Characterizing the Mobile App Ecosystem for Traffic Safety Using the Haddon Matrix: A Secondary Analysis and Theory-Based Classification

Muhammad Zafar Iqbal Hydrie^1*Owais Raza¹Nadeem Mahmood²Faiza Sattar¹

• ¹School of Public Health, Dow University of Health Sciences, Karachi, Pakistan
• ²University of Karachi, Karachi, Pakistan

Background: Road traffic accidents remain a significant public health concern. With the rapid expansion of mobile technology, smartphones have emerged as tools for both risk and prevention, including apps designed to enhance road safety and emergency response. Guided by the Haddon Matrix, our study applies a structured, quantitative analysis of road safety apps to identify gaps and opportunities in leveraging mobile technology for injury prevention. Methods: We conducted a secondary analysis of mobile app data originally collected by Aghayari et al. (2021). Apps were classified with a structured, keyword-based algorithm according to the Haddon Matrix, which cross-tabulates Human, Vehicle, Physical Environment, and Social Environment factors against Pre-event, Event, and Post-event stages. Descriptive analyses and multidimensional visualizations (Sankey and chord diagrams) were used to characterize distributions and co-occurrences across domains. Sensitivity analyses were performed to test the robustness of classifications under varying thresholds and simulated text perturbations. Results: Among 912 road safety apps, the largest share belonged to Navigation (40.6%), followed by Education (20.7%) and Travel (20.5%). Most apps were classified under the Physical Environment factor (71.3%), while Human, Social Environment and Vehicle domains were less represented. With respect to injury prevention stages, half of the apps addressed the Event stage (50.9%), compared to 30.9% for Pre-event and 7.5% for Post-event. Sankey and chord visualizations revealed that end-user categories mapped predominantly to Physical Environment features with limited integration of Human or Social domains. Sensitivity analyses showed robust classifications, with ~80% agreement even when 15% of keywords were randomly removed. Conclusion: The dual visualization approach, in which we combined hierarchical Sankey diagrams with network-based chord diagrams, revealed complex relationships that were invisible in traditional categorization methods. From a public health perspective, our findings highlight the need for more systematic approaches to mobile health technology development, evaluation, and policy guidance for road safety.