This study sought to evaluate the effectiveness of the Florida Elder Road User Program by building a crash prediction model that examines older driver crash frequency using variables obtained from quantitative variables such as maximum posted speed limit and qualitative variables such as sign visibility. Crash data was collected for 601 intersections in the state of Florida for the years 2000-2002. Detailed analysis of geometric and traffic data collected from various resources, such as crash reports, Roadway Characteristics Inventory data, and video log archives, was performed to determine the correlation between these independent factors and crash frequency. The crash distribution was found to be negative binomial; the model needed to be inflated due to the presence of many zero-crash intersections. Therefore, a Zero-Inflated Negative Binomial (ZINB) crash prediction model was developed. The response variable in the model was crash frequency, and independent variables included roadway geometrics, traffic characteristics, and intersection characteristics. Crash frequency was found to have a very strong relationship with most of the independent variables as indicated by their p-values. Traffic volumes, signalization, and sign visibility had positive coefficients, indicating that, higher traffic volumes, signalized intersections, and sign visibility were associated with increased crash frequency. Many of the improvements, including sign visibility and reflective pavement markers listed in the Florida Elder Road User Program were found to be significant but did not reduce elderly driver crashes, proving that the program is not effective. (Author/publisher)
Abstract