Patient aggression in hospitals contributes to nurse burnout and workforce attrition. Emerging AI-driven vocal biomarker technology may provide early detection, improving safety, response times, and staff well-being in clinical environments.
By Kang Hsu, Jr. MD, Chief Medical Officer, Canary Speech
Provo, Utah — October 27, 2025
Patient aggression in the hospital is a growing problem. Among nurses and other staff, violent or abusive behaviors have been linked to burnout and PTSD — two of the many reasons nurses have left the medical profession in recent years by the thousands. Preventing aggressive behavior before it can begin is perhaps the most effective, and elusive, strategy for reversing the trend.
Here’s a closer look at the problem of patient aggression, and an emerging tool that builds on technology already in use in clinical settings.
Aggression in hospitals: a growing problem
The increase in aggressive behavior toward hospital staff — mostly nurses, but also other caregivers — has been well-documented.
A 2023 survey conducted by the National Nurses United (NNU) showed nearly half of nurses (45.5 percent) reported an increase in workplace violence in the previous year.
Another study published in April revealed more than 138,000 nurses have left the workforce since 2022; more than 41 percent cited stress and burnout as the root causes. Psychiatric nurses and ER staff are particularly subject to the dangers of patient aggression.
Despite awareness of the problem, and de-escalation tactics that have arisen in response, 43 percent of nurses surveyed in 2023 reported experiencing abuse or violence in the workplace in the previous year. Patients are not the only perpetrators of aggressive behaviors, but eliminating or at least reducing it would improve clinicians’ working conditions dramatically.
The limits of today’s solutions
Many nurses have been trained on de-escalation methods, but these have their limits. The data tells the story. If today’s de-escalation methods were sufficient, violence toward clinical staff would not be on the rise.
Sometimes, a nurse trained on de-escalation might not be aware of an incident in time to de-escalate it. If a patient is being aggressive with another patient in the same room, for example, by the time a staff member arrives it’s too late. Even if no one shares the room with the aggressive patient, de-escalation gets more challenging as time passes, increasing the likelihood of a violent or abusive incident when a clinician walks in.
Nurses are often the first line of defense when patient behavior turns aggressive, but that doesn’t mean they are the best equipped to handle the situation. Unfortunately, many hospitals and health systems lack the money to pay for additional staff, security personnel, and monitoring tools that would reduce the burden and risk to nurses. Still, many health systems haven’t established adequate workflows to deal with an employee being attacked or abused on-site.
The power of vocal biomarkers
The human voice is a rich source of data. More than 2,500 distinct speech features have been identified, and are able to be instantly analyzed by today’s digital tools via ambient listening.
AI-based vocal biomarker analysis is already being used in a variety of clinical settings to detect signs of behavioral and cognitive conditions. These same platforms can be applied to detect aggression.
How?
Imagine a nurse is making their rounds, bringing medicine to patients. If one patient has been speaking aggressively — to another patient in the same room, into a cell phone, etc. — the same ambient listening tools used to record a doctor-patient conversation can detect aggression in the patient’s voice. Depending on the algorithm’s settings, it can alert support staff, security, or other personnel to de-escalate the situation prior to the nurse’s arrival.
Currently that same nurse might have little to no clue if a patient is talking aggressively, or using violent or abusive language, before they walk into a room. By the time they arrive, it might be too late to de-escalate.
Imagine also the patient who becomes aggressive while talking to a nurse in their room. Currently, calling for security is a multi-step process for that nurse. An algorithm that detects aggression in the patient’s voice can save precious time, signaling for security while the nurse does nothing more than talk to the patient.
The vocal biomarker detection technology needed to identify aggressive speech already exists. So too does the kind of AI-based ambient listening platforms needed to process that data in real time. Now it’s just a matter of applying it in high-risk settings to prevent incidents before they arise. As violent behavior threatens nurses’ well-being and continues to lead to attrition, the time is right to deploy the technology.
Challenges and Considerations
There are significant hurdles before this technology can be widely deployed:
Companies like Canary Speech are exploring feasibility of vocal biomarker detection technology in collaboration with healthcare providers, but independent validation and clinical trials are needed. If proven effective, such tools could become part of broader strategies to improve workplace safety, reduce burnout, and support staff in high-risk hospital environments.
Preventing burnout and protecting our nurses isn’t just about adding resources, but its also about using technology intelligently to provide early warning and actionable insights that make the hospital a safer place for everyone.
Kang Hsu, Jr., MD is the Chief Medical Officer (CMO) of Canary Speech, a Provo, Utah-based AI-powered developer of a speech recognition technology designed to detect neurological and cognitive diseases previously covered by TechBuzz. Canary Speech's technology monitors speech irregularities for early signs of various neurological conditions such as Parkinson's, dementia and Alzheimer's. It analyzes the neurological conditions of the patients, enabling the identification of early warning and early diagnosis of diseases that affect their cognitive functions.
Dr. Hsu oversees Canary Speech’s clinical direction, supports product development, and works closely with strategic partners including Mayo Clinic, Microsoft, Samsung, LG NOVA, and major health systems. He also contributes to the company’s research efforts, collaborating with academic and industry partners to help advance the scientific development and clinical validation of voice biomarkers.
TechBuzz welcomes contributions from operators, investors and government officials from the local community.
Learn more about Canary Speech here.
See full 2023 National Nurses United Report about nurse staffing crisis below: