Background While road network structure is key to transport planning, limited research explores the safety implications of network structure. Intersections, in particular, are of concern due to complex and conflicting traffic movements. The angle of intersection skewness (deviation of the intersection angle from 90°) and the categories of intersecting roads are crucial factors influencing intersection safety. While existing research extensively examines these factors at the micro level, there is a notable gap in investigating their impact at the macro level on traffic safety.

Objectives This study aims to (a) quantify intersection skewness along with the entropy of intersections based on road category at the macro level and (b) develop macro-level safety models at the city level to investigate the relationship between road traffic crashes and network structure.

Methods We studied 51 Indian cities with a population of more than a million. We used OpenStreetMap for the road network and the government-reported number of road deaths for the years 2018–19, which included a total of 30,127 fatalities. We employed commonly used safety analysis metrics- including meshedness coefficient (MC), betweenness centrality (Bc), closeness centrality (Cc), and average circuity (Cavg). In addition, we proposed metrics for intersection skewness (φ) and heterogeneity of intersecting leg categories (μ). We modeled the crash rate at the city level using a negative binomial regression model. The explanatory variables include various network structure metrics (Mc, Bc, Cc, Cavg, φ, and μ), number of registered vehicles, nightlight intensity (indicating economic status), and population density.

Results and Conclusions It was observed that cities with a higher deviation from the conventional 90° intersection experienced a higher rate of fatal crashes. Moreover, cities with a greater number of intersections where different road categories intersect (resulting in higher intersection leg entropy) experienced a higher rate of fatal crashes. However, cities with more central (higher Betweenness Centrality) road networks had fewer crashes. These effects were found after controlling for the number of vehicles, population density.