Current Status and Key Issues of Data Management in Tertiary Hospitals: A Case Study of Seoul National University Hospital

1. Data Types and Amount

HIS data are categorized into Electronic Medical Records (EMRs), picture archiving and communication systems (PACS), digital pathology (DP), clinical data warehouse (CDW), and research systems using a common data model (CDM). Table 1 shows the total amount of data in the HIS for each hospital.

Table 1

Total amount of data in hospital information systems (unit: GB)

2. Data Management Infrastructure

To store and manage data, hospitals choose on-premise or cloud-based approaches, or a combination of both to run their information systems. On-premise methods involve the hospital maintaining its own server, physically located onsite and managed by the IT department. Alternatively, hospitals can opt for a cloud system, in which an external cloud service provider supplies the necessary infrastructure support and security for managing data. A cloud system can be accessed through a subscription-based payment system, and resources and infrastructure can be easily expanded. The main benefit of a cloud system is reducing the initial budgetary burden for purchasing systems.

1) On-premise system

In 2004, SNUH implemented its first HIS, and in 2016, the HIS was upgraded to BestCare 2.0, a next-generation HIS. Following the plan to relocate the computer center from inside the hospital to an external location, the main computers are currently situated in an external Internet data center (IDC), with the disaster recovery server (a simplified version of the HIS) installed in a separate space within SNUH (Figure 1). Both SNUBH and SNU-BMC have also installed and are operating HISs on their premises.

Figure 1

Schematic diagram of the computer center of Seoul National University Hospital. HIS, hospital information system; PACS, picture archiving and communication system; IDC, internet data center; SDDC, software-defined data center.

2) Cloud system

As a research system, SNUH employs Amazon Web Services to establish a CDM analysis environment and a big data analysis platform called SUPERB in a cloud environment. At SNUBH, a CDM is utilized both on-premises within the hospital and in a cloud system hosted by Amazon Web Services (Figure 2).

Figure 2

Big data analysis platform at Seoul National University Bundang Hospital. AWS: Amazon Web Services.

3. Data Backup Plans

SNUH conducts daily backups of the HIS database and server, managing all data in accordance with a policy that requires retention for three months. Additionally, SNUH performs a vault backup on hard disk drives once a month, retaining them for up to 2 years. Similarly, SNUBH categorizes the HIS data into three distinct types: EMR, PACS, and optical files. Primary backups are stored within the hospital, secondary backups are physically located in a separate location, and third backups are stored on hard disk drives and tapes.

4. Organizations for Data Management

1. At SNUH, the data management organizational structure comprises the Information Planning Team, Information Development Team, Medical Records Team, and Information Security Center, all of which fall under the Office of Hospital Information. In addition to these teams, SNUH has the Department of Data Science Research, Department of Medical Device Innovation Research, and Department of Medical Device Development Support under the Biomedical Research Institute (Table 2). Other data management committees include the Healthcare Information Management Committee, which is responsible for evaluating the feasibility of projects related to information system establishment, and the Medical Records Management Committee, which oversees the quality of medical records and standardization of terminologies. The Institutional Review Board (IRB) reviews research projects that utilize medical data and the Big Data Review Board (BRB) is responsible for assessing the feasibility of utilizing big data.

2. At SNUBH, the data management organizational structure includes the Office of Hospital Information, Digital Healthcare Research Department, Healthcare ICT Research Center, Big Data Center, and Medical AI Center (Table 3). When data is used for educational and research purposes, both the IRB and the Data Review Board work together to ensure that data utilization is safe and responsible [5].

3. At SNU-BMC, the data management organizational structure comprises the Healthcare Information Team, Medical Records Team, and EMR Operation Team under the Office of Hospital Information, as well as the Data Science Center under the Biomedical Research Institute (Table 4). SNU-BMC also operates both the IRB and the Data Review Board together for safe research.

Table 2

Data management organizations of Seoul National University Hospital (SNUH)

Table 3

Data management organizations of Seoul National University Bundang Hospital

Table 4

Data management organizations of Seoul Metropolitan Government-Seoul National University Boramae Medical Center

1. Challenges of Data Management

1) Rapid data growth in data storage

In general, PACS capacity has shown an annual average increase of 11.25% at SNUH (Figure 3). Projections indicate that total storage will increase by approximately 1 petabyte (PB) per year from 2021 to 2025, about 0.5 PB per year from 2026 to 2028, and roughly 0.4 PB per year from 2029 onwards (Figure 4). The HIS, as well as the PACS, must be able to handle continuous growth in users, patients, and data while maintaining performance and efficiency. However, as the demand and cost for data storage space continue to rise, maintaining all data in the operating system may become unsustainable, considering the significant system purchase costs and management manpower expenses required. To achieve cost and manpower efficiency, it is recommended to separate clinical operational data and research data in the future to keep the speed-critical clinical database lightweight.

Figure 3

Picture archiving and communication systems (PACS) volume growth at Seoul National University Hospital.

Figure 4

Storage growth over time.

2. Government Regulations

Regarding government support and regulation, the following issues need to be considered.

Regulations on the “Retention of medical records data” are detailed in Article 15 of the Enforcement Rule of the Medical Service Act (Preservation of Medical Records) [6] as shown in Figure 5.

Figure 5

Details of Article 15 of the Enforcement Rule of the Medical Service Act (preservation of medical records).

The Medical Service Act and its Enforcement Rule systematically outline the retention period for medical records. However, the Medical Service Act does not provide separate regulations regarding the management or destruction of medical records beyond what is stipulated in the Personal Information Protection Act. This lack of clear guidance is further complicated by the ambiguous explanation of the “proviso clause” in Article 15, Paragraph 1 of the Enforcement Rule of the Medical Service Act. As a result, effectively managing data and systems in the field becomes difficult. Medical institutions, including SNUH and SNU-BMC, at which HISs have been in operation for more than 10 years, must decide whether to retain or destroy medical records in compliance with applicable regulations. However, detailed destruction guidelines and procedures are absent at this moment. Furthermore, the regulation mandating data “destruction” after the retention period has expired is based on personal information protection and does not account for other secondary uses such as long-term treatment or research. Consequently, there is a need for institutional guidelines addressing the retention, destruction, and management of various types of data in medical institutions.

In order to vitalize the digital healthcare industry, some areas need improvement in terms of health and medical data integration and utilization. A medical data-driven hospital project [2] is being carried out to exchange standardized medical data, but only data exchange between institutions within the consortium is currently possible. Through the MyHealthWay project [4], it is necessary to establish an ecosystem that integrates individual-centric medical data, allowing individuals to transmit and utilize their own data. With the amendment of the Data 3 Act, data for research can be provided after pseudonymization, but conflicts with the existing Medical Service Act [7], which prohibits the disclosure of medical records to third parties, must be resolved. Additionally, the right to request transfer of medical data to third parties should be strengthened through legislation. To regulate AI-based medical devices in Korea, the Ministry of Food and Drug Safety distributes and continuously revises guidelines for the approval and clinical efficacy evaluation of medical devices that incorporate big data and AI technology [8–10]. Furthermore, as generative AI has recently attracted attention, guidelines for trustworthy and explainable AI are also needed.

3. Need for Comprehensive Metadata Management

Effective metadata management is essential for comprehending the meaning and relationships within complex healthcare big data and enhancing data transparency. By minimizing confusion and errors resulting from unclear data, metadata management enables quick understanding and navigation of the data. This is crucial for data quality control and improvement initiatives, facilitating accurate decision-making and swift access to new insights. Therefore, to improve the knowledge and utilization of medical big data, a systematic and ongoing protocol for metadata management should be established throughout the entire hospital organization and operated accordingly. SNU hospitals use the same terminology sets for diagnosis, chief complaints, and operations through the standard protocol for adopting new terms. In the future, collaborative work across hospitals will be necessary to expand terminology standardization for other areas such as nursing and clinical lab tests. Effective metadata management, backed by standard terminology, can substantially enhance data consistency, interoperability, and quality within SNU hospitals. A comparative analysis would help highlight current strengths, identify areas for improvement, and guide a roadmap for future enhancements, including potential inter-institutional cooperation to harmonize data practices across different hospitals.

1. Promoting Data Sharing among Hospitals

Fast Healthcare Interoperability Resource (FHIR) is a standard framework for healthcare information exchange created by HL7 (Health Level 7). In Korea, the adoption of the FHIR standard is being promoted through initiatives led by the Korea Health Information Service and HL7 Korea, aiming to support online education and technology [11]. The SNUH, SNUBH, Gangnam Center, and SNU-BMC developed a standard terminology management system in 2013 and integrated it into their HIS. SNUBH plans to expand and incorporate the HL7 FHIR standard, providing a foundation for patient-centered data sharing, enhancing Health4U (a personal health record portal/app for patients) and building an infrastructure capable of delivering patient-generated health data to EMRs. At SNU-BMC, adopting the HL7 FHIR next-generation standard framework is vital, and there is also a need to consider offering mobile patient services for access via mobile devices and web portals through the standard FHIR API. If government-level guidelines are introduced in this area, an efficient system can be established.

2. Infrastructure Configuration for AI Applications

To clinically evaluate AI software developed as a research project or to integrate a solution in the process of commercialization, it is crucial to install and assess the application without disrupting the existing EMR system. Creating an environment that accommodates AI technology necessitates a thorough review of patient safety issues and ethical considerations. A digital twin of real-world data, including people, hospitals, systems, and diseases, can enable safe and effective simulations in virtual reality [12]. These simulations can enable the development of various applications such as training, education, remote medical consultations, and medical AI services. However, there is currently no dedicated department in the hospital, and implementing such an environment depends on the determination of hospital administrators to address these conditions.