San Francisco Traffic Injury Crashes with Relational Party and Victim Tables

Quarterly updated data on all crashes resulting in an injury within San Francisco, sourced from police reports and medical examiner records. The dataset has a relational structure with separate tables for crashes, involved parties, and victims, and includes geocoded locations. It is maintained by the City of San Francisco and aggregates data from multiple systems, including SWITRS and SFPD's Interim Collision System.

Use Cases

• Analyze spatial patterns of injury crashes based on geocoded location data.
• Study injury severity distributions using the relational victim table.
• Model risk factors for different road user types (drivers, pedestrians, bicyclists) based on party involvement data.
• Assess the effectiveness of traffic safety policies over time using quarterly aggregated data.

Strengths

• Data is aggregated from multiple authoritative sources, including the California Highway Patrol's SWITRS system and the San Francisco Police Department.
• Relational structure separates crash, party, and victim information, which likely allows for detailed analysis.
• Includes geocoded locations mapped to a city-maintained street centerline model.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Crashes on highways and at complex intersections are excluded or simplified, which may limit analysis for certain locations.
• Row count is unknown, which may limit suitability assessment.

Provenance

Source

City of San Francisco, aggregating data from SFPD, California Highway Patrol (SWITRS), and the Office of the Chief Medical Examiner.

Collection Method

Data is extracted from police officer crash reports, de-identified, geocoded, and loaded programmatically.

Time Range

Historical data prior to 2013, with specific systems for 2013-2017 and 2018 to present.

Freshness

Updated quarterly; last metadata update was 2026-03-08.

Geography

City of San Francisco, excluding highways.