Accident data analysis and on-field inspections: do they lead to similar conclusions?

Persia, Luca; Gigli, Roberto; Azarko, Anastasia; Usami, Davide Shingo · 2020 · OpenAlex-citations

DOI: 10.1016/j.trpro.2020.03.016

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This study investigates whether Network Safety Ranking (NSR), which relies on historical accident data, and the Road Assessment Programme (RAP), which relies on infrastructure inspections, yield similar conclusions regarding road safety. Motivated by the shift toward proactive "Safe System" approaches, the authors aim to determine if RAP indicators, specifically the Road Protection Score (RPS) from EuroRAP, can effectively predict accident frequencies and enhance the explanatory power of accident prediction models. This is particularly relevant for contexts where accident data is unavailable, such as in low-income countries or during the design phase of new roadways. The methodology involved calibrating and analyzing 96 pairs of homologous regression models using data from approximately 400 km of motorways in Lazio, Italy. The dataset combined historical injury accident records (2008–2011), EuroRAP inspection scores, and traffic flow data. To ensure robustness, the authors created six datasets with varying aggregation levels (1 km, 2–5 km, and traffic-based sections) and calculation methods for RPS (maximum vs. average values). Both Poisson and Negative Binomial models were calibrated to address potential overdispersion. The study compared models with and without EuroRAP indicators as covariates, using the Likelihood Ratio Test (LRTest) to assess statistical significance and the Akaike Information Criterion (AIC) to evaluate model fit. The results indicate that including EuroRAP indicators significantly improves the explanatory power of accident prediction models. Of the 96 model pairs, 25 (26%) met strict criteria for both statistical significance (p < 0.05) and good fit (AIC/n < 6.63). Models predicting total accident frequencies performed best, particularly when using 1 km road sections. EuroRAP indicators were most effective in explaining accidents categorized by the EuroRAP framework (head-on, run-off, and intersection accidents). However, the indicators showed limited predictive power for fatalities and injuries when analyzed as standalone dependent variables. The study also noted that Poisson models were generally sufficient, except where overdispersion was significant. The findings suggest that EuroRAP indicators are valuable tools for predicting accident frequencies and can potentially replace traditional covariates like geometric features in safety models. This supports the use of inspection-based assessments for proactive safety management. However, the authors highlight a limitation: EuroRAP does not address all accident types, such as rear-end collisions on motorways. Consequently, while EuroRAP enhances model performance for specific crash types, its explanatory power may increase if additional accident categories are incorporated into the assessment framework.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• incidence prevalence
• crash typology
• comparative international

Information type

What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).

• Empirical Findings: crash risk outcomes
• Methodological Resource: dataset resource