Testing a Training Program That Uses Virtual Reality Technology to Improve Children's Pedestrian Behaviors: A Randomized Controlled Trial

APPROACH

1. RESEARCH GOALS. The specific goals are: (1) conduct a RCT with 7-8 year olds to evaluate the relative effectiveness of 2 training programs (Virtual Reality, Street Side), compared to a no-intervention control, with a focus in training on where and how to cross. (2) Conduct an economic analysis of the programs to ascertain their relative cost effectiveness. These results will provide essential knowledge on relative program effectiveness and economic costs to inform future decisions about 'best practices' in child pedestrian injury prevention through behavioral training
2. STUDY DESIGN: 3 Groups Recruited children will receive the pre-tests and then be randomly assigned to one of three groups (2 Intervention, 1 Control), using a block (sex) randomization procedure so Intervention groups (VR training or VR-T, Street Side training or SS-T) and the no intervention Control group (No Intervention: NI-C) are balanced for sex of the child.
3. RESEARCH QUESTIONS Does performance in VR-T exceed that in NI-C? Does performance in SS-T exceed that in NI-C? Is performance equivalent in VR-T vs SS-T or does one produce better outcomes and qualify as best practice? Do any immediate gains persist over time (2 months post-intervention)? What is the cost-effectiveness of VR-T compared to SS-T?
4. POWER ANALYSIS All calculations were performed by a consulting statistician (Dr. Gerarda Darlington) using a function written in R version 3.3.2 based on a formula provided by Julious and using the pwr.anova.test function from the pwr package in R. Results from the pilot feasibility study were used to estimate mean changes for where and how to cross: after VR-T, children crossed in unsafe locations 65% fewer times and there was a 68% increase in how to cross safely, with both variables being normally distributed. With a within-group standard deviation of 0.16 (pilot data) and a sample size of 58 children per group, we would have over 80% power to confirm equivalence between the VR-T and SS-T groups, assuming no observed difference in change scores between the groups and an equivalence threshold level of 17%, with alpha=0.025. Note that the total sample size would support the analyses planned, allowing for significance at p < 0.0001 with power greater than 95%. Allowing for 10% attrition, we will recruit 64/group (N = 192).
5. SUBJECT POPULATION AND ELIGIBILITY CRITERIA [*see Introduction to Application please] A total of 192 children 7-8 years will be recruited, with 64 assigned to each group; this includes allowance for 10% attrition (usual rate=5%). Youth will be randomly sampled from an existing database of 9250 community recruited volunteers, with supplemental recruiting (as needed) through community organizations and advertising. Use of the database increases the speed and efficiency with which we can execute the research but does not change the base population-it is all a community recruited sample. Inclusion Criteria: developing normally (parent report; no learning, social, physical, mental health concerns), English speaking. Exclusion Criteria: self or anyone in family hospitalized for injury as a pedestrian, which can lead to atypical cautiousness.
6. RANDOMIZATION Blinding. Blinding to condition is not possible but greater bias may be introduced in RCTs through inadequate allocation concealment than with blinding. All participants will be blind to the study hypotheses.

Allocation of participants will occur by via a computer-generated random number sequence. 7. TIMING OF EVENTS AND MEASURES All children do an initial visit to complete the pre-test measures and are then randomized. Then SS-T and VR-T children complete 6 weekly training sessions (~30 minutes/session). All children then complete another visit for post-test immediate measures, with a longer-term visit (2 months later) to evaluate sustainability of any improvements. 8. SETTING Midblock crossing on two-lane bidirectional roads, without nearby traffic signals or crosswalks, are the focus.

For SS-T sessions, a road has been identified (50 km limit) that provides opportunity for training in where and how to cross in the same types of road situations to be presented in VR-T. Traffic volume was recently measured for 7 days and varied from light (8 cars/min) to heavy (16 cars/min). This range for traffic volume will occur in VR-T.

For VR-T sessions, these occur in the VR lab at UG campus. For pre and post measures, for all groups these occur in two settings: (a) street side (designated as "Field" below) with naturally varying traffic (children indicate where and how they will cross but do not do so), (b) in the VR lab in which children cross the virtual street in traffic using a gaming controller.