Characterizing Violent Injuries to Inform Prevention among Healthcare Workers
 
Dr. Devin Hawkins, DrS describes the initial report of a project that is designed to characterize and prevent violent injuries among healthcare workers. National Bureau of Labor Statistics data show healthcare and social assistance workers have the highest severe workplace-violence injury rates, with rising trends before COVID-19 and apparent acceleration during and after the pandemic. Risk varies by setting and job (e.g., nursing/residential care facilities; nurses and aides), and disparities are evident, with higher rates among Black healthcare practitioners and support workers. Using IRB-approved data from OSHA 300 injury logs from 14 Massachusetts hospitals (2022–2024), the research team identified 333 violent incidents among 5,035 total reported injuries/illnesses, then applied an AI/large language model schema to extract circumstances from narrative descriptions (injury type, body part, context, location, triggers, patient state, team response, weapons/objects, outcomes, and medical actions). Early patterns include being struck/kicked/punched, bites, grabbing/twisting, and injuries to the face and upper extremities; events often occur during routine care, ambulation/bathroom assistance, restraint/security interventions, or patient elopement. The team is refining categories, measuring agreement between automated and reviewed coding, and plans to translate results into transparent SAS code and test on future data (2024–2025), culminating in a public report. The presentation highlights prevention-oriented policies, including California’s healthcare violence-prevention standard and a pending Massachusetts bill emphasizing risk assessment, prevention plans, training, reporting, and worker support. Health Watch USA(sm) meeting Mar. 19, 2026. View Youtube Video: https://youtu.be/tYz0X4dFKjM

Key points
• Healthcare and social assistance workers experience the highest rates of severe workplace-violence injuries nationally, with increases seen pre-COVID and during and after the pandemic.
• Violent-injury risk varies substantially by setting and occupation (e.g., nursing/residential care; nurses, aides), and contributes to disparities (higher rates among Black healthcare workers in multiple groups).
• The project used IRB-approved OSHA 300 injury log data from 14 Massachusetts hospitals (2022–2024): 5,035 total reported injuries/illnesses, including 333 violent incidents.
• Violent cases were identified using established SAS code (NIOSH/Stephen Burke) plus manual review to capture misses.
• A large language model–based schema is used to extract structured details from free-text narratives (injury type, body part, context, location, triggers, patient state, response, weapons/objects, outcomes, medical action).
• Early recurring patterns include strikes (kick/punch), bites, grabbing/twisting, and injuries to the face and upper extremities; common contexts include routine care and ambulation/bathroom assistance.
• Additional themes include incidents during restraint/security events, preventing patient elopement, disorientation/agitation on waking or post-anesthesia, and variability in narrative detail (sometimes minimal descriptions).
• Next steps include refining categories, checking agreement between automated and reviewed coding, translating to transparent SAS code, validating on 2024–2025 data, and producing a public report to inform prevention and policy.   
 
View Youtube Video: https://youtu.be/tYz0X4dFKjM