Building a local public health firearm injury data infrastructure: a Philadelphia, PA case study

Data Analysis

The poster below was presented by me and my colleagues from the Philadelphia Department of Public Health (PDPH) Injury Prevention team at the Society for the Advancement of Violence and Injury Research (SAVIR) conference in New York, April 2025. We demonstrated feasibility of a firearm injury integrated data system -- the first of its kind in Philadelphia -- by linking hospital-based violence intervention program (HVIP) client data with Philadelphia Police Department (PPD) shooting records using Splink, a probabilistic record linkage Python package. The goal of this system is to represent, in near real time, all firearm injuries in Philadelphia.

Eventually, multiple data sources identifying those who have been injured will be incorporated, including hospital records and additional community violence interruption program client data. PPD data were used first because they are the most comprehensive firearm injury data that our team currently has access to; however, we acknowledge that relying heavily on police data creates a biased understanding of firearm violence. This integrated data system will ideally allow for a more holistic and public health-oriented understanding of firearm violence. Supplemental data meant to better understand firearm injury determinants, such as human services and housing data, will also be incorporated.

At the time of this presentation, the Splink model was performing well, with 74% of HVIP clients matching to their corresponding firearm injury recorded by PPD. Since this presentation, we began receiving all firearm injury data from PPD, including "non-criminal" shooting records, and incorporating firearm fatality data from the Medical Examiner's Office (MEO) in the data system. The data system shows promise; once more fleshed out, it will be invaluable to our team at PDPH as well as the violence intervention programs we support.