A systematic review of traffic safety data collection methods and challenges: From crash databases to AI-augmented...

Road crashes continue to be one of the leading causes of death and injury around the globe. Yet, despite decades of studying the problem, safety surveillance is still largely dependent on crash reports that are at best incomplete and at worst outdated, leaving daily risk in critical blind spots.

This systematic review synthesizes 89 peer-reviewed studies published between 1995 and 2024. We grouped these studies into five categories: (a) crash databases, (b) automated imagery, (c) onboard and mobile sensors, (d) naturalistic and simulated studies, and (e) AI-augmented approaches. They measured each method on four major axes: monetary cost, scale-up cost, data granularity, and ethics.

The findings show a clear space of trade-off between traditional databases, which are still relevant and used for long-term monitoring in the field, but lack detail on behavior, and sensors and AI-based approaches, which provide more detailed and real-time information on individual-level data but are plagued by cost, privacy, and accessibility issues. Some gaps remain, such as underreporting in police statistics, privacy concerns in naturalistic research, and technological inequalities in low- and middle-income countries. This review, therefore, brings together methodological considerations and outlines steps for identifying a contextualized approach.

Together, these findings highlight the importance of creating scalable and ethical paradigms that shift traffic safety research from a response-based to a risk-based framework. To facilitate the systematic comparison of these various approaches, we compile evidence within a Strengths-Weaknesses-Opportunities-Threats (SWOT) based framework that highlights the trade-offs among methods.

These review results are intended to inform future research and eventual outcomes related to road safety and reduced traffic crashes.