PRIVACY:
Generative AI cannot be allowed in managing health records, as it would need to upload personal information to an external database, a health official said
By Wu Liang-yi
and Chiu Chih-jou / Staff Reporters
While the use of artificial intelligence (AI) has revolutionized the healthcare industry, generative AI is prohibited from producing patient health records to protect people’s personal information, the Ministry of Health and Welfare said.
According to the “Guidelines on the Qualification and Classification of Software Used in Healthcare” (醫用軟體分類分級參考指引), the use of processing software for collecting, saving, analyzing, displaying and transferring health conditions, physiological parameters and medical records should be within healthcare facilities, personal homes and telemedicine; while AI-assisted medicine is managed by medical device software.
The Food and Drug Administration (FDA) in May 2021 established a smart medical devices office, providing consultation to people in the industry. Most of the items that have undergone review by the office are AI-assisted medical device software for screening, detecting and diagnosis, listed as Class 2 (moderate risk) medical devices. Such devices are new and complicated, and there are more than 100 cases waiting to be reviewed.
Photo: Reuters
AI-assisted medical devices that have obtained FDA approval included ones analyzing images to diagnose cardiovascular diseases, tumors, eye disorders, bone fractures, and chest and abdominal abnormalities, and many hospitals have been collaborating with medical device manufacturing companies.
Companies such as Ever Fortune AI Co Ltd (長佳智能) have developed the NeuroSuite computed tomography intracerebral hemorrhage assessment system and obtained US FDA approval last week.
The Taipei Veterans General Hospital and Taiwan AI Labs (臺灣人工智慧實驗室) also collaborated in developing DeepMets, a brain metastasis detection module, which received approval in Vietnam in August and is on track to enter the Southeast Asian medical device market.
However, the use of generative AI, such as the chatbot ChatGPT, is prohibited from producing health records.
According to Article 12 of the Physicians Act (醫師法), physicians should produce patient health records, either through handwriting or typing, and they cannot use software to produce them, the ministry’s Department of Medical Affairs deputy director Liu Yu-ching (劉玉菁) said.
As generative AI would need to upload patient health records with personal information to another database other than the National Health Insurance’s, it cannot be allowed, she said.
Some hospitals say they use AI-assisted speech tools for producing health records, but they are actually speech-to-text software, not AI, Liu said.
“However, as technology advances rapidly, if there are doubts about the use of certain technology, it will be discussed and reviewed,” she said. “The current regulations are not unadaptable, but so far the ministry still has limits on AI application in the healthcare industry.”
Hope Foundation for Cancer Care chairman Wang Cheng-hsu (王正旭), a physician, said the efficacy of AI-assisted diagnosis technology needs to be carefully tested and verified, while laws need to be updated to keep up with new technologies.
Physicians must be responsible to their patients, so being transparent about how information is collected is important, Wang said, adding that explaining to them how the treatment procedures would be assisted by AI, and if it has obtained government approval, it would ensure that patients have confidence in the treatment.
While many people often complain about “three longs and two shorts” — waiting too long to book an appointment, seeing a doctor and retrieving the prescribed drugs, and having too short time to ask questions to the doctor and getting explanations — AI-assisted technology can reduce the time doctors need to spend on low-level administrative work, Taiwan Young Patient Association vice president Eric Liu (劉桓睿) said.
Additional reporting by Lin Hui-chin and Chen Yung-chi
Comments will be moderated. Keep comments relevant to the article. Remarks containing abusive and obscene language, personal attacks of any kind or promotion will be removed and the user banned. Final decision will be at the discretion of the Taipei Times.
