In a new study published online today in JAMA Network Open, researchers used artificial intelligence to analyze more than 1.4 million electronic health record (HER) emails to physicians — and the results aren’t pretty.
Among the emails, 43% were from patients; the remainder were mostly from other physicians or clinicians, or automated. The content of the messages wasn’t associated with doctor burnout, as the researchers had hypothesized. And only about 5% of the messages had negative sentiment.
But the researchers were struck by the hostility of that sentiment, displayed in messages like these that surely would be distressing for physicians to read:
“I hope and expect that you will spend eternity in he**. You are an abusive, nasty, cheap person.”
“Your office is full of liars, hypocrites and I will do everything in my power to prevent anyone from going to your bullsh** office again.”
About 5% of emails had an overall negative sentiment, with high-frequency words like “cancel,” “pain,” or “problem.” Among patient messages, 3% were negative and contained words and expletives suggesting hatred, hostility, or violence.
“F***” was the most common expletive used by patients.
Researchers provided examples of profanity-laced messages, including one patient who said, “I am so upset that I was told the blood work would include the gender of the baby. I have been waiting 5 [days] to find it, and it wasn’t even fu**ing tested!!!! What a disappointment in your office and the bullsh** I was told. I will be switching plans because this is sh**!”
Researchers also noted some high-frequency words associated with violence, such as “shoot,” “fight,” and “kill.”
“This is concerning, especially given documentation of patient-inflicted violence against physicians. Health systems should be proactive in ensuring that the in-basket does not become a venue for physician abuse and cyberbullying,” the researchers wrote.
“Posting reminders in EHR patient portals to use kind language when sending messages, applying filters for expletives or threatening words, and creating frameworks for identifying patients who frequently send negative messages are potential strategies for mitigating this risk.”
Using a form of artificial intelligence technology called natural language processing (NLP), researchers at the University of California San Diego analyzed the characteristics of more than 1.4 million emails received by the university’s physicians, 43% of them from patients. They specifically looked at the volume of messages, word count, and overall sentiment.
Whereas other studies have examined the growing burden of EHR messaging for doctors, this type of email sentiment analysis could help in creating solutions. Researchers say that one such solution could involve applying filters for expletives or threatening words. It also could help identify fixable health system issues that make patients so angry, the researchers say.
Among the emails from physicians to physicians, just over half reported burnout, which correlated to the following phrases: “I am beginning to burn out and have one or more symptoms of burnout” and “I feel completely burned out [and] am at the point where I may need to seek help.”
On average, physicians who reported burnout received a greater volume of patient messages. The odds of burnout were significantly higher among Hispanic/Latinx physicians and females. Physicians with more than 15 years of clinical practice had markedly lower burnout.
Despite physicians now spending more time on EHR in-basket tasks than they did before the pandemic, the study found no significant associations between message characteristics and burnout.
Data for the cross-sectional study were collected from multiple specialties from April to September 2020. Physicians then completed a survey and assessed their burnout on a 5-point scale. Of the 609 physician responses, approximately 49% of participants were women, 56% were White, and 64% worked in outpatient settings. About 70% of the doctors had been in practice for 15 years or less.
The sentiment score was based on word content as well as the use of negation, punctuation, degree modifiers, all caps, emoticons, emojis, and acronyms. Positive patient messages were more likely to convey gratitude and thanks, along with casual expressions, such as “fyi” and “lol.”
Steph Weber is a Midwest-based freelance journalist specializing in healthcare and law.