The artificial intelligence (AI) era in medicine is well underway, and it’s hard to imagine a more suitable showcase for the potential of AI than clinical care. With its layers of complexity, embodied in the classic steps of presentation, diagnosis, and treatment, medical practice should only benefit from what I currently see as AI’s most important contribution to our field—more and better information. Among other things, AI has already shown promise in medical image interpretation, a boon for multiple specialties, including oncology, gastroenterology, and cardiology.1 The deployment of large language models to help manage and integrate administrative data and information in its myriad forms should improve efficiency and patient care.2 

It is clear that AI, a new set of highly sophisticated, ever-evolving tools, is going to change many things about the way medicine is done.3 What is not going to change, however, is the very human experience of ill and injured people seeking help from strangers, which is what we are to most of those who come through our doors. Patients in pain do not really care about the cool new systems and methods we use to figure out what is wrong with them and how to fix it. They just want to feel better. They’re grateful for our help, of course. But, for many, the hospital is an inherently stressful place where everybody but the patient seems to be in control. Busy and crowded, with unavoidable delays, it is exactly the wrong setting for patients seeking compassionate family-centered medical and surgical care and for staff who are frustrated and stressed by their work environment.

One troubling aspect of this inescapable truth is examined in this issue of Pediatrics by Waltzman et al,4 who developed an AI tool to identify unreported incidents of verbal and physical violence against nurses. Such outbursts are all too common, as any health care provider can confirm. They are also routinely regarded as an unfortunate fact of life on the clinical frontlines and notoriously underreported.

To address this problem, the researchers used natural language processing, a type of AI that scans text to assess meaning, to help review of more than 19 000 nursing notes involving 2827 inpatients at an urban community hospital. The notes covered the 6-month period from July through December 2022. Of the 26 violent episodes identified that would be considered as workplace violence according to the hospital’s criteria, just 7 had been reported.

I am not surprised that 19 incidents, 73% of the total, went unreported. I do not think I ever reported any of my own experiences with violence, yet there were plenty. I remember once being threatened by a family who had to wait a long time for care, and then being told by them that they were going to go home and get a gun. That was an empty threat, fortunately. Although no one every physically hit me, people would push me and scream at me. I am over 6 feet tall, and was 30 years younger then, but that did not stop patients or their loved ones from lashing out at me. Such behavior remains an issue today for health care workers of every kind, and some have it worse than I did.

The Waltzman et al study reflects the central role nurses play in health care and the outsized risk they face when they are on the job. It is the nurses who spend the most time and have the most interaction with patients and families, caring for them and documenting the daily course of clinical events. They are there at the beginning, when much is still unknown and anxiety levels are high; they are there when the news is bad; they are there when there is no news and the hours drag on. And they are there when the worry and frustration peak, and people snap. Sixty percent of respondents to the American Nurses Foundation 2022 workplace survey reported 1 or more incidents of bullying or incivility in the past year, whereas 29% reported 1 or more incidences of violence.5 Another report estimated that US nurses experience nearly 60 assaults every day.6 

Emergency departments have long been recognized as high-risk zones for violence directed at nurses, doctors, and other health care providers, as anyone who has ever worked in one knows all too well.7 According to the American College of Emergency Physicians, the situation is only getting worse. Fully 85% of American College of Emergency Physicians members surveyed in 2022 said emergency department violence had increased in the previous 5 years, and 45% said it had greatly increased.8 Most of the assaults were verbal, but the doctors also reported being slapped, spit on, kicked, and scratched. Sadly, 89% said violence had harmed patient care.

How could it not affect patient care? Beyond the immediate shock, fear, and anger, a violent episode can make you more guarded with patients, less friendly, and more cautious, none of which are helpful when it comes to forging a strong therapeutic bond. At the institutional level, it is bad for morale and contributes to burnout.

The big challenge is how to balance being warm and welcoming with an appropriate level of safety and security. At Montefiore, we have deescalation training to teach our staff how to deal with angry, upset people. Because reporting incidents is so important,9 we encourage people to make verbal reports to a supervisor or onsite security and to use the proprietary electronic software reporting system. The electronic reports are picked up by our compliance, legal, and security departments, as well as human resources. When there is an incident, we go out of our way to take prompt action. We have support teams for employees who have experienced trauma. And patients who have been abusive repeatedly with staff are put on a no-fly list, meaning their future nonemergent care needs can be refused. We make exceptions for patients with mental health disorders and take special measures with them to make sure that everybody is safe.

Our health system, like others around the country, has installed an advanced weapons detections systems when patients enter our facilities and has removed weapons and sharp objects from patients before their being registered to be seen or friends or relatives coming to see a loved one who is in one of our hospitals. Although this can help avert serious incidence with deadly weapons, it does not stop the more common and insidious and often underreported violence that happens on a regular basis.

Like a bin full of confiscated weapons, the details described here, in the Waltzman et al paper, and in the many other papers about violence in health care settings,10 offer a decidedly grim picture of hospital life. But facing difficult truths is what we do in medicine. We face them and we deal with them. What Waltzman et al have done is to take a fresh look at an old problem with a new tool: natural language processing. And it worked. The technology used uncovered important, overlooked information revealing unreported violence in the workplace, using a method with the potential to make us all safer. Good research raises as many questions as it answers. What else can we learn about the circumstances that lead to violent outbursts by patients? What other contributing factors, missed warnings and telltale patterns of behavior are buried in the sea of notes, reports, and test results that clinical care generates? I am certain that it cannot hurt to ask these questions so we can get better answers to them than currently exist.

COMPANION PAPER: A companion to this article can be found online at www.pediatrics.org/cgi/doi/10.1542/peds.2023-063059.

AI

artificial intelligence

1
Rajpurkar
P
,
Chen
E
,
Banerjee
O
,
Topol
EJ
.
AI in health and medicine
.
Nat Med
.
2022
;
28
(
1
):
31
38
2
Thirunavukarasu
AJ
,
Ting
DSJ
,
Elangovan
K
,
Gutierrez
L
,
Tan
TF
,
Ting
DSW
.
Large language models in medicine
.
Nat Med
.
2023
;
29
(
8
):
1930
1940
3
Secinaro
S
,
Calandra
D
,
Secinaro
A
,
Muthurangu
V
,
Biancone
P
.
The role of artificial intelligence in healthcare: a structured literature review
.
BMC Med Inform Decis Mak
.
2021
;
21
(
1
):
125
4
Waltzman
ML
,
Ozonoff
AI
,
Fournier
KA
, et al
.
Surveillance of health care-associated violence using natural language processing
.
Pediatrics
.
2024
;
154
(
2
):
e2023063059
5
American Nurses Foundation
.
Pulse on the Nation’s Nurses Survey Series: 2022 Workplace Survey. Nurses not feeling heard, ongoing staffing and workplace issues contributing to unhealthy work environment
. Available at: https://www.nursingworld.org/~4a209f/globalassets/covid19/anf-2022-workforce-written-report-final.pdf. Accessed April 25, 2024
6
Carbajal
E
.
2 nurses assaulted every hour, Press Ganey analyses shows
. Available at: https://www.beckershospitalreview.com/nursing/2-nurses-assaulted-every-hour-press-ganey-analysis-shows.html. Accessed April 25, 2024
7
Aljohani
B
,
Burkholder
J
,
Tran
QK
,
Chen
C
,
Beisenova
K
,
Pourmand
A
.
Workplace violence in the emergency department: a systematic review and meta-analysis
.
Public Health
.
2021
;
196
:
186
197
8
Marketing General
.
ACEP emergency department violence poll results
. Available at: https://www.emergencyphysicians.org/siteassets/emphysicians/all-pdfs/acep-emergency-department-violence-report-2022-abridged.pdf. Accessed April 25, 2024
9
Rosenthal
LJ
,
Byerly
A
,
Taylor
AD
,
Martinovich
Z
.
Impact and prevalence of physical and verbal violence toward healthcare workers
.
Psychosomatics
.
2018
;
59
(
6
):
584
590
10
The International Association for Healthcare Security and Safety (IAHSS) Foundation
.
2023 Healthcare Crime Survey
. Available at: https://iahssf.org/assets/2023-Healthcare-Crime-Survey.pdf. Accessed April 25, 2024

Competing Interests

CONFLICT OF INTEREST DISCLOSURES: The author has indicated he has no potential conflicts of interest to disclose.