Hancheol adds that building a healthcare data platform began with increasing the value of data collected in medical settings and sharing it in a usable environment. The goal was to create a collaborative network so that participating hospitals could safely manage data while allowing it to be used in various environments where data is needed. To this end, Kakao Healthcare conducted a project to advance data solutions with a local university hospital.

The system was initially designed around an open source database based on PostgreSQL to be run on-premises within the hospital for versatility. However, because this hospital was already using an in-memory data analysis appliance, it already had an environment with very fast data processing speed. By applying the open source-based PostgreSQL as is, it was challenging to upgrade or establish a data platform at the level of the existing environment.

“We were able to achieve similar performance through database tuning and hardware expansion, but we decided that it would not be appropriate to apply the same approach to all other hospitals. In the end, we decided that the cloud was most appropriate for all hospitals to quickly and easily build the same environment. This is when we considered Google Cloud, which specializes in data loading and analysis.”

Kakao Healthcare signed a business agreement with Google to strengthen its data and healthcare business, and through this, established itself as a partner that thinks and creates a healthcare data platform business together, taking their relationship beyond just being a cloud provider and user.

PostgreSQL is optimized for online transaction processing (OLTP), so if there are real-time analysis requirements, performance limitations would occur even if more central processing units (CPU) and memory resources were added. Meanwhile, AlloyDB can achieve higher performance with fewer resources, and has the advantage of being able to utilize the Columnar Engine without additional costs. Since AlloyDB is a fully-managed service, Kakao Healthcare saved time on maintenance and administrative tasks during the initial deployment.

Data extraction, conversion, and loading were done smoothly through Dataproc, making it advantageous to form a collaborative network where data from multiple hospitals were gathered to create greater value. Kakao Healthcare drew up an integrated data environment architecture using various data services from Google Cloud, and established a foundation that could be smoothly deployed and operated in actual medical settings.

Now, Kakao Healthcare has introduced its data platform to hospitals and records all clinical data. Through appropriate data preprocessing, all information that can identify patients is anonymized to fully protect privacy, and then loaded into AlloyDB in real time.

Through this data platform, patients’ diseases, treatment details, drug prescriptions, rehabilitation, hospitalization period, and cure and survival rates are organically loaded and created in AlloyDB. The clinical results of the treatment process or prescription can be analyzed, and through AlloyDB, a database is provided to judge various hypotheses and results in the medical environment.

“I have high expectations for federated learning based on collaborative networks. Even clinical trials for drug development cannot be conducted within a single hospital to sample patients taking a specific drug. Only when the number of participating hospitals and the size of data increases, will the effects of learning and analysis become meaningful. Google Cloud has provided an environment optimized for this federated learning,” says Hancheol.

With Google Cloud, Kakao Healthcare is now trying to collect hospital data. However, collaborative networks are not easily achievable because data is not only hospital assets, but also patients’ private information. Therefore, data sharing becomes challenging, not only in terms of regulation, but also data ethics and technology.

When the Google Cloud-based data platform is operated at each hospital, a model tailored to the research topic is developed, and this can be applied to each data platform to learn and improve the model to find the desired results. AlloyDB has become a way to increase participation in healthcare organizations without the data being combined, shared or transmitted. To this end, Kakao Healthcare is currently supplying data platforms to major hospitals in Korea and building a collaborative network.

“The AlloyDB-based data platform is being deployed to tertiary level general hospitals, including university hospitals and large hospitals, starting in early 2024. AlloyDB delivers performance comparable to commercial in-memory databases at a reasonable cost. From the developer’s perspective, we believe that data handling technologies are more easily accessible from a variety of angles, laying a solid foundation for upgrading solutions,” says Hancheol.

He explains that AlloyDB creates an environment where data can be loaded, processed appropriately, and analyzed. Beyond data processing, Google Cloud carried a series of processes to effectively introduce analysis results internally to other researchers.

In the long run, Kakao Healthcare plans to bring the value of medical data beyond Korea, into the global medical market through its data platform. Currently, clinical data is concentrated in specific countries, but if data analysis is conducted in countries where patients actually receive treatment and take medicine, the accuracy of data can be greatly improved.

Hancheol also sees an opportunity for medical institutions starting other businesses based on datasets. The company is focused on the possibility of turning the value of data into a business model.

“Combining AlloyDB, Dataproc, and AI services, we can create value for pharmaceutical and medical development companies that actually need extensive clinical data, participating in the development of a better medical environment while being profitable.”

Hancheol says that an environment in which hospitals and the medical market can continue to grow is needed. Medical IT is a sensitive and difficult field because it deals with life. For this reason, it has become natural to take a cautious and conservative approach to data management. However, the role of medical IT is to ensure that data creates great value for humanity. 

Kakao Healthcare is also upgrading various other services to create a data environment that is accessible to benefit everyone, along with its responsibility to provide medical care. Together with Google Cloud data services, including AlloyDB, the company is moving toward that goal.