【明報專訊】AMID the growing popularity of generative artificial intelligence (GenAI) applications, the Hospital Authority (HA) will begin a trial using GenAI to aid doctors in drafting medical reports at six hospitals next month. Adopting AI to enhance work efficiency is already a major trend in public service and administrative governance. The Shenzhen authorities have recently deployed their first cohort of 70 "AI-powered digital employees" to assist in document processing and public services. Looking back at Hong Kong, the SAR government began the trial of a document editing AI co-pilot application for the civil service last year, but only a very limited number of users have participated in the trial, and it is uncertain when its use will be extended to all governmental departments. The SAR authorities have to gather pace and speed up their digital transformation.
The HA announced yesterday (20 February) that a GenAI system will be adopted at six hospitals, including Princess Margaret Hospital, United Christian Hospital and Queen Elizabeth Hospital, to assist in drafting medical reports. Developed internally by the HA, the system can summarise patients' clinical data and provide quick drafts of medical records or reports, for review and revision by doctors before finalising the content. If implemented smoothly, the system is expected to operate at all public hospitals within this year.
The workload of public healthcare professionals has been immense. Aside from daily outpatient consultations and ward rounds to communicate with patients and their families, doctors also have to attend different meetings, fill out various sorts of forms and process documents like patient discharge slips, among others. The overloading of paperwork has made it difficult to spend more time seeing patients. In 2023, the HA received 110,000 counts of medical report requests related to patients' insurance claims and others. Medical report writing, on average, consumed 8% to 10% of doctors' working time. Among all specialties, orthopaedic surgeons spent particularly more time on report writing.
Not only can the introduction of AI tools to quickly draft preliminary medical reports significantly ease the related workload on doctors, but it can also reduce the processing time of documents and provide convenience to patients and related people. Of course, AI tools are not error-free. The HA has also clearly delineated that AI only serves an auxiliary role; doctors confirm the reports in the final step and are responsible for the content.
As GenAI develops in leaps and bounds, many mainland enterprises have joined the rush to adopt DeepSeek. Even governmental bodies are no exceptions. In Shenzhen, the Futian district government has taken the lead in launching AI tools based on DeepSeek. Its first batch of 70 "AI-powered digital employees" have just "assumed duties" recently, deployed in areas including document processing, public services and emergency management. Despite their auxiliary nature, they undoubtedly offer great help.
The SAR government announced early last year that it would roll out over 100 digital government and smart city initiatives one by one, including enhancing security in public cargo working areas with AI, and extending the AI chatbot services in the 1823 enquiry service. A domestically developed AI document processing co-pilot tool has also been made available to staff of various bureaux and departments for trial.
However, as of November last year, only about 800 government officers of different grades from different departments had participated in the pilot programme, mainly for document processing tasks like drafting, translating and summarising. So far there is no specific timetable for using the system at a larger scale. It is also a concern whether the implementation of the over 100 digital government initiatives is behind schedule. The authorities should push ahead with the transformation into a smart government at full speed, instead of dragging their feet anymore.
明報社評 2025.02.21：生成式AI提升效率 特區政府須加快推進
生成式人工智能(AI)應用日益普及，醫管局下月起將於6間醫院試行以生成式AI協助醫生撰寫醫療報告。引入AI提升工作效率，已成為公共服務與政府治理的大趨勢，深圳當局近日就引入首批70個「AI數智員工」，協助政府公文處理及民生服務。回看香港，特區政府去年開始試用公務員文書AI輔助系統，惟參與試用的人員仍很有限，未知何時才能推廣至所有部門，特區當局有必要加快腳步，加快智能化轉型。
醫管局昨天表示，在瑪嘉烈醫院、聯合醫院、伊利沙伯醫院等6間醫院，引入生成式AI智能系統，協助撰寫醫療報告。系統由局方內部開發，可以綜合病人臨牀資料，快速起草醫療紀錄或報告，醫生會再檢視及修改，最終確定報告內容。如運作暢順，預計今年內在所有醫院推行。
本港公院醫護工作繁重，醫生除了平日門診看病、巡房跟病人及家屬溝通，又要出席各式會議，以及填寫各類表格、處理病人出院文書等工作。文書工作太多，無法抽出更多時間看病人。2023年醫管局收到逾11萬份跟病人保險索償等有關的醫療報告申請，撰寫醫療報告平均佔去醫生8%至10%的工作時間，當中又以骨科醫生需要撰寫較多報告。
引入AI工具快速起草出初步醫療報告，不僅可以顯著減少醫生相關工作，也可以縮短處理報告時間，方便病人及相關人士。當然，AI仍有機會出錯，醫管局也有相當清楚的界定，AI僅屬輔助，醫生作為報告最終確認者，需要對報告內容負責。
生成式AI飛躍發展，不少內地企業都紛紛接上DeepSeek，就連政府機關也不例外。深圳福田區就率先推出基於DeepSeek的AI輔助工具，首批70個「AI數智員工」近日正式「上崗」，應用範疇包括公文處理、民生服務、應急管理等，雖然僅屬輔助，作用卻毋庸置疑。
特區政府去年初宣布，將陸續推出數字政府及智慧城市「百項方案」，包括利用AI提升公眾貨物裝卸區保安，以及在1823查詢服務中擴展AI聊天機械人服務等。本地研發的AI文書輔助工具，亦開始供各局及部門員工試用。
但截至去年11月，只有800多名不同部門職系人員參與試用，主要進行文件撰寫、翻譯、摘要等文書工作。到底何時可以較大規模應用，暫時仍未見有具體時間表；「百項方案」推展進度有否落後，同樣惹人關注。當局應全速推動政府的智能化轉型，不能再歎慢板。
■ Glossary 生字 /
delineate : to describe, draw or explain sth in detail
auxiliary : giving help or support to the main group of workers
drag your feet : to be deliberately slow in doing sth or in making a decision