November 10, 2023 – You can have used ChatGPT-4 or certainly one of the opposite latest artificial intelligence chatbots to ask an issue about your health. Or perhaps your doctor uses ChatGPT-4 to create a summary of what happened at your last visit. Your doctor may even have the diagnosis of your illness checked again using a chatbot.
But at this stage in the event of this latest technology, experts say each consumers and doctors could be clever to proceed with caution. Despite the reliability with which an AI chatbot delivers the requested information, it shouldn't be all the time accurate.
As the usage of AI chatbots spreads rapidly, each in healthcare and elsewhere, there are increasing calls for the federal government to control the technology to guard the general public from the potential unintended consequences of AI.
The federal government recently took a primary step on this direction under President Joe Biden issue an implementing regulation This requires government agencies to seek out ways to control the usage of AI. In the world of healthcare, the order directs the Department of Health and Human Services to advance responsible AI innovations that “promote the well-being of patients and healthcare workers.”
Among other things, the agency should arrange a task force for health AI inside a 12 months. This task force will develop a plan to control the usage of AI and AI-enabled applications in healthcare, public health, and drug and medical device research and development and safety.
The strategic plan may even address “the long-term security and real-world performance monitoring of AI-powered technologies.” The department must also develop a way to find out whether AI-powered technologies “maintain an appropriate level of quality.” And, working with other government and patient safety organizations, health and social services must create a framework to discover errors “resulting from the use of AI in clinical settings.”
Biden's executive order is “a good first step,” said Ida Sim, MD, PhD, professor of medication and computational precision health and chief research informatics officer on the University of California, San Francisco.
John W. Ayers, PhD, associate director of informatics on the Altman Clinical and Translational Research Institute on the University of California San Diego, agreed. He said that while the healthcare industry is subject to strict oversight, there aren't any specific regulations for the usage of AI in healthcare.
“This unique situation arises from the fact that AI is evolving rapidly and regulators cannot keep up,” he said. However, it will be significant to tread fastidiously on this area or latest regulations could hinder medical progress, he said.
The problem of “hallucination” concerns the AI
In the 12 months since ChatGPT-4's launch, the chatbot and others impressed with its human-like conversation and knowledge on a wide selection of topics have firmly established themselves amongst healthcare professionals. Fourteen percent of doctors according to a survey Already use these “conversational agents” to diagnose patients, create treatment plans and communicate with patients online. The chatbots are also used to compile information from patient records before visits and summarize visit notes for patients.
Consumers have also began using chatbots to go looking for health information. Understanding insurance benefit noticesand to research numbers from laboratory tests.
The fundamental problem with all of that is that the AI chatbots should not all the time right. Sometimes they devise things that don't exist – they “hallucinate,” as some observers call it. According to a current study by VectaraAccording to a startup founded by former Google employees, chatbots provide information at the least 3% of the time – depending on the bot, even 27% of the time. Another report got here to similar conclusions.
That's to not say that chatbots aren't great at finding the proper answer more often than not. In one attempt33 doctors in 17 specialties asked chatbots 284 medical questions of various complexity and evaluated their answers. More than half of the answers were rated as almost correct or completely correct. However, the answers to fifteen questions were rated as completely incorrect.
Google has created a chatbot called Med-PaLM that's tailored to medical knowledge. This chatbot, which has passed a medical licensing exam, has a 92.6% accuracy rate when answering medical questions, which is concerning the same as doctors. according to a Google study.
Ayers and his colleagues conducted a study Comparing chatbots' and doctors' responses to questions asked by patients online. Healthcare professionals evaluated the responses and preferred the chatbot response over the physician response in nearly 80% of exchanges. Doctors' responses were rated lower when it comes to each quality and empathy. The researchers suspected that the doctors can have been less empathetic due to practice stress they were under.
Garbage in, garbage out
Chatbots could be used to discover rare diagnoses or explain unusual symptoms, and may also be consulted to make sure doctors don't miss obvious diagnostic opportunities. To be available for these purposes, they must be embedded right into a clinic's electronic medical record system. Microsoft has already done it embedded ChatGPT-4 in Epic Systems' most generally used health records system.
A challenge for any chatbot is that the records contain some misinformation and sometimes data is missing. Many diagnostic errors are related to poorly recorded patient histories and incomplete physical examinations documented within the electronic health record. And these records typically don't contain much or any information from the records of other physicians who've seen the patient. Simply due to insufficient data within the patient's record, it might probably be difficult for a human or artificial intelligence to attract the proper conclusions in a given case, Ayers said. This is where a physician's experience and knowledge of the patient could be invaluable.
But chatbots are quite good at communicating with patients, Ayers' study showed. Under human supervision, he said, it's likely that these conversational agents could help relieve doctors of the burden of online messaging with patients. And he said this might improve the standard of care.
“A conversational agent isn’t just something that can handle your inbox or the load on your inbox. It can turn your inbox into an outbox through proactive messaging to patients,” Ayers said.
The bots can send patients personal messages tailored to their records and the doctors' anticipated needs. “What would that mean for patients?” Ayers said. “There is enormous potential here to transform the way patients interact with their healthcare providers.”
Advantages and drawbacks of chatbots
If chatbots could be used to generate messages to patients, they can even play a key role within the management of chronic diseases, which affect as much as 1,000 people 60% of all Americans.
Sim, who can also be a family doctor, explains it this manner: “Chronic illness is something you have 24/7. I see my sickest patients for an average of 20 minutes each month, so I'm not the one doing most of the chronic care.”
She encourages her patients to exercise, control their weight and take their medications as prescribed.
“But I don’t offer support at home,” Sim said. “AI chatbots, because of their ability to use natural language, can be there for patients in ways we doctors can’t.”
She said that along with counseling patients and their caregivers, conversation agents can analyze data from monitoring sensors and ask day by day questions on a patient's condition. Although none of this may occur within the near future, it's a “great opportunity,” she said.
Ayers agreed, but cautioned that randomized controlled trials must be conducted to find out whether an AI-powered messaging service can actually improve patient outcomes.
“If we don’t do a thorough scientific study of these conversational agents, I can imagine scenarios where they will be implemented and cause harm,” he said.
In general, Ayers said, the national AI strategy must be patient-centered and never focused on how chatbots help doctors or reduce administrative costs.
From a consumer perspective, Ayers said he is anxious about AI programs “making generic recommendations to patients that could be insignificant or even bad.”
Sim also emphasized that buyers mustn't depend on the answers that chatbots provide to health questions.
“There needs to be lots of caution. These things are so compelling in the way in which they use natural language. I believe it's an enormous risk. At the very least, the general public must be told, 'There's a chatbot here, and it could possibly be unsuitable.'”