ChatGPT could free doctors to spend more time with patients
Artificial intelligence could be used to free up hospital doctors and surgeons from repetitive paperwork so they can spend more time with patients, according to a new trial.
Doctors have used the online AI programme ChatGPT to write detailed surgical discharge letters that give information to patients and their GP about their care after procedures.
The AI-written letters were deemed to be almost indistinguishable, and in some cases were classed as better in terms of the medical information they contained, than real hospital letters written by humans.
It raises the potential for discharge letters, which NHS doctors tend to write out individually for each patient in many hospitals, to be automatically generated using AI.
The study is being presented today (12 July 2023) at the first ever Health and Care Analytics Conference, which is taking place in Birmingham on 11-12 July.
Dr Chris Bodimeade, a Core Surgical Trainee in South East England, who led the work, said: “I am half way through my core surgical training and one of the frustrations I have with the NHS is the amount of time I have to spend on a computer.
“Doctors are increasingly becoming tethered to a computer doing mundane and repetitive paperwork tasks rather than spending time with patients. But at the same time I’ve been excited by the development of AI technologies like ChatGPT, so I wondered if there was a way of using it to make some of these tasks easier.”
Surgical discharge letters for routine operations usually contain a summary of the treatment a patient has received during their stay in hospital, alongside any potential complications and how their care should be managed after they leave hospital.
Dr Bodimeade decided to ask ChatGPT to write a surgical discharge letter for two hypothetical patients who had undergone an elective tonsillectomy. To generate the letters, he gave the AI chatbot the simple instruction: “Please write a hospital discharge letter for patient undergone a tonsillectomy.” No patient information was included.
He then showed the resulting letters to 19 other consultants and doctors at his current hospital alongside two real letters written for patients that had been anonymised. The doctors and surgeons found the letters indistinguishable and one of the letters was on average rated to contain better medical information and was easier to read than the real discharge letters.
Dr Bodimeade said that while his study was small, he hopes it might lead to greater exploration of how artificial intelligence might lighten the workload of NHS staff.
He stressed that any letters would all still need to be checked by a doctor before being sent and said colleagues in other surgical departments were already conducting their own experiments to use ChatGPT to write discharge letters.
He said: “I’m hoping this work can open up a discussion about how technology that is freely available can be used within the confines of confidentiality and patient safety to make our jobs much more efficient and improve a service that is under tremendous stress.”