Childhood injury remains among the leading causes for childhood morbidity and mortality.^1,2 Such injuries result in significant human and societal costs, with both short and long term consequences ranging from interruptions in school/sports activities to significant cognitive and behavior disorders, depression, and deterioration of family functioning, physical morbidity, and sometimes death.^3–7 However, researchers have long advocated that as much as 90% of injuries could be prevented through coordinated and effective education, engineering, legislation, and enforcement strategies.^7,8

GROUP INTERVENTIONS AMONG YOUNG CHILDREN

There has continued to be much debate about the effect of injury prevention strategies targeted directly towards children, in particular, preschool age children. Some argue that children in this age group do not possess the cognitive ability and attention to permit the integration of injury prevention messages into their repertoire of skills.¹¹ However, other researchers advocate that taking advantage of early intervention may provide a foundation for long term application of safety behavior in preventing injuries.^12–14

Systematic reviews have made it possible to formally and comprehensively assess the quality of studies and draw significant conclusions that permit further advances in the area of inquiry. Systematic reviews not only provide comprehensive and objective evidence to guide practice, often resolving controversies, but they can also direct future research efforts.^9,10 The intent of this systematic review is to explore group based injury prevention interventions that targeted young children (under age 6 years) to determine the effectiveness of such strategies in enhancing children’s safety behaviors. Group level interventions refer to strategies that provided interventions in a systematic manner to a group of children, primarily in classroom or day care settings. Interventions could include a broad range of strategies such as skills training, interactive games, activities, modeling and rehearsal, written material, and videos, but not a community component. Interventions that focused on children but may have also incorporated parents as secondary targets were included. Individual interventions were not included as they are typically delivered under different circumstances, usually in a healthcare setting where parents are the primary target population (for reviews see DiGuiseppi & Roberts).²

METHODS

Inclusion criteria

Studies were included in the systematic review if they met the following criteria: (1) the intervention engaged children under the age of 6 years; (2) included a control group; (3) used a group intervention approach; (4) were written in English; (5) addressed unintentional injuries, and (6) outcome measures included injuries, knowledge, or safety behaviors. The decision to include all experimental studies whether randomized or not was due to the nature of group level interventions that most often occur in natural settings where it is often difficult to systematically assign participants randomly. Only studies with control groups were included in an effort to minimize bias. Only published English studies were included due to the limitations of the researcher to access translation services. No studies of intentional injuries (violence, bullying, and abuse) were included as strategies are often unique to this safety issue and may not be appropriate for unintentional injuries. Knowledge, safety behaviours, and injury events were included although knowledge and safety behaviors are known to be the outcomes predominantly measured in these types of studies.⁹

Identification of studies

The search strategy used an electronic search of several databases including The Cochrane Library, CINAHL (Cumulative Index to Nursing and Allied Health Literature), PUBMED, Cambridge Scientific Abstracts, Dissertation Abstracts, and PsycINFO for the years 1980–2002 in an effort to enhance comparability between study designs. Search strategy included terms “child”, “risk taking”, “education”, “evaluation”, “prevention and control”, “safety”, and “injury prevention”. In addition, bibliographies of published studies and reviews were searched for relevant studies that met the inclusion criteria. A review of additional websites included; http://depts.washington.edu/hiprc/childinjury, http://www.safetylit.org,www.thecommunityguide.org, and http://www.cdc.gov/ncipc/injweb/websites.htm. Abstracts from the 6th World Conferences on Injury Prevention and Control were also reviewed. Authors of relevant studies and reviews were contacted as well as recognized experts in the field of childhood injury prevention research. After screening titles and abstracts, 307 studies were identified as potentially eligible for inclusion. Potentially appropriate articles were further reviewed to determine if they met the inclusion criteria, yielding nine studies for full review.

Data extraction

Two researchers independently extracted the data after full articles were screened for inclusion. Data detailing study design, participants, and types of injury and interventions were documented (see table 1). Any disagreements were resolved by consensus. Quality assessments were not conducted due to the diverse study designs and the lack of randomized control groups. Descriptive analysis was chosen rather than meta-analysis as the data findings and statistical analysis reported for outcome measures varied considerably, making the combining of results inappropriate. Primary outcome measures were compared for differences between pre and post intervention between the control and intervention groups. Variability in study findings was examined with respect to study design and intervention characteristics.

RESULTS

The areas of safety interventions included four road crossing,^15,16,18,20 two car restraint,^12,19 one spinal cord safety,¹³ one poison safety,¹⁴ and one combination of 911 (North American emergency number)/stranger danger/street crossing.¹⁷ The types of interventions included videos, interactive activities, cartoons, stories, puppets, singing, coloring, games, simulation games, demonstrations, modeling/role playing and rehearsal practice using seat belts, and model and real street crossing. The parent directed activities included workshops, home activities, and written information. In addition to the variety of interventions delivered, the duration of delivery varied substantially from a minimum of one 40 minute session to two weeks’ intensive classroom instruction of unspecified session duration.

The participants ranged in age from 3–6 years of age. Eight studies engaged preschool students (3–5 years old) and one study¹⁴ was conducted with kindergarten children (5–6 years old). All studies had a control group. The majority of school sites were chosen based on the cooperation of the school board/community resource agencies or the school/teacher willingness to follow protocol and the suitability of school environments for intervention. Some study sites were selected for their location, in high risk neighbourhoods or in different socioeconomic communities. The number of children participating in each study was broad, ranging from 30 to 829 children. Several studies either matched for gender and socioeconomic status or controlled for these potential confounding factors in the analysis phase. The interventions occurred in classroom groups in all nine studies with teachers delivering the intervention in four studies,^12–14,19 trained parent/volunteer in three,^16,17,20 the researcher in one,¹⁵ and unknown in one study.¹⁸ Most studies provided interventions to the entire class with only two studies using small groups of five¹⁵ and three.¹⁶ Table 1 includes descriptions of each study’s population, intervention, and outcome measures.

Four studies reported random selection, two by school^12,17 and two by student.^15,18 Measurement data were collected by teachers for the most part, and in two studies by the researchers themselves.^15,16 Only one study reported blinded observations.¹⁶ Outcome measurement occurred anywhere from immediately following the intervention to six months after intervention. Overall, the review revealed a positive effect for five of the studies;^{12,14–16,19} three had mixed effects^13,18,20 and one reported no effect.¹⁷ Of those studies demonstrating an effect, differences in behaviour,^12,15,16,19 attitude,¹⁸ and knowledge^14,18,20 were reported. No studies measured changes in injuries.

Table 2 includes descriptions of the intervention and control groups and the strengths and weaknesses of each study.

Potential bias

Many of the schools were chosen based on convenience and willingness of the school to permit research activities to take place, whereas other schools were chosen for their representativeness of different geographic areas and socioeconomic demography. Although attempts to select schools representative of the population are useful, the convenience sampling was one of practicality and consideration must be given to the potential for bias when convenience samples are used. Cluster randomization is often considered to be the gold standard in these types of studies but can be difficult to achieve and complex to implement in community settings.²¹ Only two of the studies used cluster randomization.^12,17

Although most studies indicated that considerable effort had been made to ensure the individual responsible for the delivery of the intervention was well prepared before initiation of the intervention, some studies were vague about preparation^17,18 and monitoring for consistency of program delivery. In fact, most studies reported that the delivery varied depending on the individual. Despite the reported attempt to establish a minimum expectation, investigators did not report on the comprehensiveness of delivery beyond the minimum nor did they monitor for fidelity of intervention that ranged from one 40 minute session to two weeks of intense learning opportunities. Furthermore, some studies actively engaged parents whereas others did not report on the level of parental involvement. In other respects, the follow up measurement varied considerably from immediately following the intervention to six months after intervention. Clearly the intensity and duration of intervention follow up and preparation of the interventionist may influence the outcome assessment.²¹ Outcomes were usually measured using observation techniques or knowledge tests based on the intervention content. Most assessments were primarily developed for the specific study and were not rigorously tested. Although observation is a useful strategy, it was not clear in the majority of the studies whether the observers were blind to the group assignment of the children. Lack of blinding could bias reporting. More generally, potential for publication bias must be considered given that only studies published in English since 1980 were included in this review.

DISCUSSION

This review suggests that group interventions could enhance children’s safety behaviors during early childhood. The studies were all conducted with children aged 3–6 years from a diversity of neighbourhoods and socioeconomic backgrounds. Although the diversity of samples makes comparison of studies difficult it strengthens the external validity of the consistent effect of the interventions in different populations. In addition, regardless of the safety issue addressed, positive results were demonstrated providing evidence that children may respond to such strategies. The rigor of the study designs were similar in that they all included a control group and all but two conducted baseline testing for comparison. The weakness of the interventions was in the lack of control over the intensity and duration of the intervention, and preparation of the interventionist. None included an adequate measure of fidelity of the intervention. It is difficult to conclusively analyze the findings without this level of rigor that could significantly contribute to differences in outcomes. In addition, no study attempted to measure rates of injuries, so it remains unclear whether such interventions have the potential to actually alter injury rates among children. Furthermore, the potential for long term impact, particularly in relation to injury rates cannot be determined, as no study included a long term follow up. Future efforts should focus on developing collaborative research teams that include community partners such as teachers and parents using studies of enhanced rigor. Such initiatives could strengthen opportunities to conduct long term studies that measure key factors linked to injuries such as parental involvement, teacher awareness and prevention strategies, school and community attributes, child behaviors, peer relationships, and supervision. Such efforts may yield more accurate evidence of associations and lead to development and measurement of causal relationships.

The elements of the successful programs included group sessions that incorporated multiple interactive learning tools^{12,14–16,19} including group activities and rehearsal opportunities. Typically, the leaders were well trained and could be teachers, parents, or volunteers. The optimal dose is not clear and requires further assessment. Furthermore, it is not obvious whether engaging parents as reinforcement has any significant effect. Practice appears key to positive outcomes in that children need to be exposed to opportunities to develop problem solving skills rather than content specific knowledge alone.^15,16,20 Equally important is the interactive component rather than the traditional didactic approach to increasing children’s knowledge of safety behavior.^{13–16,18–20} A comprehensive approach that integrates such learning approaches into curriculum in the early years may be worthy of further research.

Regardless of the safety topic that was addressed, the majority of studies demonstrated some positive effects. This suggests that potential exists to positively influence the development of safety behaviors among young children using group interventions. The challenge remains as to whether these behaviors are transferable and can be incorporated into a child’s repertoire of decision making skills about injury risks. This could create the potential to influence the rate of injuries.

Future research needs to standardize intervention strategies in a variety of situations to better assess the ability of such interventions to impact on children’s safety behaviors and ultimately on the risk of injury. Although the challenges are numerous and often discouraging in community settings, it remains important to establish scientific rigor through the random allocation of children or groups of children to treatment and control groups. Studies that incorporate established data collection tools, blind observation, and long term follow up would be useful to further our understanding. It would be valuable to explore whether outcome assessments are responsive to dose intensity, duration, or parental participation. If such strategies can be proved to be effective, it may be possible through early intervention to influence the development of more effective risk taking strategies for young children. Developing partnerships between researchers and education and community agencies could provide for future collaborative opportunities and help researchers to conduct more rigorous, long term studies.