It is increasingly difficult to find a profession untouched by artificial intelligence, or AI. The medical field is no exception. How are doctors using AI in clinical practice?
A U.S. research team analyzed 271 million questions asked over 2.5 years by approximately 950,000 healthcare professionals on OpenEvidence, an AI-powered clinical information search platform. Of all the healthcare professionals who asked questions, 556,200 were doctors, while the remainder consisted of nurses, pharmacists and medical support personnel, PA, among others.
Calculated based on monthly active users, nurses and medical support personnel, PA, asked an average of 44.5 questions per month, while doctors asked an average of 37.1.
When examined by years of experience, the doctors who used AI most actively were those who had only recently entered clinical practice. Doctors who had graduated from medical school within the previous 10 years accounted for 24.3% of the overall physician workforce, but 45.2% of AI users, and they accounted for 61.8% of all questions asked by doctors. Residents and fellows made up 14.8% of the physician workforce and asked 27.2% of all questions posed by doctors.
By specialty, AI use was active among oncology and hematology, which accounted for 1.9% of all doctors and 3.8% of questions, internal medicine, which accounted for 14.5% of all doctors and 27.3% of questions, and family medicine, which accounted for 12.5% of all doctors and 22.5% of questions. Specialties with lower usage rates included ophthalmology, radiology, orthopedics and anesthesiology.
An analysis of the questions showed that, among all healthcare professionals except pharmacists, questions about clinical decision-making were the primary purpose for using AI.
Among doctors’ questions, those about clinical decision-making accounted for 51.9%. Within clinical decision-making, questions about treatment and management, 33.1%, medication information, 14.1%, and diagnosis, 13.9%, were the most common.
The main purposes of questions also varied by medical specialty. For example, while questions about medication information accounted for 14.1% of all questions from doctors, they accounted for 29.1% of questions from psychiatrists. The share of questions about research and experimental results was higher among oncology and hematology specialists, at 3.5%, than among doctors overall, at 0.7%. Questions about clinical guidelines accounted for 2.1% of all questions from doctors, but the proportions were relatively high in gastroenterology, at 4.2%, and cardiology, at 3.6%.
In addition, doctors with more than 20 years of clinical experience asked fewer questions about treatment and management than doctors with less than 10 years of experience, while asking more questions about literature searches, risks and disease prognosis, and medication information.
The research team said, “Healthcare professionals have traditionally been expected to acquire vast amounts of knowledge through training and maintain it through continuing education,” adding, “The findings of this study show that a substantial part of the approach to knowledge is shifting from ‘memory’ to ‘search.’”
A U.S. research team analyzed 271 million questions asked over 2.5 years by approximately 950,000 healthcare professionals on OpenEvidence, an AI-powered clinical information search platform. Of all the healthcare professionals who asked questions, 556,200 were doctors, while the remainder consisted of nurses, pharmacists and medical support personnel, PA, among others.
Calculated based on monthly active users, nurses and medical support personnel, PA, asked an average of 44.5 questions per month, while doctors asked an average of 37.1.
When examined by years of experience, the doctors who used AI most actively were those who had only recently entered clinical practice. Doctors who had graduated from medical school within the previous 10 years accounted for 24.3% of the overall physician workforce, but 45.2% of AI users, and they accounted for 61.8% of all questions asked by doctors. Residents and fellows made up 14.8% of the physician workforce and asked 27.2% of all questions posed by doctors.
By specialty, AI use was active among oncology and hematology, which accounted for 1.9% of all doctors and 3.8% of questions, internal medicine, which accounted for 14.5% of all doctors and 27.3% of questions, and family medicine, which accounted for 12.5% of all doctors and 22.5% of questions. Specialties with lower usage rates included ophthalmology, radiology, orthopedics and anesthesiology.
An analysis of the questions showed that, among all healthcare professionals except pharmacists, questions about clinical decision-making were the primary purpose for using AI.
Among doctors’ questions, those about clinical decision-making accounted for 51.9%. Within clinical decision-making, questions about treatment and management, 33.1%, medication information, 14.1%, and diagnosis, 13.9%, were the most common.
The main purposes of questions also varied by medical specialty. For example, while questions about medication information accounted for 14.1% of all questions from doctors, they accounted for 29.1% of questions from psychiatrists. The share of questions about research and experimental results was higher among oncology and hematology specialists, at 3.5%, than among doctors overall, at 0.7%. Questions about clinical guidelines accounted for 2.1% of all questions from doctors, but the proportions were relatively high in gastroenterology, at 4.2%, and cardiology, at 3.6%.
In addition, doctors with more than 20 years of clinical experience asked fewer questions about treatment and management than doctors with less than 10 years of experience, while asking more questions about literature searches, risks and disease prognosis, and medication information.
The research team said, “Healthcare professionals have traditionally been expected to acquire vast amounts of knowledge through training and maintain it through continuing education,” adding, “The findings of this study show that a substantial part of the approach to knowledge is shifting from ‘memory’ to ‘search.’”
The findings of the study were recently published in the Journal of the American Medical Association, JAMA.