Case Study Intel Xeon Processor E5-2600 Family Regional Medical Bigdata/Cloud Services Regional Medical Big Data Whitepaper UFIDA Medical Big Data Application Case for a Healthy City Strategy in China A Trend Toward Big Data Analysis in China s Growing Regional Health Information Networks China s fast-growing regional health information exchange networks (RHINs) are beginning to show exciting opportunities for data analytics to help regional health authorities meet the Chinese Government s ambitious health improvement goals. Health information technology (IT) is one of eight pillars of the CPC Central Committee and State Council on Deepening Medical and Health System Reform and the Key Implementation Plan of the State Council for Deepening Medical and Health System Reform in Near Future (2009-2011). It is explicitly stated in the Plan that many efforts will be made to establish a practical and shared medical and health information system. To this end, RHINs are a foundation for best practices in the electronic sharing of patient summary information. Developing a smart healthcare system in China is part of a global trend to address the unsustainable growth in healthcare costs as the population ages. The Chinese governments and hospitals have made huge investments in health IT in recent years, rapidly improving the availability of patient information and services. In general, however, China s health IT development is still in an its initial state, with significant challenges in the reliability of data sharing mechanisms among hospitals, unbalanced regional development, a lack of uniform standards, etc. For this reason, it is important to demonstrate early progress not only in the sharing of patient summary information but also the establishment of big data platforms to develop knowledge from the growing data sets. This paper examines key big data implementation challenges and solutions through the experience of the software vendor UFIDA s use of the Intel Distribution of Hadoop to serve the regional health authority Healthy City plans in Jinzhou, with a population of 3.1 million in West Liaoning Province. The overall goals of the Chinese Government s medical and health system reform are proposed in the Key Implementation Plan are, to establish and improve a basic medical and health system covering urban and rural residents; to provide safe, effective, convenient and affordable medical and health services; to make basic medical and health services available to all people, and to further improve the health of all people. Besides health information technology, the other seven pillars of the Government s health reform plans are: Administration mechanism Operation mechanism Investment mechanism Pricing mechanism Surveillance mechanism Science-tech and HR guarantee Legal system Overall construction goals are also set forth in the Healthy China 2020 research report, namely to improve health condition of urban and rural residents, improve health and life quality of nationals, reduce disparities in health among different regions and to attain major health indicators of moderately developed countries. Under this plan: By 2015, the basic medical and health system will take initial shape and all nationals will enjoy basic medical security and basic public health services. The availability of medical and health services will be significantly improved and differences in health conditions and resource allocation among various regions will be
demonstrably narrowed. China will lead the developing countries by national health level indicators. By 2020, a basic medical and health system covering urban and rural residents will be established. All people will enjoy basic medical and health services, and the level of medical security will be steadily improved. The utilization of health services will be significantly improved and regional health disparities will be further narrowed. China will advance to national health levels of moderately developed countries. The Ministry of Health compiled the 12th Five-year Plan for National Health Informatization (3521-2 Program for short) in 2010. Specifically, 3 stands for national-, provincial- and municipal-level health information platform. 5 stands for five business sectors of health informatization, namely medical services, public health, medical security, drug security and integrated management. 2 stands for two basic resource libraries electronic health records (EHR), which in China means networked information for sharing and and electronic medical record (EMR), which in China means clinical data locally stored in hospitals. 1 stands for dedicated health information network. And latter 2 stands for information security system and standard specification system. Municipal-level health information platform, namely regional health information network (RHIN), is one of core elements of health IT strategy. Based on the 35212 Program blueprint, pilots of grassroot medical and health information system have begun under the support of the state in five provinces (regions and municipalities), namely Shanghai, Zhejiang, Anhui, Chongqing and Xinjiang. Various local governments have also been actively devoted to construction of regional medical information platform based on resident electronic health record (EHR) and accumulated a large number of EHR data. Big Data in China s Medical Industry As shown in the China Medical Industry IT Solution Forecast and Analysis 2012-2016 released by IDC, in 2011, IT expenditures of China s medical industry stood at RMB14.63 billion, an increase of 28.9% over the previous year. IDC predicates that IT expenditures of China s medical industry will reach RMB 33.99 billion in 2016, with 2011-2016 CAGR of 18.4%. All of this growth will inevitably generate massive amounts of data derived from medical business activities, health examinations, nine public health services and other medical and health services, as well as huge datasets from EMRs at hospitals, resident health records acquired by regional health information platforms, etc. There are a large number of unstructured/semistructured data, including images, Office file, XML structure files, etc. With the construction of 3-5-2-1 regional medical system actively advocated by China, hundreds of medical data centers are expected to appear in China, and each one will carry medical and health data of nearly a 10 million population as well as provide various services. According to estimates, a medium-sized city in China (10 million population) will accumulate 10 petabytes of medical data in 50 years (source: CCIDNET). In addition, with the advancement of personal health management, more and more personal daily health monitoring information will be generated, resulting in a data size and growth rate beyond imagination. Despite being stored electronically, the overwhelming majority of medical data are stored in archives unavailable for real-time query. It is very complicated to rapidly search data dispersed across different business systems. In the past, these unmet needs for data integration had no proper technical solutions. Simple application of traditional storage technology and analysis methods has two problems: 1) diversified and changeable data formats, and 2) slow access to acquiring mass data. Application of Big Data in the Medical Industry Big data solutions will have numerous application scenarios in the medical industry. In China s medical industry, the most practical application of big data technology lies in management and services of resident health record data meaning the management of personal full-life-cycle medical/health data. From the perspective of doctors and health administrators, full-life-cycle health record retrieval is of practical significance. For example, for patients with chronic diseases, previous pathogenesis changes and treatment processes play a helpful 2
role in assisting doctor s diagnosis and treatment. Data of allergic history and adverse reactions also play a positive role in avoiding malpractice and medical negligence. Statistics of and analysis on mass medical and health data provide a more scientific basis for management decision making and supervision implementation. Traditional clinical research are based on sample survey, while increasing health record data will help greatly reduce workload and improve the quality and quantity of scientific research data as well as data processing efficiency. Big data offers promising opportunities in the future when applied to clinical diagnosis, scientific research, public health decision making by governments, and individual health management. Challenges of Big Data Application Big data also faces various challenges in the medical industry. According to our observations, these challenges go far beyond technical and include the following areas: Awareness: We must fully realize the necessity, complexity and imperativeness of data services. Maybe some corporate users still design information system guided by previous thinking and methods because they lack such awareness. E-commerce and mobile Internet developed suddenly and rapidly. As a result, we will inevitably be unable to cope with booming medical data and growing data service needs if we don t apply forward-looking thinking. Talent: There are now scarce talents in big data. Only a few companies and talents master big data technology. Powerful companies and enterprises are expected to actively promote the application of big data technology in the medical industry and drive big data development and talent cultivation. Exploration of business application model: Many enterprises are not familiar with the medical business and thus are unable to discover key business intelligence with big data application in medical practices, constraining big data development in the sector. Overall Framework of Regional Medical Big Data Figure 1 RHIN/Grass Root solution with Big data (Hadoop) 3
Purposes Most developers adopt traditional database to process health record storage and services in the implementation of the Regional Health Information Network (RHIN) Project. Some service providers also use No SQL systems such as MongoDB to manage unstructured health records. However, the integrated solution cannot be formed using a single component or product. An Intel big data approach to health record management handles the relationship between data services and business production systems by organically separating them, laying a solid foundation for performance, extensibility and dynamic expansion. Positioning Big data solution is mainly positioned at health record management and service in the regional medical platform, which will be implemented along with the implementation of regional medical platform. As shown in Figure 1, the big data solution is positioned as a data service system structured on HIAL, which effectively stores and manages medical data acquired by HIAL and provides integrated data services for doctors, managers and patients. The big data solution is integrated with grassroots medical institution information systems based on virtualized cloud technology to acquire relevant grassroots medical data on one hand and provide holistic data services for grassroot medical institutions on the other hand. The services are integrated into business system interfaces of doctors, managers and patients to enable users to smoothly get data services in routine operations. Functions The functions of a big data solution in the regional medical service include basic services, data analysis and compliance management. Basic services include storage, inquiry and scanning. Themed-based data analysis means to analyze clinical data, public health management data, performance assessment data, payment data of rural cooperative medical services so as to find possible trends and risks in mass data of different structures. Compliance management represents the advanced stage of big data service. As an example, it leverages data across business activities recorded in business systems to trigger off data service requests, or retrieves and analyzes big data sets to generate intelligence to improve business operations. Such big data service is possible in a real-time, accurate and available way only under the support of big data technology. Value of defining a Solution Stack is to provide reference for data management and service system construction for regional medical platforms being built in various regions. The value of the Solution Stack is to lay a solid foundation for future regional medical health record services, greatly saving time and money and reducing risk by starting with well-defined requirements for extensibility. Intel Big Data Solution Intel provides hardware platforms with powerful computing capability as well as Hadoop-based big data software platforms to form complete big data solutions serving corporate customers. Intel supports big data with the Xeon E5 Processor for distributed processing and high-performance computing, while the Xeon E7 can serve relational databases as well as business intelligence technologies and applications. Efficiency, openness, flexibility, universality, and ease of optimization are characteristics that make Intel Xeon Processorbased double- and multi-route platforms ideal for datacenters and supercomputing. The Intel Xeon Processor has unparalleled advantages in high scalability needed by big data processing for business intelligence. Targeting the needs of big data distribution and management, Intel provides Hadoop products and services for Intel platform optimization. Hadoop One of the most important big data platforms from opensource Apache project. Use of a simple programming model to support distributed processing of big data sets on the computer cluster. Complete technology stack includes a general utility program, a distributed file system, analysis and data storage platforms and an application layer managing distributed processing, parallel computing, working process and configuration management. In addition to high availability, the Hadoop framework can also process mass, complicated or unstructured data sets in a more cost-efficient way than traditional methods, and provides excellent extendibility and speed. Targeting the diversified and massive amounts ofmedical business data, big data technology can create unique value. Thepurpose 4
Figure 2 Intel Distribution of Hadoop Intel Distribution of Hadoop includes a complete distribution of Apache Hadoop Open Source Project and value-added monitoring and management control console--intel Manager for Hadoop* software. Intel updates and fully supports the Intel Distribution of Hadoop. Its advantages include: Intel Distribution of Hadoop is tested, verified, deployed and operated in the customer s production environment to ensure 7x24 continuous operation. Intel Distribution of Hadoop has gained from experience through the development of solutions to problems in actual application by customers, and includes a large number of improvements and performance optimizations by Intel. These improvements mitigate defects and shortcomings of open source Hadoop. Intel tools for cluster management and settings simplify Hadoop installation and configuration. Optimal cluster configuration can be automatically generated according to the user s hardware environment to give full play to the cluster s computing capability. Based on Intel s accumulated experience in cloud computing R&D, Intel Distribution of Hadoop provides professional consulting services in various stages, ranging from project planning to implementation, to help customers build highscalability and high-performance distributed systems. With Intel hardware platforms, Intel Distribution of Hadoop provides comprehensive hardware and software solution design. Big Data Application Case of UFIDA Regional Medical Platform With a registered population of about 3.1 million, Jinzhou is a regional central city and important port in West Liaoning Province. In March 2012, the Jinzhou Municipal Government as part of its healthy city strategy decided to build city-based regional health data centers and cover the main business sectors of the RHIN by leveraging a resident health card. Data to be stored include relevant portions of resident health records, EMR, public health, integrated management, etc. The datacenters are expected to reach the petabyte level in about 20 years. As the healthy city concept advances, there also will be increases in individual health data along with increasing personal health services. Challenges As main constructor of regional health data center of the healthy city strategy in Jinzhou, UFIDA Medical and Health Information System Co., Ltd. ( UFIDA Medical ) has rich successful experience in construction of provincial-, municipal- and prefecture-level smart healthy city regional health solution projects. This experience has given UFIDA Medical a window into potential problems if a traditional relational database construct were to be used to build the regional health data center in Jinzhou: Mass data The city-based regional health datacenter will store EHR, EMR and health management databases in a medium-sized city with 3 million population, meaning its health data center will reach the petabyte level in 20 years. 5
Traditional relational data has limitations for realization of big data storage; a performance problem exists when more than 500GB data are stored in a sheet of table. Complicated and changeable data types The regional health data center in Jinzhou will store a large number of unstructured data and semi-structured data. Traditional relational database will bring about numerous tough problems: The sizes of unstructured data generated by PACS images, B ultrasound, pathologic analysis and other payloads vary greatly from hundreds of KB to hundreds of MB. It may be necessary to store and retrieve hundreds of image files from simply one diagnostic episode for a single patient. Clinical EMR data are generally in standard XML file format of HL7 CDA. The file formats will keep evolving over time. The complexity of medical and health business operations makes it difficult to develop unified data standards, bringing a new challenge to data access and exchange. Additionally, future data processing will face great challenges. For example, storage, backup and capacity expansion are likely for mass data in the future; retrieval requirements for specific data as well as high-efficiency data exchange will inevitably generate new needs for the regional health datacenter in Jinzhou. Hadoop-based Mass Data Processing and Analysis To help avoid the above-mentioned problems, Intel is assisting UFIDA Medical by providing an architectural analysis and guidance based on in-depth explorations and experience in solving big data analysis problems. Intel and UFIDA Medical jointly developed goals of the regional health datacenter construction based on Intel s big data solutions: The datacenter will enable rapid file retrieval. The datacenter will be capable of updates that do not create incompatibilities in the storage and data models. The datacenter will be capable of horizontal extension and transparency to the application program, isolation of bottom-layer extension from upper-layer business, as well as proportional transparent extension capacity and performance by means of adding more servers. Figure 3 Data Processing Flow of Hadoop-based Regional Health Information Platform Various medical and health service institutions in the region will call on services provided by the regional health information platform to upload health record files (in XML format) designed by HL7 CDA template to the regional health information platform. CDA file types acquired include registration of basic personal health record, outpatient abstract, prescription record, diabetes visit, diabetes follow-up visit and other types of data (files). After Registrations of basic personal health information: 10,000,000 Outpatient abstracts: 30,000,000 Prescription records: 100,000,000 Diabetes visits: 2,000,000 Diabetes follow-up visits: 24,000,000 6
files are uploaded, the key data element (META-DATA) will be abstracted via XML analysis to form a health record document library stored as XML files in Hbase. By doubling as a file retrieval service and a big data analysis service, the regional health information platform will serve many different users, including individuals (residents and patients), doctors, and health managers. Conclusion Intel and UFIDA Medical leveraged an Intel Xeon E5 Processor platform and Intel Distribution of Hadoop through repeated single business load tests, big data tests, optimizations and other technical means to successfully build a complete regional medical big data computing architecture at the Jinzhou Regional Health Data Center. The architecture can meet performance requirements of high concurrency retrieval and real-time data analysis of mass data (with more than 100 million records). The architecture based on Intel Xeon E5 Processor platform and Intel Distribution of Hadoop provides a smart health cloud service platform for data processing, retrieval, analysis and other data services to meet Jinzhou s healthy city goals. Values Business value optimizations through Intel and UFIDA Medical cooperation realize: Mass data storage: meets storage capacity needs of more than100 million records (files). Dynamic extension of data format: columnar storage provided by Hbase can easily cope with flexible adjustment of data format. Rapid retrieval of mass data: high concurrency parallel mass data retrieval meets real-time health record retrieval needs of residents and doctors. Statistical analysis: open statistical analysis framework is close to real-time statistical analysis capability. Smooth capacity expansion: outstanding horizontal expansion performance copes with ever-growing business and data sizes. Economic value Reduction in cost of hosting storage construction: compared with traditional databases and minicomputer solutions, Intel big data solution boasts better openness and better economy. High scalability and integrated graphic interface: provides convenient node management and horizontal expansion, greatly reducing management cost. Full utilization of existing hardware resources: reduces upgrading costs. The business value of an integrated solution based on big data optimizations achieved through cooperation by Intel and UFIDA Medical are as follows: Enables doctors to rapidly retrieve basic personal information, existing medical history, treatments, prescriptions and other important information of patients, helping doctors make more accurate diagnosis via auxiliary information, and effectively avoiding repeated medication, adverse drug reaction, etc. Supports high concurrency big data retrieval, making it possible for individuals (residents and patients) to access to their own health records in a convenient and rapid way, enabling health education to help residents develop an awareness of personal health management and transform from a treatment approach to a prevention approach. Supports flexible data analysis modeling and data mining, providing data analysis capabilities for regional health managers, and can be widely applied in disease control, medical behavior supervision, medical quality management and other fields. IT value Integrated hardware and software solutions based on big data 7
References 1) Guide for Health-record-based Regional Health Information Platform Construction (Trial), Ministry of Health 2) Technology Solution of Health-record-based Regional Health Information Platform Construction (Trial), Ministry of Health 3) Technology Solution of EMR-based Hospital Information Platform Construction (Version 1.0), Ministry of Health 4) Regional Health Information Platform Whitepaper, Intel 5) Healthy China 2020 Strategic Research Report, Editorial Committee of Healthy China 2020 Strategic Research Report 6) China Medical Industry IT Solution Market Forecast and Analysis 2012-2016, IDC INFORMATION IN THIS DOCUMENT IS PROVIDED IN CONNECTION WITH INTEL PRODUCTS. NO LICENSE, EXPRESS OR IMPLIED, BY ESTOPPEL OR OTHERWISE, TO ANY INTELLECTUAL PROPERTY RIGHTS IS GRANTED BY THIS DOCUMENT. EXCEPT AS PROVIDED IN INTEL'S TERMS AND CONDITIONS OF SALE FOR SUCH PRODUCTS, INTEL ASSUMES NO LIABILITY WHATSOEVER AND INTEL DISCLAIMS ANY EXPRESS OR IMPLIED WARRANTY, RELATING TO SALE AND/OR USE OF INTEL PRODUCTS INCLUDING LIABILITY OR WARRANTIES RELATING TO FITNESS FOR A PARTICULAR PURPOSE, MERCHANTABILITY, OR INFRINGEMENT OF ANY PATENT, COPYRIGHT OR OTHER INTELLECTUAL PROPERTY RIGHT. A "Mission Critical Application" is any application in which failure of the Intel Product could result, directly or indirectly, in personal injury or death. SHOULD YOU PURCHASE OR USE INTEL'S PRODUCTS FOR ANY SUCH MISSION CRITICAL APPLICATION, YOU SHALL INDEMNIFY AND HOLD INTEL AND ITS SUBSIDIARIES, SUBCONTRACTORS AND AFFILIATES, AND THE DIRECTORS, OFFICERS, AND EMPLOYEES OF EACH, HARMLESS AGAINST ALL CLAIMS COSTS, DAMAGES, AND EXPENSES AND REASONABLE ATTORNEYS' FEES ARISING OUT OF, DIRECTLY OR INDIRECTLY, ANY CLAIM OF PRODUCT LIABILITY, PERSONAL INJURY, OR DEATH ARISING IN ANY WAY OUT OF SUCH MISSION CRITICAL APPLICATION, WHETHER OR NOT INTEL OR ITS SUBCONTRACTOR WAS NEGLIGENT IN THE DESIGN, MANUFACTURE, OR WARNING OF THE INTEL PRODUCT OR ANY OF ITS PARTS. Intel may make changes to specifications and product descriptions at any time, without notice. Designers must not rely on the absence or characteristics of any features or instructions marked "reserved" or "undefined". Intel reserves these for future definition and shall have no responsibility whatsoever for conflicts or incompatibilities arising from future changes to them. The information here is subject to change without notice. Do not finalize a design with this information. The products described in this document may contain design defects or errors known as errata which may cause the product to deviate from published specifications. Current characterized errata are available on request. Contact your local Intel sales office or your distributor to obtain the latest specifications and before placing your product order. Copies of documents which have an order number and are referenced in this document, or other Intel literature, may be obtained by calling 1-800-548-4725, or go to: http://www.intel.com/design/literature.htm%20 Intel, the Intel logo, Xeon, are trademarks of Intel Corporation in the U.S. and/or other countries. *Other names and brands may be claimed as the property of others. Copyright 2013 Intel Corporation. All rights reserved.