국가임상시험지원재단

● Data Ecosystems and Data Sharing to Enable Practical Medical Analytics

Kenneth Paik(Massachusetts Institute of Technology)

Machine learning in healthcare is rapidly expanding and will be one of the most impactful tools to effect significant improvements in care delivery. While clinical trials continue to be the gold standard for inferring causality, they are not adept at demonstrating small effect sizes across a population, particularly given the typical heterogeneity of treatment effect. With the proliferation of digital data, including monitors and EHR data, rigorous observational analysis can be a wellspring for process improvement and knowledge discovery.
However, the underlying challenge is that the potential of any analytics depends on the availability of data, more than the actual methods. While many institutions are generating massive amounts of data, very few have the infrastructure or resources in place to manage and leverage the data. There needs to be increased data access and sharing, but this in itself has significant technical and political challenges, not the least of which include privacy and security. A future health data ecosystem requires a widespread commitment of all the stakeholders across government and industry to realize the true value of health data.

● Medical Big Data in the Field

Hyung Jin Yoon(Seoul National University)

In many areas, big data has driven a number of technological innovations, and now big data is no longer a buzzword. Unlike earlierthere is still a big gap between the bench and the bed. The delay of big data's innovation in the medical arena may be due to many reasons including privacy issues, difficulties of proving causality, and other technical issues. Despite these difficulties, the gap between the bench and the bed is gradually being filled, and some of these efforts are introduced here.
Patient data stored in many hospitals is probably the most important source of medical big data, but unfortunately, these data
are scattered among hospitals. Patient data stored in each hospital is often too small to be used in artificial intelligence, so it is necessary to collect the data in one place. Collecting data from multiple hospitals to the one place is often very difficult due to technical reasons, such as the difference in the database structure of each hospital, as well as privacy and conflicting interests among hospitals. Many common data models have been proposed to overcome this difficulty. The Observational Medical Outcomes Partnership (OMOP) Common Data Model (CDM) of Observational Health Data Sciences and Informatics (OHDSI) program is a representative example. OHDSI program, which is a multi-stakeholder, interdisciplinary collaborative effort to bring out the value of health data through large-scale analytics, has established an international network of researchers and observational health databases. Each hospital converts its electronic medical records to the OMOP CDM structure, stores it on a separate server, and periodically updates it. If a researcher wishes to analyze the data of the hospitals in the network, he or she can write a query statement according to the structure of the OMOP CDM and send it to each hospital. The medical institutions that have received the query statement independently determine whether or not to participate in the study, and at the hospitals that have decided to participate, the query statement is executed in the OMOP CDM server and the results are sent to the requesting researchers. The researcher who originally sent the query statement finishes the study by pooling the query results. With this system, the raw data of the hospital is kept within the hospital, and it is free from personal privacy issues and at the same time, it can prevent data being used for other than its intended purpose so that shared data can greatly contribute to the facilitation of research. 

1. Kenneth Paik(Massachusetts Institute of Technology)

2. Hyung Jin Yoon(Seoul National University)

상기 과정은 보수과정으로 직능별, 단계별 인정가능시간이 상이함으로 관련규정 '[별표1] 임상시험등 종사자별 교육과정 및 이수시간'을 꼭 확인하시기 바랍니다.

[별표 1]

임상시험등 종사자별 교육과정 및 이수시간

(제5조 관련)