Aim 1n Chapter 3, we examined routinely-available data sources (police crash records, hospital emergency department (ED) visits, and death certificates) that captured pedestrian injuries and explored how determinants of pedestrian injuries differed across data sources. Comparing the pedestrian injury distributions across the three data sets, we noted that although the absolute number of events reported does not align perfectly (due to the nature of the data collection, e.g., crash vs. patient visit records), the relative frequencies and general distribution of major demographic and temporal/seasonal characteristics were remarkably similar across the data sets. As an exception to this, the distribution of events by age group in ED and police data did not align closely for non-fatally injured pedestrians, particularly among the age groups at either end of the age spectrum.
Due to the nature of the data collection processes, ED visit data overcounted pedestrian injury incidence (relative to police data), while police data undercounted pedestrian injury incidence (relative to ED data). Nevertheless, across all data sets, the calendar quarter from October-December had the highest frequency of pedestrian injury and fatality events. Male pedestrians were involved in more than 60 percent of the injury events (and 70 percent for the fatal events) for all data sources. For all data sources (at all levels of injury), crash rates were highest among 20-24 year olds, though fatal crash frequencies were highest among adults aged 40-59. Understanding pedestrian injury distributions such as these—and the data sources that
underlie them—is important for informing the development of pedestrian interventions such as the Watch for Me NC program.
Aim 2: The Watch for Me NC intervention (described in Chapter 4), was developed based on a number of behavioral theories and also accounted for pedestrian crash trends identified through a review of police-reported crash data. For example, the program was
designed to launch in August (with enforcement efforts peaking in October) and run through the end of the year, in accordance with trends documenting that time period as the highest-crash season of the year. The intervention involved 10 municipalities and within those, eight universities delivered the Watch for Me NC program on their campuses, helping to target the young adults most involved in pedestrian crashes.
To describe the intervention delivery (Aim 2.1), we obtained records of paid media, earned media, website usage, law enforcement operations, and public engagement activities. We found that communities and campuses with more staff—in particular those with devoted
pedestrian coordinators—and those with a longer history of commitment to pedestrian initiatives were more likely to support key intervention components, including public outreach and
enforcement, as measured by a range of implementation records. In this study, we also identified a set of measures that can be used by others to increase the consistency and comparability of multifaceted program delivery.
In addition to the evaluation of program delivery, we used a pretest-post-test comparative survey design to assess the effects of officer participation in a pedestrian safety training course on self-reported capacity to support the program and perform pedestrian safety operations (Aim 2.2). Results showed an increase in the number of correct responses regarding pedestrian and driver yielding requirements and an improved recognition of NC laws regarding pedestrians.
After the training courses, more officers stated that they had adequate resources, training, time, and ability to perform pedestrian operations, and several stated plans to conduct targeted enforcement in the next six months to a year. This indicated that providing training to officers was an important step in building organizational interest and ability to deliver enforcement- oriented pedestrian safety interventions.
The process results described in Chapter 4 provide important evidence of theoretical considerations and real-world lessons in intervention delivery, and complement the evaluation of program effects on driver behaviors (Aim 2.3) described in Chapter 5. For this final aim, we used a pre-post design with a control group, comparing locations receiving enforcement and low-cost engineering treatments with untreated locations to examine changes in driver yielding over a six- month period. We found that driver yielding rates improved (between four and seven percentage points on average) at locations enhanced by enforcement and engineering, while remaining unchanged at untreated sites. Some covariates were found to have strong associations with higher rates of driver yielding, in particular: high visibility crosswalk markings, crosswalks spanning fewer than three lanes, crossings on low-speed roads (i.e., posted speed limits of less than 30 miles per hour, or MPH), and crossings in smaller communities (less than 60,000 in population). An examination of potential modifiers found mixed results across naturalistic and staged
crossing types. We also found mixed results from an assessment of the relationship between driver yielding rates and Walk Score® (a summary index of built environment features such as land use and connectivity), which may affect driver yielding and pedestrian safety related outcomes in different ways.
Overall, this dissertation provides recommendations for making relevant comparisons between police, ED, and death certificate data, and provides a better understanding of the
discrepancies that exist between data sources. It suggests a set of process measures to increase the consistency and comparability of multifaceted intervention delivery, and provides evidence that enhanced enforcement/engineering, as a part of a broader program, can increase driver yielding to pedestrians in marked crosswalks. These results and lessons can guide researchers and decision-makers in developing and evaluating similar programs in other geographic locations.