The Strike Out model consists of a carefully planned sequence of educational and dissemination strategies grounded in behavioral theories. Implemented in a community sports context, the intervention includes booster seat promotion by local champions, coalition development, capacity building, media, and direct education. Social marketing techniques of Kotler and Anderson were applied in the initial development and evaluation of the intervention.²² The diffusion of innovation model^23,24 was applied to explain the adoption of booster seat use as a community norm through multiple social systems (eg, health improvement coalitions, civic leaders, media, and baseball coaches). Consumer Information Processing model²⁵ was used to frame and promote booster seat use in a way that parents could readily draw conclusions about the importance of appropriate restraint selection and use. All materials developed for the program were tailored to reflect the baseball theme and to focus on using communication through the weekly local newspapers and community groups common in smaller communities.

A total of 20 communities participated in this trial, leading to 1,775 observations on children aged 4–7 (1,014 pre-intervention (baseline or T1) and 761 post-intervention (T2)). Demographic, vehicle, and AAP compliance data are summarized for pre-intervention (Table 2) and post-intervention (Table 3) by state. Pre-intervention variables-age of child, gender of child, gender of driver, driver is parent of child, driver using restraint, and type of vehicle driven-were significantly different. Gender of driver was only recorded if the driver was the parent of the children in the car; otherwise, only the relationship of the driver to the child was recorded. To minimize the number of missing observations in the model, only parent as driver was included in outcome models. Arkansas had the highest proportion of trucks assessed (20.4%) and the lowest percentage of children appropriately restrained at both the pre-intervention and the post-intervention assessment (21.0% and 36.4%, respectively).

Figure 2 shows the changes in proportions of children restrained in each state. For 3 of 4 states, the intervention resulted in increases in recommended child restraint use (Alabama +15.5%, Arkansas +16.1%, Illinois +11.0%). Communities in one state (Indiana) did not have a positive response (-9.2%). Overall, unadjusted restraint use increased 10.2% in intervention and 1.7% in control communities (P = .02).

Appropriate restraint use using AAP criteria was further assessed at post-intervention, stratified by state using CMH. Those Arkansas communities in the intervention group were 4 times more likely than those in the control group to have properly restrained children (Odds Ratio = 3.99, 95% Confidence Interval (2.22, 7.19), P < .001). Arkansas was the only state to have a statistically significant effect (Alabama OR = 1.17, P = .603; Illinois OR = 1.23, P = .567; Indiana OR = 0.73, P = .317), although across all states the overall compliance was significantly improved by the intervention, with intervention communities being 1.56 times more likely than those in the control group to have properly restrained children (OR = 1.56, 95% CI (1.16, 2.10), P = .003).

The sensitivity analysis using both the “complete case” and “all variable” GLMMs are summarized in Table 4. Overall, the missing data contributed to a reduction of 3.3% in the overall sample between the complete case GLMM (n=1,775) vs the all variable GLMM (n=1,715). The “all variable” model is a GLMM with all explanatory variables; it was estimated using only 1,715 observations as data were missing for the variables vehicle type and driver using restraint. The “complete case” is a GLMM estimated without these 2 explanatory variables that used all 1,775 observations. The estimation of both models is reported as a measure of sensitivity, but the GLMM with all variables is considered the best model. Both of the GLMMs estimated a significant effect of the intervention in the state of Arkansas, where children in the intervention communities were more than twice as likely to be correctly restrained compared to those in control communities. The overall effect was less pronounced across the 4 participating states. Using the CMH approach, the intervention resulted in a 56% increased likelihood of correct restraint overall (Odds Ratio 1.56, 95% Confidence Interval (1.16, 2.10)); results demonstrated a strong trend toward improvement using other models but did not achieve statistical significance.

Other variables independent of intervention were also associated with compliance with best practices for restraint use. Younger children were more likely to be properly restrained compared to an older child, with odds ranging from 1.4 (4-year-olds vs 6-year-olds) to 2.3 (5-year-olds vs 7-year-olds). Correct restraint use was 1.7 times higher if the driver of the car was the parent of the child and 2.8 times higher if the driver was also restrained. Children transported in all car types (minivan, SUV, sedan) were more likely to be properly restrained than children in trucks, while the odds of compliance increased 1.6 times if the child was in a minivan as compared to a sedan.

Premature graduation to lap and/or shoulder belts is a leading reason for injury among restrained children in survivable crashes, and booster seat use reduces likelihood of death and prevents lap belt syndrome associated with lap and/or shoulder belts alone.^33–35 Booster seat use decreases with the age of the child, and overall use remains approximately 41% and has actually decreased slightly in recent years.^36,37 Variation in education and resources, road environment, and cultural and socioeconomic factors combine to contribute to the lack of optimal restraint and increased mortality risk for individuals in rural communities.^38,39 Practical factors including lower population density, higher costs, and a relative lack of trained personnel have resulted in few interventions focusing on the high-risk rural population.

Our study aimed to overcome these barriers to booster seat adoption in rural communities. Strike Out involved a convenient, culturally tailored, and highly efficient intervention designed to provide information and personalized child passenger safety education within a family activity and in a setting where relatively large numbers of target age children and families would be present. Grounded in behavioral theory and practical in its approach, Strike Out was able to reach rural families in fairly large numbers by building on an existing activity where children and their parents were already gathering.

The intervention demonstrated substantial short-term changes in child restraint practices. Strike Out appeared to be most effective in the state (Arkansas) with the lowest pre-intervention rates of booster seat use, and it may therefore be most successfully replicated in other rural areas where booster seat use is low. In 2 states where baseline booster seat use was greater, the increases in appropriate use approached but did not reach statistical significance. For reasons that are not clear, the small sample in one state (Indiana) did not respond to the intervention. The study also supports previous research which demonstrated that adults who wear seatbelts, parent compared to non-parent drivers, and drivers of minivans and sedans are more likely to restrain children correctly. Conversely, children in light trucks were much less likely to be restrained in booster seats, reinforcing the need for increased intervention and education in rural areas, where such vehicles are common.

The Strike Out intervention was well received in participating communities and was an impetus for sustained child passenger safety education in numerous communities. An implementation guide has been developed to aid in replication of the program. The guide includes step-by-step instructions based on the experience gained in the research project. This guide is being distributed to direct expansion of the program in interested communities in Arkansas. Other future directions of interest include expanding the sports-themed construct to include other team sports, such as soccer or football.

Several limitations should be noted in our study. First, the post-intervention surveys were conducted within a short time after the intervention activities were concluded, and therefore the longer-term effects of the intervention cannot be determined from the current study. Second, while the teams observed remained largely consistent during the season, no longitudinal individual measurements were made, and the results therefore reflect overall proportions of correctly restrained children observed in the communities. The 25% reduction in subjects from pre-intervention to post-intervention paralleled attrition in participation in the baseball programs that occurs over the course of the season as the school year ends and competing summer activities ensue. Unanticipated complicating factors (ie, severe weather, competing community interests) resulted in alterations of the planned sequence of Strike Out components in some communities, although every possible effort was made to not compromise the core components of the model. Finally, it is not known if comparable community engagement or effectiveness will be found in communities with different demographics, in urban settings, or when delivered through organized sports other than baseball.

A tailored intervention using baseball programs increased appropriate restraint use among rural children of booster seat age, both overall and in 3 of 4 states studied. Effects were most pronounced in the state with the lowest pre-intervention booster seat use. Such interventions hold promise for expansion into other sports and populations, particularly in rural areas with low adoption of booster seat technology.