Identifying Type II workplace violence from clinical notes using natural language processing

Area of Science:

• Healthcare safety
• Natural Language Processing
• Workplace violence

Background:

• Type II workplace violence by patients towards home healthcare nurses is a critical health and safety concern.
• A substantial number of these violent incidents go unreported through official channels.
• Natural language processing (NLP) offers a method to identify underreported violence from clinical notes.

Purpose of the Study:

• To calculate the 12-month prevalence of Type II workplace violence experienced by home healthcare nurses.
• To develop and implement an NLP system for detecting unreported workplace violence in clinical notes.

Main Methods:

• Analysis of nearly 600,000 clinical visit notes from two U.S. home healthcare agencies (January 1, 2019 - December 31, 2019).
• Application of rule-based and machine-learning NLP algorithms to identify descriptions of workplace violence.
• Comparison of NLP findings with official incident report data.

Main Results:

• NLP identified 236 clinical notes detailing Type II workplace violence against home healthcare nurses.
• Prevalence rates per 10,000 home visits: physical violence (0.067), nonphysical violence (3.76), any violence (4.0).
• Official incident reports documented zero Type II workplace violence incidents during the study period.

Conclusions:

• NLP is effective in uncovering "hidden cases" of workplace violence by analyzing clinical notes.
• NLP can supplement formal reporting systems, providing a more accurate picture of violence risks.
• Utilizing NLP can help healthcare managers and clinicians enhance safety in practice environments.