthics Committee provided ethics approval for this
project.The trial involved three metropolitan communitiesin Perth, Western Australia. Two received an interven-tion, and the third acted as a comparison community.
ET AL.
Community 1 received the “high” intervention that com-prised a school- and home-based pedestrian safety edu-cation program as well as a community education andenvironmental intervention. The community/environ-mental intervention comprised the establishment of acommunity road safety committee which advocated fortraffic management strategies; changes to speed limitsaround schools; traffic calming in local streets; and aprogram to map and mark footprints to show childrensafer pedestrian routes.
Community 2 received the “moderate” intervention,which included the same school- and home-based pedes-trian education program but no community or environ-mental interventions. The comparison group, Commu-nity 3, received a nutrition education program, as wellas the standard Western Australian health educationprogram that contains several road-safety-related ac-tivities [18].
In each of the 3 study years, the school- and home-based intervention comprised nine 40-minute pedes-trian safety lessons and nine home activities. Teacherswere asked to implement these lessons and send homethe corresponding home activities in three clusters orboosters of three 40-minute lessons at specified inter-vals throughout the school year (at the beginning of thefirst three school terms when children returned fromtheir vacation breaks). The key focus of the home andclassroom activities was pedestrian skills training in areal road environment using school access and localtraffic roads. All strategies were cross-curricular anddesigned to be student-centered and interactive and toactively involve parents. To maximize teacher imple-mentation, a half-day training was conducted at thebeginning of each of the 3 study years. Teachers alsoreceived passive (user initiated) posttraining support,such as content information sent by telephone andfacsimile.
Sample
The sample comprised students, parents, and teach-ers from three Perth Local Government Areas (LGAs).Perth City is the capital of the state of Western Austra-lia and has a population of approximately 1.3 million.The three LGAs were selected based on relatively simi-lar rates of child pedestrian injury in each community(Community 1: 3.3/10,000; Community 2: 2.3/10,000,and Community 3: 2.3/10,000 licensed motor vehiclesper year) [2]. These three LGAs did, however, vary inpopulation size and area (population 80,000, 75,000,and 160,000, respectively, and an area of 12.5 km2, 1024km2, and 96.5 km2, respectively). The CPIPP communi-
ties are separated by at least one other LGA on allboundaries.The sample size estimate was based on children’sknowledge of road safety data from a case–control study
R
CHILD PEDESTRIAN INJUconducted in Perth between 1991 and 1993 [19]. In thisstudy approximately 32% of children aged 6 to 14 yearscould provide more than one correct road safety instruc-tion. Based on this value, under simple random sam-pling, comparisons between samples of 430 childreneach ensured a 90% power at a (two-sided) level of 0.05to detect a difference in the child’s pedestrian safetyknowledge of 10% or more. After adjusting for the de-sign effects of attrition and clustering (based on a classsize of 25 students and an intraclass correlation valueof 0.02), a sample of 778 (32%), 785 (33%), and 847(35%) Grade 2 children were identified as needed inthe high, moderate, and comparison groups, respec-tively. The total student sample size recruited from eachof the communities is presented in Table 1. The 1603students who responded at baseline and all three post-tests comprise the “student cohort.”
Instrumentation
Two instruments were developed to collect studentdata, one to assess students’ pedestrian-related knowl-edge and self-reported road behaviors and another tovalidate the behavioral road-crossing findings. The chil-dren’s self report questionnaire comprised 23 questions;10 measured their knowledge of pedestrian safety (e.g.,safer places to cross, adult accompaniment, trafficsearch strategies); 3 questions measured their playingbehaviors on or near the road; 3 asked about their roadcrossing behavior; 2 assessed their exposure to the roadenvironment (e.g., mode of transport, frequency of roadcrossing); and the remaining questions were related toparental instruction (e.g., “How often have your parentstalked to you about how to cross the road safely?”).
A trained project staff member administered thequestionnaire to all students in each group at baseline(May 1995) and at posttest in November of 1995, 1996,
and 1997. Because of the students’ age (6 years at base-line) and associated reading level, the questionnairesModerate-intervention group 767 702Comparison group 825 744
Total 2356 (98%) 2146 (91%
a Percentage of baseline respondents who responded at all posttestsb Percentage of baseline respondents who responded at all posttests
Y PREVENTION PROJECT 181
all items on the questionnaire aloud to students andmost questionnaire responses were illustrated.
To ensure standardization, questionnaire adminis-trators received a 2-h training for the baseline and allposttests. The questionnaires took approximately 25minutes to administer to students. The school principalprovided consent for student participation. These tech-niques are consistent with other school-based re-search [20,21].
Prior to baseline data collection, the questionnairewas pilot tested with a group of similar aged students,not included in the study. Test–retest reliability andthe face and content validity of this instrument wereassessed. Similar to other primary-school-based re-search, the reliability analysis was lower than wouldbe expected in older children or adults [22–24]. The kstatistic ranged from 0.05 to 0.7. Haas suggests that0.4 is a minimal level of acceptable concordance [25].A combination of limited variability in several itemsand the majority of students providing similar re-sponses increased the percentage of agreement bychance, which lead to a lower than desirable k for someitems. The test–retest correlation for the 10-itemknowledge index, the 3-item road crossing behavior in-dex, and the 3-item road playing behavior index (usingtotal score) were 0.42, 0.61, and 0.42, respectively.
Using the results from the measures of reliability,item difficulty, and feedback from the expert panel andthe pilot administration, a number of changes weremade to the wording, illustrations, and some responsesin the questionnaire. Further details regarding the de-velopment of this questionnaire are described else-where [26].
To unobtrusively observe a randomly selected sub-sample (n 5 80) of students’ road crossing behavior asvalidation of the self-report behavioral responses in the
student questionnaire, an observational schedule wasdeveloped and administered. This subsample of stu- dents included only those children from the longitudi-and the administration procedures were designed withnal cohort who indicated that they walked to school andsome unique features. To reduce the impact of somestudents’ poor reading levels, the administrators read whose parents gave permission for research staff to
TABLE 1
Student Response Rates—Baseline to Posttest 1997
Posttest 1 Posttest 2 Posttest 3 LongitudinalBaseline (1995) (1996) (1997) cohort
Number surveyed 2414 2356 2146 1863High-intervention group 764 700 610 535 535 (70%)a
a
601 514 514 (67%)652 554 554 (67%)a) 1863 (87%) 1603 (86%) 1603 (68%)b
by experimental group.from a total of 2356 surveyed.
E
which is nested within the school. As described above,
9%, whose names were still on class lists, were sick or
182 CROSS
follow them. The subsample was also interviewed intheir classroom immediately after their arrival atschool on the day they were observed. The interviewerasked them two questions related to whether an adultaccompanied them when walking to school and crossingthe road.
Trained administrators conducted the observationsfollowing active parent permission. The observers’ in-terrater reliability was assessed using video scenariosat the training and 3 weeks after the observation periodhad commenced. Observer agreement assessed for eachadministration period (baseline, posttest 1995, andposttest 1997) ranged from 81 to 92%.
To assess the validity of student self-reported roadcrossing behavior, these data were correlated with roadcrossing observational data for those students fromwhom observational data were collected (n 5 80). Thebehaviors of primary interest were whether an adultaccompanied the walking child to school and across theroad. Moderate correlations (0.5 and 0.52) were foundwhen observations were compared with the studentself-report questionnaire. During the classroom inter-view, the observers asked the children whether an adultaccompanied them as they walked to school andwhether an adult helped them to cross the road. Whenthe self-reported behaviors at interview data were com-pared to the observational data, correlations of 0.83 and0.8 were found, respectively. The subsample observedreflected the cohort on key characteristics, namely so-cioeconomic status, age, and sex.
To assess the potential confounder of socioeconomicstatus, each student was assigned a socioeconomic sta-tus figure. This figure was determined using the Indexof Relative Socio-Economic Disadvantage as describedby the Australian Bureau of Statistics, based on thezip code of the school students attend [27]. This indexuses data collected as part of the 1991 Census andsummarizes variables related to households’ economicresources, education, and occupation. Students typi-cally live within a 5-km radius of their school.
Statistical Analysis
A description of the demographic characteristics ofthe sample and the distribution of the dependent vari-ables at baseline was conducted. The first dependentvariable, students’ pedestrian safety knowledge, was asummation of 10 pedestrian knowledge-based itemswhich were scored 1 for a correct answer or 0 for anincorrect answer. The 3 items which measured thechild’s self-reported road, crossing behavior were sum-mated to provide the dependent variable, road crossing
behavior (0, greatest risk, to 3, lowest risk). A score of0 (greatest risk) for this dependent variable indicatedthat students were walking along and crossing roads intheir neighborhood and on their way to school withoutT AL.
adult help. The 3 items measuring the child’s self-re-ported road playing behavior were also summated toprovide the dependent variable, road playing behavior(0, greatest risk, to 3, lowest risk). The risk behaviorscore of 0 indicated that the students were playing onthe footpath, on the road, and on the driveway. For bothof these indices the score of 3, lowest risk, indicatedthat they reported taking none of the risks identified.
The dependent variables of interest were assessed todetermine any changes from baseline to 3-year follow-up between the three groups. Using the three depen-dent variables, namely knowledge of pedestrian safety,road crossing behavior, and road playing behavior, threerepeated-measures ANOVA were conducted using Time(baseline and follow-ups—1995, 1996, and 1997),Group (whether intervention or comparison groups),and Time*Group interaction as fixed effects. The ran-dom effects included Student nested within Schoolnested within Group, and Time nested within Schoolnested within Group. Using the same dependent andindependent variables, a further three models were con-ducted controlling for the potential moderators of expo-sure to the road environment and socioeconomic status.
Univariate analyses were conducted using SPSS forWindows software [28]. As the student responses werenot independent of their classrooms or schools, the mul-tivariate analyses were conducted using SAS ProcMixed [29]. This program adjusts for the intraclass cor-relation that arises when analyzing individual re-sponses when the sampling unit is not the individual.In this study, the sampling unit is the school and theindividual (student) is nested within the classroom,
the multivariate models included fixed and random-effect terms. The variable of interest, which tests forthe school intervention effects, is the Group*Timeinteraction.
RESULTS
Of the 2414 students available at baseline, 2356(98%) responded. Seven hundred sixty-four of the 2356were from the high-intervention group, 767 from themoderate-intervention group, and the remaining 825from the comparison group. Sixty-eight percent (n 51603) of baseline students (the longitudinal cohort)completed all three posttest questionnaires in Novem-ber of 1995, 1996, and 1997 (see Table 1). Of thosestudents lost to follow-up, 23% moved to schools thatwere not in the same community, and the remaining
unable to complete the questionnaire during the admin-istration and follow-up period. As is evident in Table1, the attrition of students over the 3 years was similaracross the three groups.
R
CHILD PEDESTRIAN INJUAlthough the proportion of students’ lost to follow-up over the 3 years was similar across the groups, differ-ential attrition was examined. This was assessed bydetermining whether there were statistically signifi-cant differences between the three groups across thedependent variables: pedestrian safety knowledge, roadcrossing behavior, and road-playing behavior at base-line. All analyses were adjusted for socioeconomic sta-tus. No significant differences were evident betweenthree groups for pedestrian safety knowledge (F 50.414, df 5 230, P 5 0.661), road crossing behavior(F 5 1.46, df 233, P 5 0.234), and road playing behavior(F 5 0.807, df 5 229, P 5 0.448).
The mean age of the students was the same in allthree groups, namely 6.4 years (at baseline). Whilethere was no significant difference in the sex distribu-tion between the three groups, a difference was evidentbetween the intervention groups and the comparisongroup in relation to the socioeconomic status (SES) [27].The high-intervention group comprised predominantlymiddle to higher SES suburbs, 39 and 38%, respec-tively; the moderate-intervention group was predomi-nantly lower SES, 50%; while approximately 79% offamilies in the comparison group resided in higherSES suburbs.
Of the 2356 students who responded at baseline 41%reported they walked to school or rode their bike onmost days, and 41% reported crossing one or more roadsalone each day. For both of these exposure variablesthere were no significant differences among the threestudy groups.
To avoid Type III error, three measures of classroomprogram implementation were collected using a teacherlesson log, postimplementation teacher self-reportquestionnaire, and assessment of student workbooks.
These instruments measured the proportion of the nineMean (standard error)Road playing behavior
0 (greatest risk) to 3 (lowest risk) 2.0 (0.030Meana (standard error)
a Mean not adjusted for the effects of clustering.
Y PREVENTION PROJECT 183
the program in 1995 and 75 and 59% in 1996 and 1997,respectively. The parent self-report questionnaire re-vealed that, of the total number of home activities, par-ents completed 55% in 1995, 56% in 1996, and 44%in 1997.
There were no significant differences between theintervention groups and comparison group at baselineacross the three dependent variables (see Table 2). Asno statistical difference was detected between the high-and moderate-intervention groups over time for thevariable pedestrian safety knowledge, these two groupswere pooled for the analyses and compared to the com-parison group. As is evident in Table 3, there were nosignificant differences in the mean number of pedes-trian safety knowledge items answered correctly by stu-dents in the intervention groups compared to studentsin the comparison group between baseline and 1997follow-up, although significant differences were evidentwithin both the intervention and the comparison groupsand between baseline and posttest 1995 and 1996.
The data from the high- and moderate-interventiongroups for the variable road-crossing behavior were alsopooled as no statistical differences were detected be-tween the two groups over time. A statistically signifi-cant Group*Time interaction favoring the interventiongroup was evident over the 3-year period for the depen-dent variable, road-crossing behavior (Table 3). Thepost hoc analyses revealed a decline in the child’s re-ported adult accompaniment when crossing the road,both between the groups at each follow-up and withinthe groups. However, the decline was greatest in thecomparison group, suggesting even less adult supervi-sion in the road environment.
For the variable road playing behavior differenceswere detected between the high- and moderate-inter-
vention groups over time. Hence, this variable was ana-lessons teachers implemented. These three proportions lyzed separately in the high- and moderate-interven-tion groups and compared to the comparison group. Awere averaged for each teacher/class to give a composite
measure of implementation. This aggregate score of statistically significant Group*Time interaction fa-voring the high- and moderate-intervention groups wascompleteness indicated that teachers delivered 84% of
TABLE 2
Distribution of Dependent Variables at Baseline
Baseline values
High-intervention group Moderate-intervention group Comparison groupVariables (n 5 764) (n 5 767) (n 5 825)
Pedestrian safety knowledge (score out of 10)Meana (standard error) 6.1 (0.073) 6.1 (0.072) 6.2 (0.062)
Road crossing behavior0 (greatest risk) to 3 (lowest risk) 2.4 (0.048) 2.4 (0.047) 2.4 (0.046)
a
) 2.2 (0.027) 2.1 (0.026)
Posttest 1996 1.79* 1.51*Posttest 1997 1.21 1.04
ex
r
compared with the comparison group (low exposure:
a Mean is adjusted for socioeconomic status and the child’s level ofmean, adjusted for clustering.
* Significant change (P , 0.01) between time periods, within the g
evident over time for the dependent variable road play-ing behavior (see Table 4). While the proportion of chil-dren who reported undertaking activities which placedthem at greater risk of injury increased in all threegroups (for example, playing on or near the road), theintervention groups reported fewer “risky” activitiesthan the comparison group. The moderate-interventiongroup reported the safest playing behavior. Further-more, in contrast to the intervention groups, significantdifferences within the comparison group were evidentat each follow-up period, indicating a significant declinein safe self-reported road playing behavior at each fol-low-up period.
Further analysis was undertaken to determinewhether children who had high or low levels of exposure
to the road environment differed across the three depen-Post 1997 1.46
a Mean is adjusted for socioeconomic status and the child’s level of exmean, adjusted for clustering.
† Significant change (P , 0.05) between time periods, within the gr* Significant change (P , 0.01) between time periods, within the gr
posure to the road environment. Mean reported is the least-squares
oup.
exposure (mean score 2.17 versus 1.06) and the differ-ence was significantly better in the high- and moderate-intervention groups compared to the comparison group(F 5 4.64, P 5 0.01). Children with either high or lowlevels of exposure reported similar levels of road playingbehavior. However, significantly fewer risky playing be-haviors were reported between the intervention groups
184 CROSS ET AL.
TABLE 3
Meana Change in Dependent Variables—Baseline to 1997 Follow-up
P valueIntervention groups Comparison (for Time by Group
Dependent variables combined group interaction; F value)
Pedestrian safety knowledge (score out of 10) 0.084Baseline 6.18* 6.24*Posttest 1995 7.74* 7.48*Posttest 1996 8.29 8.16Posttest 1997 8.23 7.97
Road crossing behavior (0, greatest risk, to 3, lowest risk) 0.013Baseline 2.43* 2.44*Posttest 1995 2.23* 2.04*
F 5 16.32, P 5 0.00; high exposure: F 5 34.0, P 5 0.00).Finally, no significant difference in pedestrian safetyknowledge was detected between children with highlevels of exposure and those with low levels.
DISCUSSION
Overall, it can be seen that while no significant differ-
ence was found over the 3-year period for pedestriandent variables, as well as between the groups. Theseanalyses found that children who had low levels of expo- safety knowledge, a difference was evident between theintervention and comparison groups for the two behav-sure to the road environment reported better road cross-
ing behavior than did children with higher levels of iors: crossing the road and playing on or near the road.
TABLE 4
Meana Change in Road Playing Behavior—Baseline to 1997 Follow-up
P value (for Time byHigh-intervention Moderate-intervention Comparison Group interaction,
Dependent variable group group group F value)
Road playing behavior (0, greatest risk, 0.000to 3, lowest risk)
Baseline 2.09* 2.19† 2.16†Post 1995 2.06 2.24 2.00*Post 1996 1.97* 2.12* 1.67*
1.91 1.43
posure to the road environment. Mean reported is the least-squares
oup.oup.
R
CHILD PEDESTRIAN INJUThese findings, however, are tempered by several limi-tations of the study related to its design, selective attri-tion, and instrumentation.
As the three study communities could not be ran-domly assigned to conditions, several potential sourcesof bias may have been introduced. For example, thecommunities assigned to high or moderate interven-tions may have considered road safety education a moreimportant issue than the comparison community. Whilethis limitation cannot be controlled statistically, base-line data collected from teachers in each of the threestudy groups indicate similar attitudes to road safetyeducation. The majority of teachers ranked road safetyeducation the second most important health topic fortheir students. Additionally, no significant differencesamong teachers in each of the three study groups werefound for their previous levels of road safety teaching.Further, no significant differences were found at base-line among students for any of the dependent variables.The only difference found between the groups was forsocioeconomic status. All analyses were subsequentlyadjusted to account for this difference.
Generalizability of the data is also limited by theselective attrition. While the attrition rate was rela-tively low (32% over the 3 years of the study) and ade-quately accounted for in the power calculations, morestudents lost to follow-up were likely to walk and/orcross the road unaccompanied by an adult than cohortstudents. Also, students lost to follow-up in the moder-ate and comparison groups had significantly lower pe-destrian safety knowledge. This loss may have spuri-ously deflated the program effects.
Additionally, because the observed program effectsrelate to self-reported behaviors, student responsesmay reflect their knowledge (social desirability bias)rather than actual behaviors. Further, the low reliabil-ity of the instrument suggests some measurement error.However, the behavioral change was uniformly in thecriterion direction which suggests marginal instrumentvalidity (albeit perhaps only as a measure of knowl-edge). The high association between observed crossingbehavior and the behavioral interview on the subsam-ple of students also provides support that the behav-ioral findings, at least for road crossing, reflect actualbehaviors. The moderate association between observedcrossing behavior and the self-report questionnairedata on the subsample suggests that this may not havebeen their usual behavior and may be a reason for notfinding greater intervention effects.
The environmental component of CPIPP, reportedelsewhere [18,30], provided to only the high-interven-
tion group, did not appear to have influenced observedchanges in student knowledge or crossing or playingbehavior. This result is not unexpected as it took thefirst 18 months of the 3-year study to establish theCommunity Advisory Committee and for this group toY PREVENTION PROJECT 185
decide on its plan of action. Further, by design, most ofthe strategies proposed and implemented by this com-mittee targeted community residents, the road environ-ment, and drivers, not the student cohort directly. Moretime is needed to determine the contribution of theseactivities to student road-related knowledge and behav-ior and child pedestrian injury rates.
Pedestrian Safety Knowledge
Previous pedestrian safety education programs havenot reported statistically significant differences in pe-destrian safety knowledge as a result of their interven-tions [14,21,31]. In the first 2 years of CPIPP a signifi-cant difference in pedestrian knowledge was observedbetween the intervention and comparison groups, albeitby the final year this difference had diminished. A num-ber of factors may have contributed to this lack of effectin the final year. To create a questionnaire sufficientlysimple for 6 to 7 year olds to read and understand,yet developmentally and cognitively sensitive to detectchanges in knowledge and conceptual understandingswhen these children are 8 to 9 years of age was challeng-ing. Many versions of the questionnaire were pilotedto find a balance between these two factors. An age-appropriate questionnaire that increases in difficultyfor each study year was not developed as it was toodifficult to achieve equivalence. The results suggestthat the questionnaire was too simple for students. Itmay also be true that in pedestrian safety educationthere is a “ceiling effect” whereby students reach athreshold of utility knowledge from which they thendevelop their cognitive and motor skills in the roadenvironment. Students in the intervention group mayhave reached this threshold in the first year of thestudy when significant differences between groups wereapparent (77.4% correct in the intervention groups com-pared to 74.8% correct in the comparison group). Fur-ther, while knowledge is necessary to change behavior,theorists argue that on its own, it is not sufficient. Atti-tude change, skill development, and practice are consid-ered integral (with knowledge) to achieve behaviorchange [32–34]. These results may reflect the disassoci-ation between knowledge and behavior generally ob-served in road safety [14].
Adult Accompaniment Behavior in the RoadEnvironment
As children under 10 years of age may not have therequisite cognitive and perceptual skills to cross roadssafely, the key message in the CPIPP program was thatthis age group of children be accompanied by a trusted
adult when walking along and crossing a road. Despitethe significant positive difference observed between theintervention and comparison groups for both road cross-ing and road playing behaviors, the data nonetheless
from Rollins School of Public Health, Emory University, for his helpful
186 CROSS
indicate that in all groups’ student risk taking in-creased from the first year to the third year of the study.These findings are similar to those associated withother risk behaviors such as smoking. Children typi-cally increase their risk-taking behavior as they growolder. Parents may also feel that there is less need toaccompany and supervise their children in the roadenvironment as the children get older. The results forboth of these behaviors may indicate the proportionaldifference in risk taking between the intervention andcomparison groups. Further, while this trend may bedifficult to prevent, the findings suggest that these risk-taking behaviors can be moderated to reduce possibleharm.
The observational and interview validation of a sub-sample of student self-report data provides support thatthese behavioral findings are not an artifact of socialdesirability bias or good knowledge. Additionally, theanalyses that examined the effect of exposure on roadcrossing behavior provide evidence that it was not nec-essarily a factor in safer road crossing behavior.
Road Playing Behavior
While the self-reported playing behaviors were notvalidated, the high correlation found with the crossingbehavior data suggests that self-reported behavior maybe a valid means of assessing road playing behavior inthis age group. As discussed previously, both interven-tion groups reported significantly safer playing behav-iors than the comparison group. Previous research hasfound that many children in this age group, especially
those in lower SES groups, spend a larger proportion of their time playing on or near the road environment[14]. As all analyses controlled for exposure to the roadenvironment and SES, these factors are not an alter-nate explanation.CONCLUSION
Like most community trials, CPIPP lacks the abilityto control for activities in the community, which maycontribute, indirectly, to the effectiveness of the inter-vention. Notwithstanding this, every attempt was madeto document programs and/or activities, external toours, which may have impacted the observed outcomes.There were few identified extraneous factors that couldhave explained the differences detected between theintervention groups and comparison group. Therefore,as there was equivalence of the groups at baseline;appropriate multivariate analyses undertaken; suffi-cient statistical power; good follow-up (68%); and evi-
dence that the program was implemented as planned,one can have confidence in the findings from the study.At the completion of this intervention trial the CPIPPschool- and home-based pedestrian safety educationmaterials were incorporated into a comprehensive
ET AL.
kindergarten to grade 7 school road safety educationresource called Kids and Roads and disseminatedstatewide.
ACKNOWLEDGMENTS
The authors acknowledge their colleagues, Curtin UniversityHealth Promotion students, teachers, principals, school students, andtheir parents, Steve Jones, Anna Flintoff, and the Project AdvisoryCommittee for their valued contributions to CPIPP and the contentsof this paper. Thanks are also conveyed to Professor Ken Resnicow
comments on the manuscript. The Western Australian Health Promo-tion Foundation (Healthway); Main Roads Western Australia; andthe Western Australian Office of Road Safety supported the CPIPPstudy and the preparation of the manuscript. The Project was initi-ated and analyzed by the investigators.
REFERENCES
1. Federal Office of Road Safety Australia, personal communica-tion, 1998.
2. National Injury Surveillance Unit. Injury deaths Australia1991–1995. Case numbers and rates per 100,000 population byyear of death registration, age and sex. Adelaide: NISU, 1997.
3. Roberts I. Injuries to child pedestrians. Br J Med 1995;310:413–4.4. Department of Human Services and Health. National goals, tar-
gets and strategies for injury prevention and control. Canberra:DHHS, 1994.
5. Injury Control Program Disease Control Health Department ofWestern Australia. An Overview of Injury in Western Australiathe 1995 Update. Perth: Health Department of Western Austra-lia, 1997.
6. Rivara FP. Child pedestrian injuries in the United States: currentstatus of the problem, potential interventions for future researchneeds. Am J Dis Child 1990; 144:692–6.
7. Gillam C, Stevenson MR. Evaluation of pedestrian safety educa-tion programs for young children: a review of methodologies. AsiaPacific J Public Health 1995;8(1):40–5.
8. Sandels S. Children in traffic. Revised edition. Surrey: Elek BookLtd., 1975.
9. Cross D, Darby J, Jones S, Noble S, Baxendale S. FormativeEvaluation Report to the Western Australian Road Safety inSchools Project. Perth: Curtin University Printing Service, 1995.
10. Stevenson MR, Lo SK, Laing B, Jamrozik K. Childhood pedes-trian injuries in the Perth metropolitan area. Med J Aust1992;156(4):234–8.
11. Cross RT, Pitkethly A. Concept modification approach to pedes-trian safety: a strategy for modifying young children’s existingconceptual framework of speed. Res Sci Technol Educ 1991;9(1):93–106.
12. Cross RT, Pitkethly A. A curriculum model to improve youngchildren’s concept of speed to reduce their pedestrian accidentvulnerability. School Sci Math 1989;89(4):285–92.
13. Dalis GT. Effective health instruction: both a science and an art.J Health Educ 1994;25(5):289–94.
14. Dunne RG, Asher KN, Rivara FP. Behavior and parental expecta-tions of child pedestrians. Pediatrics 1992;89(3):486–90.
15. Rivara FP, Bergman AB, Drake C. Parental attitudes and prac-
tices toward children as pedestrians. Pediatrics 1989;84:1017–21.16. West R, Sammons P, West A. Effects of a Traffic Club on roadsafety knowledge and self-reported behavior of young childrenand parents. Accident Anal Prev 1993;25(5):609–18.
CHILD PEDESTRIAN INJUR
17. Stevenson M, Iredell H, Howat P, Cross D, Hall M. Measuringcommunity/environmental interventions: The Child PedestrianInjury Prevention Project. Injury Prev 1999;5:26–30.
18. Education Department of WA. Health education K–10 syllabus.Perth: William C. Brown Printers, 1985.
19. Stevenson MR, Jamrozik KD, Spittle JA. A case–control studyof traffic risk factors and child pedestrian injury. Int J Epide-miol 1995;24(5):957–64.
20. Resnicow K, Cohn L, Reinhardt J, Cross D, Wynder E, AllegranteJ. A three-year evaluation of the know your body program ininner-city school children. Health Educ Q 1992;19(4):463–80.
21. Anataki C, Morris P, Flude B. The effectiveness of the Tufty Clubin road safety education. Br J Educ Psychol 1991;56:363–5.
22. Rivara FP, Booth CL, Bergman AB, Rogers LW, Weiss J. Preven-tion of pedestrian injuries to children: effectiveness of a schooltraining program. Pediatrics 1991;88(4):770–5.
23. Parcel GS, Edmunson E, Perry CL, Feldman HA, O’Hara-Tomp-kins N, et al. Measurement of self-efficacy for diet related behav-iors among elementary school children. J School Health 1995;65(1):23–7.
24. Gore CJ, Owen N, Pederson D. Case study of an educational andenvironmental intervention in school health promotion. HealthPromot J Aust 1992;2(3):38–40.
25. Haas M. Statistical methodology for reliability studies. J ManipPhysiol Ther 1991;14(2):119–32.
Y PREVENTION PROJECT 187
26. Burns S, Cross D, Stevenson M., Kenaston T. Measuring theimpact of the Child Pedestrian Injury Prevention Project on lowerprimary school-age children. Health Promot J Aust 1998;8(3):195–201.
27. Castles I. Socio-economic indexes for areas. Catalogue No. 2912.0.Canberra: Australian Bureau of Statistics, 1994.
28. Norusis MJ. SPSS professional statistics 7.5. Chicago (IL): SPSSInc., 1997.
29. Statistical Analysis System. SAS computer programs. Cary (NC):SAS Institute, Inc., 1996.
30. Cross D, Stevenson M, Howat P. Hall M, Officer J. (1998) ChildPedestrian Injury Prevention Project Final Report. Perth: CurtinUniversity Printing Service, 1998.
31. Pruesser D, Blomberg R. Reducing child pedestrian accidentsthrough public education. J Safety Res 1984;15:47–56.
32. Gore CJ, Owen N, Pederson D, Clarke A. Educational and envi-ronmental interventions for cardio-vascular health promotionin socially disadvantaged primary schools. Aust N Z J PublicHealth 1996;20(2):188–94.
33. Newman IM, Anderson CS, Farrell KA. Role rehearsal and effi-cacy: two 15-month evaluations of a ninth-grade alcohol educa-tion program. J Drug Educ 1992;22(1):55–67.
34. St Leger L. The origins, growth and challenges of health educa-tion in Victorian schools. In: Went S. (editor): A healthy start:holistic approaches to health promotion in school communities,second ed. Melbourne: Sands and MacDougall, 1992:17–28.