Evaluation of an Education, Restraint Distribution, and Fitting Program to Promote Correct Use of Age-Appropriate Child Restraints for Children Aged 3 to 5 Years: A Cluster Randomized Trial

Abstract

Objectives. We evaluated an education, distribution, and fitting program for increasing age-appropriate and correct child restraint use.

Methods. We performed a cluster randomized trial involving 28 early childhood education centers in low socioeconomic status areas in Sydney, Australia. The main outcome was optimal restraint use defined as age-appropriate restraints, installed into the vehicle correctly and used correctly.

Results. One service withdrew after randomization, so data are presented for 689 child passengers, aged 3 to 5 years, from 27 centers. More children attending intervention centers were optimally restrained (43% vs 31%; P = .01; allowing for clustering). More 3-year-olds were using forward-facing seats rather than booster seats, more 4- to 5-year-olds were using booster seats instead of seat belts alone, and there were fewer errors in use at intervention centers. Among non–English-speaking families, more children attending intervention centers were optimally restrained (43% vs 17%; P = .002; allowing for clustering).

Conclusions. The program increased use of age-appropriate restraints and correct use of restraints, which translates to improved crash injury protection. Multifaceted education, seat distribution, and fitting enhanced legislation effects, and the effect size was larger in non–English-speaking families.

Injury in road traffic crashes is a major cause of death and disability in children. In the United States each year more than 800 children younger than 10 years are killed,¹ and many more are seriously injured as passengers in road crashes. Although restraint systems appropriate for a child’s size are proven to reduce crash injury,^2–5 best-practice child restraint use continues to be low.^6–8

Reductions in fatalities and serious injuries are possible through increasing both age-appropriate restraint use and correct use of these systems.⁵ Legislation has an important role^9,10 and in 2009 the Australian Road Rules were changed, specifying different restraint systems by age for children aged up to 7 years.¹¹ However, achieving widespread best-practice child restraint use requires education and restraint distribution programs.¹² Furthermore, legislation has been shown to be less effective in low-income and minority groups.¹³

Among the few well-designed studies that have evaluated community-based programs to increase child restraint use, success has been mixed.^9,10,14 To our knowledge, no study has investigated the impact of such programs on correct installation and securing of the child within the restraint, which is critical for performance of a child restraint system.¹⁵ We developed a multifaceted program aimed at enhancing legislative effects to increase best-practice child restraint use and used a cluster randomized trial to evaluate the program. Cluster randomized trials are well suited to evaluation of injury prevention initiatives that are typically delivered to groups or communities. This design recognizes that there are societal norms for child restraint use and that peers can influence behavior.

METHODS

We conducted a cluster randomized trial involving children aged 3 to 5 years attending preschools, which operate between 9 am and 3 pm, and long daycare centers, which have extended hours, in areas of low socioeconomic status in Sydney, Australia.¹⁶ We confined our study to the 2010 calendar year to minimize change of enrollments in participating centers. Baseline surveys were completed in February and March 2010 and final assessments in September through December 2010.

Center Recruitment

We identified centers from a 2009 database of all preschools and long daycare centers. Eligible centers were within Statistical Local Areas of west and southwestern Sydney with a Socio-Economic Index For Areas (SEIFA)¹⁷ in the lower 30% of metropolitan Sydney (SEIFA < 1010). This is the most culturally and linguistically diverse region of Sydney.¹⁸ Only community-based preschools and long daycare centers were eligible, and we excluded 42 state-administered centers and 15 that were work-based, mobile, or drop-in services.

We excluded centers with fewer than 20 families,¹⁶ as well as centers that were unsuitable for observational surveys for space or safety reasons. We approached eligible centers in random order from 2 lists of eligible preschools (n = 19) and long daycare centers (n = 72), until we had enrolled 10 preschools and 18 long daycare centers (Figure 1). We conducted baseline self-report surveys at all centers.¹⁹

FIGURE 1—

Flow of preschools and long daycare centers through the trial: Buckle-Up Safely program evaluation, Sydney, Australia, 2010.

Participants

All children enrolled at intervention centers and their families were exposed to the program. Children aged 3 to 5 years who arrived by motor vehicle during the postintervention period of observation were eligible to participate in the final outcome assessment.⁶

All arrivals were documented together with reasons for nonparticipation. General characteristics of these arrivals were recorded including gender of driver, use of seatbelt by driver, whether preschool children sat in the front seat, and number of child passengers. We compared the characteristics of families who did and did not participate to estimate possible bias in sample selection.

Intervention Program

We designed the Buckle-Up Safely program²⁰ to increase both age-appropriate and correct use of child restraint systems. We developed the program by using the precaution adoption process model.²¹ We refined program content by using focus groups in our target population, appraisal of other programs,^9,10,14,22 review of known barriers to child restraint use in Australia,^23,24 and pilot testing in 2009.

The Buckle-Up Safely program consisted of a workshop for educators, distribution of printed educational material in English and 6 community languages, parent information sessions, supply of child restraints at subsidized cost, and free restraint checks. Intervention centers were offered a 2-hour professional development workshop and support to develop safe arrival and departure policies. These aspects were delivered by the Kids and Traffic Road Safety Education Program, which is a state-funded organization, providing education to early childhood education services in New South Wales. A parent information session was conducted 6 to 8 weeks after the staff workshop to allow for program implementation. At this session, an educational DVD developed for this program (available at: http://www.youtube.com/user/TheGeorgeInstitute#p/u/17/pdkZ-ZDxwtA), was screened and child restraints were demonstrated. Each family, regardless of attendance at the information session, was provided with an information pack containing the DVD, printed educational material, and a voucher for a free fitting check at a local authorized child restraint fitter (value A$50). A limited number of child restraints (approximately 10%) were also offered at a subsidized cost of A$50, approximately 25% of the recommended retail price.

Centers not allocated to the intervention continued with their usual educational programming in 2010 and were offered the Buckle-Up Safely program on study completion. The center staff were informed of group assignment.

Objectives

The objective of the trial was to measure the impact of the Buckle-Up Safely program at an individual level for children aged 3 to 5 years attending preschool or long daycare. We hypothesized that the program would increase age-appropriate child restraint use and decrease misuse of child restraint systems.

Randomization was by center. We used restricted randomization²⁵ to balance service types (preschool or long daycare) and to limit differences in the proportion of children using age-appropriate restraints at baseline (self-reported) to within 5% and families with household income less than A$60 000 per annum, families who speak a language other than English at home, and the number of children enrolled at the centers to within 10% between the 2 arms of the study. Randomization was performed by a study investigator (J. M. S.) who was not involved in recruitment or contact with the centers and who was blind to center identity.

Measurement of Restraint Use

Observers measured trial outcomes by direct observation of children as they arrived at their preschool or long daycare center.⁶ Observers were blind to the center allocation, though this was not evaluated and it was possible that they could identify intervention centers by visible program material. Each observer collected data at both intervention and control sites to minimize the impact of observer bias. Bilingual research assistants were employed with fluency in community languages to conduct interviews. Translated versions of the consent form and interview were available in Chinese, Vietnamese, Arabic, Hindi, Turkish, and Dinka languages. Where able, respondents scribed their own responses for income, education, and other demographic information.

Research assistants attended the centers for approximately 2 hours at the arrival times on consecutive days over a 1-week period. When the enrollment target of 20 observations was reached or more than 70% of eligible children at that center had been included, recruitment was ceased. If targets were not met, researchers attended the center for up to 3 days the following week. The duration of recruitment was limited as the presence of the researchers may serve to change how people restrain their children in cars on subsequent days.

We selected 1 child from each vehicle. If there was more than 1 child aged 3 to 5 years, we selected the child with the next birthday. At the same time as the interview a second researcher observed the child in situ, noting type of restraint and how the child was secured. After the child had left the vehicle, installation of the restraint was inspected. Experienced restraint fitters or research assistants who had completed an accredited restraint fitting course to ensure necessary technical skills collected the data.²⁶

We defined age-appropriate restraint use as a forward-facing child restraint for children aged up to 4 years and a forward-facing seat or booster seat up to age 7 years. Using the front seat was considered inappropriate for children younger than 4 years and was only appropriate for children aged 4 to 7 years if no rear seat was available, consistent with new child passenger legislation.¹¹

We dichotomized usage and installation errors into significant (2 or more minor errors or at least 1 major error)¹⁵ or not significant (no errors or 1 minor error). For analysis, we combined age appropriateness and errors to give 4 ordered categories: (1) optimal (age-appropriate restraint and no significant errors), (2) suboptimal (not age-appropriate but no significant errors), (3) significant errors with an age-appropriate restraint, and (4) significant errors without an age-appropriate restraint. We based the order of these categories on likely impact of not using an age-appropriate restraint and errors in use.¹⁵

Sample Size and Statistical Analysis

Allowing for 30% drop-out, we required 14 intervention and 14 control centers each with at least 20 children aged 3 to 5 years to detect a 20% change in age-appropriate restraint or errors of usage (from a baseline of 60%) with 80% power at 5% significance level, assuming an intraclass correlation coefficient (ICC) of 0.06.²⁷

We performed all analyses with either SAS version 9.1 (SAS Institute, Cary, NC) or Stata version 11 (StataCorp LP, College Station, TX). We used Fisher’s exact test to compare characteristics of participants and nonparticipants. We tested the proportional odds assumption for the ordinal outcome variable for each predictor variable by using the OMODEL command followed by the Brant test. We analyzed all intervention effects by intention to treat and adjusted for clustering by center by using a random effects model with robust standard errors (GLLAMM command with OLOGIT link), from which we also estimated the ICC.

We conducted a posthoc subgroup analysis by whether English was spoken at home, using a test for interaction between language and treatment group. This was because families who spoke a language other than English at home had low baseline age-appropriate restraint use¹⁹ and other similar programs have had lesser impact in minority groups.¹⁴

RESULTS

During center recruitment in October through December 2009, 28 of 34 (82%) centers that were approached agreed to participate, although 1 center withdrew and was replaced before randomization in March 2010. We conducted baseline surveys at all 28 centers (Table 1). After randomization, 1 intervention center withdrew citing time commitment concerns and it was not replaced. We did not complete any observations at this center, so data from only 27 centers were available for analysis (Figure 1). As there was a school term break in April, the staff workshops were delivered between May and July, parent sessions followed 6 to 8 weeks later, and child restraints distribution was completed by mid-November. At least 1 staff member from each intervention center attended 1 of the 12 workshops conducted. Printed material was displayed at all services and more than 800 information bags were distributed. In total, 16 parent information sessions were held and 164 restraints were sold to families and professionally installed. In addition, 40 free restraint check vouchers were redeemed.²⁰

TABLE 1—

Baseline Characteristics of the Preschools and Long Daycare Centers and Children Attending Intervention and Control Centers in February and March 2010: Buckle-Up Safely Program Evaluation, Sydney, Australia

Characteristics	Intervention Group, No. (%) or Mean ± SD (Range)	Control Group, No. (%) or Mean ±SD (Range)
Center
Number	14	14
Type of servicea
Preschool	5	5
Long daycare	9	9
Enrollments in ages 3–5 y in 2010b	62 ±21 (24–98)	58 ±24 (28–105)
Staff teaching children aged 3–5 y in 2010	9 ±3 (5–15)	9 ±5 (4–23)
Participants
Number	534	529
Age of children, y
3	219 (41)	231 (44)
4	261 (49)	242 (46)
5	54 (10)	56 (11)
Gender of child (7 missing data)
Male	257 (48)	282 (54)
Female	274 (52)	243 (46)
Age-appropriate restraintb	381 (71)	389 (74)
Relationship of survey respondent to childc
Mother	371 (69)	378 (71)
Father	112 (21)	109 (21)
Grandparent	30 (6)	25 (5)
Other	21 (4)	17 (3)
Children in family aged < 18 y	2.3 ±1.1	2.2 ±1.0
Level of parental education (57 missing data)
None, primary, or some secondary school	53 (10)	58 (12)
Completed secondary school	102 (20)	117 (24)
Some university or technical and further education	161 (31)	177 (36)
University graduate	137 (27)	101 (20)
Postgraduate degree	60 (12)	40 (8)
Household incomed (171 missing), A$
<20 000	36 (8)	40 (9)
20 001–40 000	65 (14)	73 (17)
40 001–60 000	106 (23)	78 (18)
60 001–80 000	70 (15)	74 (17)
80 001–100 000	69 (15)	86 (20)
100 001–150 000	73 (16)	60 (14)
> 150 000	38 (8)	26 (6)
Speak a language other than English in homec	124 (24)	96 (19)

Note. Some percentages may not add to 100% because of rounding.

^aRandomization balanced between intervention and control centers.

^bRestricted randomization to produce no more than 5% difference between intervention and control centers.

^cFour surveys for the intervention and 3 for the control centers were completed with both parents present so numbers add up to more than 100%.

^dRestricted randomization to produce to no more than 10% difference between intervention and control centers.

Observations were conducted over 4 to 8 days at each center (median 5 days) and were completed between September and December 2010. During data collection, 2473 vehicles arrived at the 27 centers. Of these, 518 had already participated, 100 could not be recruited because no research assistant was available, and 427 were not eligible (266 no child passengers, 127 no children in age range, 29 not attending center, and 5 unspecified). Among the remaining arrivals, 736 of 1428 (52%) agreed to participate. Because there were opportunities to participate on multiple days, this is a conservative response rate estimate. Reasons for nonparticipation included lack of time (567 of 692; 82%), refusals with no reason (101 of 692; 15%), language difficulties (17 of 692; 2%), arrivals by taxi (2 of 692; < 1%), and miscellaneous (5 of 692; < 1%). Refusal rates were similar in intervention and control centers (305 of 660; 46% vs 387 of 758; 50%; P = .12).

The characteristics of participants and those who declined are summarized in Table 2. Nonparticipants were more likely to have a preschool child in the front seat (42 of 692; 6% vs 28 of 736; 3%; P  = .02) and somewhat more likely to not be using a child restraint (97 of 692; 15% vs 81 of 736; 11%; P  = .06).

TABLE 2—

Characteristics of Families Participating in the Evaluation of Child Restraint Use and Those Who Declined to Participate From Observation on Arrival at the Preschool or Long Daycare Center: Buckle-Up Safely Program Evaluation, Sydney, Australia, 2010

Characteristics	Participants (n = 736), No. (%)	Nonparticipants (n = 692), No. (%)	P
Male drivers	198 (27)	205 (30)	.26
Driver not using seatbelt	2 (0.3)	2 (0.3)	> .99
Preschool child or children not in child restraint	81 (11)	97 (15)	.06
Using restraint incorrectly	192 (34)	167 (38)	.18
Preschool children in front seat	24 (3)	42 (6)	.02
≥ 3 children	28 (4)	28 (4)	.89

Although 736 agreed to participate, the final analytic data set was 689 observations (4 no parent interview, 1 child restraint type not identified, 3 missing data for age, 39 children younger than 3 years). There were 238 (35%) 3-year-old children, 303 (44%) 4-year-old children, and 148 (21%) 5-year-old children. We met our target of at least 20 children aged 3 to 5 years at 21 of 27 centers. Those where fewer than 20 families participated were smaller centers (28 to 50 enrollments) or centers where a large proportion of families did not arrive by vehicle.

Among the 3-year-old children, 99 (41%) and of 4- to 5-year-old children, 57 (13%) were not in an age-appropriate child restraint. Age-appropriate restraint use was more commonly observed in children from intervention centers than control centers (270 of 328; 82% vs 263 of 361; 73%; P = .02; ICC = 0.034).

Significant errors in either installation or use of the restraint were observed in 333 of 689 (48%) of children and more often in control centers (194 of 361; 54% vs 146 of 328; 45%; P = .02; ICC = 0.008). The most common types of securing errors were nonuse or partial use of the harness or seat belt and loose, twisted, or damaged webbing. The most common installation errors were not having the top tether attached to the anchorage point or failure to engage the buckle or thread the seat belt through the restraint.

Although the majority of children in the intervention and control centers were in age-appropriate child restraints, fewer than half were optimally restrained (age-appropriate and no errors in use). The proportion optimally restrained was 43% at intervention centers and 31% at control centers (Figure 2).

FIGURE 2—

Restraint use by age in the case and control centers: Buckle-Up Safely program evaluation, Sydney, Australia, 2010

Children attending the intervention centers were more likely to be optimally restrained after we adjusted for clustering effects, low income, education level, and language spoken at home (P = .002; Table 3). They had 1.73 times (95% confidence interval [CI] = 1.22, 2.45) the odds of being in a better category of the 4-point ordinal scale than those in the control group. The effect size was larger in the subgroup who spoke a language other than English at home (adjusted odds ratio = 3.65; 95% CI = 1.53, 8.73 vs 1.50; 95% CI = 1.05, 2.14 for English speakers; P = .01 for interaction). This difference between intervention and control centers within the non–English-speaking families was driven by a reduction in usage errors in the intervention subgroup rather than a difference in age-appropriate restraint use.

TABLE 3—

Improvement in Restraint Use in Children Attending Preschools or Long Daycare Centers That Received an Integrated Program Including Education, Subsidized Restraint Distribution, and Restraint-Fitting Vouchers: Buckle-Up Safely Program Evaluation, Sydney, Australia, 2010

Variable	Intervention, No. (%)	Control, No. (%)	Intracluster Correlation Coefficient	OR for Improvement in Best-Practice Restraint Use (95% CI)	P
No. of centers	13	14
Primary outcome (ordinal)
Optimal restraint use (1)	140 (43)	111 (31)
Inappropriate restraint, no errors in use (2)	42 (13)	56 (16)
Appropriate restraint, errors in use (3)	130 (40)	152 (42)
Inappropriate restraint, errors in use (4)	16 (5)	42 (12)
Total children	328 (100)	361 (100)
All centers
Unadjusted			0.026	1.67 (1.17, 2.39)	.005
Adjusteda			0.012	1.73 (1.22, 2.45)	.002
English spoken at home	282	255
Unadjusted			0.023	1.49 (1.03, 2.18)	.04
Adjustedb			0.008	1.50 (1.05, 2.14)	.03
Other language spoken at home	58	60
Unadjusted			0	2.87 (1.46, 5.66)	.002
Adjustedb			0	3.65 (1.53, 8.73)	.004

Note. CI = confidence interval; ICC = intracluster correlation coefficient; OR = odds ratio. Analysis stratified by whether English was spoken in the home (n = 34 missing data). All models used ordinal regression with a random effect to allow for clustering and satisfied the proportional odds assumption. P = .01 for test for interaction between intervention arm and language spoken at home. n = 110 missing data for income and n = 42 missing data for level of education, so adjusted analyses are for n = 575. Some percentages may not add to 100% because of rounding.

^aAdjusted for low income (< A$60 000 annual family income), education level (≤ secondary school), and language spoken at home.

^bAdjusted for low income (< A$60 000 annual family income) and education level (≤ secondary school).

DISCUSSION

When small children are secured in a correctly installed, age-appropriate child restraint they are far less likely to be seriously injured in a crash.^2–5 Although laws that mandate age-appropriate restraint use have increased use for child passengers¹³ and have reduced severe crash injuries,²⁸ high rates of compliance have not been achieved.^8,12,29 Furthermore, it is nearly impossible to legislate for correct restraint installation and use. Here we have demonstrated that a program integrating education, hands-on demonstration, subsidized restraints, and free fitting checks can increase use of age-appropriate restraints and reduce errors in use. This is the first trial to our knowledge that has evaluated both usage errors and type of restraint.

The Buckle-Up Safely program showed an impact in the same year that new legislation on age-appropriate child restraints was enacted in Australia.¹¹ Our findings suggest that the impact of legislation can be enhanced, particularly in relation to the correct use of child restraints.

Our program showed a greater effect in families who spoke a language other than English at home, unlike a previous trial of similar design in the United States where there was a reduced effect in minority groups.¹⁴ Our success in effecting change in this group is important as these families have lower awareness of new laws,¹⁹ different patterns in access to information,¹⁹ and lower rates of age-appropriate restraint¹⁹ and correct use.²³ In the United States, 1 study found no change in minority groups while legislation increased age-appropriate restraint use overall by close to 20%.¹³

Our educational resources explained graphically how age-appropriate and correct use of restraints is important to injury prevention in crashes. Previous cross-sectional studies have demonstrated that many parents believed they knew all they needed to know to safely restrain their children, yet failed to follow best practice.²⁴ We believe the program was successful in engaging families on safety benefits.

Education has previously been shown to increase use of age-appropriate restraints, but a systematic review found larger effects when incentives were offered.³⁰ Furthermore, an association between nonuse of an age-appropriate restraint and low socioeconomic status has consistently been reported.^8,13 It is likely that the provision of reduced-cost seats was a critical part of the success of our program even though restraints were made available to less than 10% of families at intervention centers. Without this assistance, these families may not have moved into the stage of actually acquiring the restraint even if they were convinced they should do so.²¹ The costs associated with subsidizing restraints and administering a restraint distribution scheme are significant and have implications for the wider use of the Buckle-Up Safely program. A more sustainable approach may be to support production of low-cost, high-quality restraints.

Another important feature of our program that we believe contributed to the success was the integration with existing services such as the Kids and Traffic Road Safety Education program, which has been providing road safety education in this sector for approximately 20 years. We believe that using this existing service provider, expert in the early childhood education delivery area, added to the overall credibility of the program.

Limitations

There were some limitations in our approach. We evaluated the impact of the program in the same year that it was implemented and therefore we do not know if there was a long-term impact. The high refusal rate is understandable as we were approaching families at a busy time of day and time was the primary reason for nonparticipation. However, those who did not participate were more likely to have a preschool-aged child in front, and this suggests that we are likely to be underreporting use of the front seat. It is not certain whether the bias extends to age-appropriate restraint use, although it is possible. Our results may therefore present child passenger seating as better than it is. Refusal to participate and front seating in refusal cases was similar in the control and intervention centers, so is not likely to have had an impact on the overall trial result.

Our participants were children attending an early learning center and those children who arrived at the center by vehicle. Although approximately 59% of children attend preschool and a further 25% long daycare in Australia, rates of attendance are known to be lower in low-income households.³¹ This has implications for generalizability of the results.

Although we were not able to reach recruitment targets at all centers, the study was sufficiently powered to allow for this drop in sample size and the loss of 1 center before delivery of the intervention. Further strengths of our study design were the randomized design, masking of observers to group allocation, and objective measure of study outcome.

Conclusions

This trial provides much-needed community-based data on the efficacy of an integrated program to promote best-practice child-restraint use. The program has improved child restraint practices in this setting, including those for whom English is not the primary language. This is an important achievement when one considers the challenges to communicating effectively with this group who has previously been reported to be at highest risk for non–age-appropriate restraints and usage errors. Most importantly, the success of this intervention demonstrates how comprehensive community-based interventions that target best-practice child-restraint barriers can work to enhance the effect of legislation. The trial results have implications for policy in this area and could ultimately contribute to reductions in child passenger injuries.