Practice and Exploration of Smart Outpatient Services at the Second Xiangya Hospital: Innovations in AI-Driven Medical Records, Post-Diagnosis Management, and Microservices Architecture
On August 17, 2019, the Smart Healthcare Special Session of the Sangfor Innovation Conference Sub-Forum (co-organized by Chinese Digital Medicine magazine) was held in Shenzhen. At the event, Zhu Hongtao, Director at the Second Xiangya Hospital of Central South University, delivered a keynote speech titled “Practice and Exploration of Smart Outpatient Clinic Construction.” VCBeat has compiled and edited the highlights from his presentation.
Zhu Hongtao: Director of the Information Network Center, Xiangya Second Hospital of Central South University
Director Zhu Hongtao shared Xiangya Second Hospital’s experience in building smart medical outpatient services, covering four aspects: reflections on the development of information technology, practical implementation of smart outpatient clinics, application of emerging technologies in smart outpatient construction, and challenges and explorations in this field.
A review of the entire process of hospital informatization reveals that each update, iteration, and evolution of information technology has significantly promoted improvements in hospital operational efficiency and service delivery models.
In 2018, the national government issued the “Notice on Printing and Distributing Three Documents, Including the Administrative Measures for Internet Diagnosis and Treatment (Trial).” In March 2019, it released the “Smart Hospital Service Grading Assessment Standard System (Trial).” As emerging information technologies are leveraged to continuously improve service quality and efficiency and enhance patients’ sense of gain in medical care, the pressure on information systems has been gradually increasing.
Currently, the smart healthcare outpatient applications that are widely recognized primarily cover non-core medical processes such as appointment registration, payment, information inquiry, and light consultations. Medical services targeting core clinical workflows remain scarce. The transition from general services to in-depth medical care, ultimately achieving comprehensive intelligent healthcare services, requires further exploration.
I believe we can conduct explorations and trials addressing the “Three Longs and One Short” (long wait times for registration and payment, long waiting times for consultations, long lead times for scheduling examinations, and short consultation durations). Current medical service initiatives primarily target the “Three Longs,” and I believe it is feasible to eliminate them. However, eliminating the “One Short” is not realistic, as outpatient volumes at some large hospitals continue to grow. Instead, we can optimize the efficiency of consultations without extending their duration. Furthermore, we can enhance the continuity of care by integrating services end-to-end, aligning with national policies and exploring innovations in areas such as tiered diagnosis and treatment.
Since 2016, the Second Xiangya Hospital has launched explorations and practices of “Internet + Smart Outpatient Services” in two National Clinical Research Centers: Mental Health and Metabolic Endocrinology.
Technological Innovation: To address the challenges of high outpatient volumes and the difficulties in implementing and applying electronic medical records (EMR) in outpatient settings, we leveraged “Internet + Artificial Intelligence” technologies to develop a proactive outpatient EMR system, which significantly improved the efficiency and quality of EMR documentation by outpatient physicians.
Service Innovation: Given the large number of out-of-town patients and their strong demand for post-consultation management and services, an intelligent post-consultation management system was developed based on the proactive outpatient service platform to achieve full-course patient management.
Mode Innovation: In alignment with the gradual rollout of disease-specific outpatient clinics at our hospital, we have developed an integrated specialty service platform for outpatient care, scientific research, and follow-up, based on the proactive outpatient electronic medical record system.
The concept of proactive medical records leverages internet and artificial intelligence technologies. After patients register for appointments via platforms such as WeChat or Alipay, the system proactively initiates an intelligent consultation based on the registration information. It collects patients’ basic medical history through a question-and-answer format and employs natural language processing (NLP) and other techniques to generate medical records that are readable and comprehensible to physicians. This approach significantly reduces physicians’ consultation time while ensuring the completeness of outpatient medical records.
Meanwhile, through an integrated workstation design leveraging application components and container technologies, configurable specialty-specific applications are enabled. Furthermore, by restructuring the outpatient physician workstation, the time physicians spend documenting medical records and issuing orders is significantly reduced, thereby increasing effective doctor-patient communication time and enhancing patients’ sense of healthcare satisfaction.
In terms of patient services, we have developed a virtual robotic assistant to shift the paradigm from "patients seeking information" to "information seeking patients." This approach provides more humanized and precise optimized decision support throughout the entire medical consultation process, including recommended appointment times, intelligent in-hospital navigation, and smart scheduling. These features minimize unnecessary patient movement, ensure that each visit is more effective, and achieve a "dual reduction" in both the overall cost of care and time costs.
In post-consultation management, we assess each individual’s self-management capabilities and level of knowledge mastery. Based on the assessment, physicians prescribe structured educational interventions tailored to the patient, which are linked to the monitoring and intervention of key post-consultation indicators. This enables personalized patient education and management, while precise daily monitoring and rehabilitation plans are developed and pushed to the patient’s mobile device.
Post-consultation management prescriptions are generated through a human-AI collaborative approach, providing patients with automated services such as medication reminders, indicator monitoring, daily data collection, health education, and tiered interventions. Patients can view their personalized rehabilitation goals set by physicians via WeChat from home and provide timely feedback on their daily conditions. This feedback is transmitted back to the hospital’s electronic medical record system, creating a comprehensive full-process patient profile. Physicians can then conduct post-consultation interventions and management through mobile devices or physician workstations. This post-consultation management model has been successfully implemented in the Department of Endocrinology at our hospital.
We are also exploring the use of artificial intelligence technologies to support functional decision-making and clinical diagnosis and treatment. Preliminary applications include assisted diagnosis, medication support, cluster analysis, and disease onset prediction; however, there is still a gap before these solutions can be fully implemented in practice.
Referring to the “Graded Evaluation Standard System for Smart Hospital Services” released in March this year, we have further established an informatics capability model for smart hospitals. By leveraging technologies such as microservices and containers, we have implemented the system architecture for smart outpatient services, enabling the rapid implementation and deployment of intelligent healthcare services.
In terms of top-level technical architecture deployment, we have established a microservices architecture. This architecture enables the construction of stable microservice applications within the system, allowing services to invoke one another through unified service contracts. It facilitates rapid response to changes in clinical requirements and flexibly addresses the differentiated needs across various medical specialties.
Microservices require an underlying infrastructure to operate, referred to as the “microservice foundation.” This foundation comprises multiple frameworks that provide unified audit logging, service registration, centralized scheduling, and a unified access gateway. All microservices operate on top of this “microservice foundation,” much like apps running on an operating system. Each subordinate system within the microservice architecture can function as an independent microservice.
How Should Microservices Be Deployed? Since their emergence, there have been three primary approaches to microservice deployment.
One approach is the multi-service model on a single host, which is rarely used nowadays. The second approach involves deploying one service instance per container. Its advantage lies in the complete isolation of each service instance, with fixed memory allocation that prevents resource theft from other services. However, this approach results in relatively low resource utilization in microservice deployments.
The currently prevalent approach is one instance per container. Unlike virtual machines, containers are a lightweight technology that can be built and started up very rapidly.
We are also exploring the construction of this model on the scientific research big data platform. Currently, some vendors are using microservices architecture to build their platforms and Hospital Information Systems (HIS). However, migrating all vendors to this architecture remains a lengthy process.
Currently, our hospital’s smart outpatient services still face certain challenges. Due to significant variations in physicians’ emphasis on outpatient medical records, participation from some specialties in constructing intelligent consultation knowledge graphs has been low, resulting in incomplete collection of patient medical histories. Furthermore, the lack of clear pricing for whole-course disease management has dampened physicians’ willingness to contribute to post-consultation management knowledge bases and provide value-added services. Consequently, the overall adoption level of these smart solutions within the hospital remains insufficient.
Another critical issue is data security. The leakage of registration information could trigger an earthquake within the healthcare industry. Security breaches during the internet transmission of diagnostic and treatment data would lead to severe consequences. Therefore, ensuring the security of medical records is paramount. Additionally, with the adoption of these new technologies, safeguarding code security has also become a pressing challenge that must be addressed.
In terms of exploration, our next step is to leverage artificial intelligence technology to comprehensively optimize the medical consultation process; utilize 5G and smart hardware products to achieve a closed loop of medical and health data both inside and outside the hospital; based on big data, provide patients with personalized, precise, and intelligent medical services; and offer physicians efficient and convenient clinical decision support and patient service tools.
I believe that comprehensively enhancing physicians' service efficiency and quality will become the primary direction for smart healthcare as a whole.