Savana Builds World’s Only Natural Language Processing Clinical Research Network on AWS

Madrid-based Savana helps healthcare providers to unlock the value of their electronic medical records (EMRs) for research purposes. It combines research-grade methodology with natural language processing (NLP) and predictive analytics to obtain relevant results for healthcare and life science providers investigating disease prediction and treatment. Using Amazon Web Services (AWS), it can process the large volumes of data required to run machine learning algorithms. It can also scale its infrastructure to support its rapid growth across global markets, including the US, Latin America, and Western Europe, while satisfying local regulatory requirements. It has reduced IT costs by up to 90 percent and cut processing times for studies by 25 percent compared to its previous technology.

International medical technology company Savana provides a cloud-based solution that processes and analyzes unstructured clinical information contained in electronic medical records (EMRs) combined with other structured databases as genomics. Based in Madrid, Savana operates one of the largest artificial intelligence (AI), multicentric research networks in the world, which is comprised of more than 180 hospitals across 15 countries.

Savana processes billions of EMRs and data points through an automatic data pipeline populated by hospitals that are investigating disease prediction and treatment. Doing this successfully requires a reliable architecture that supports the volume of information and has the processing power needed to execute complex algorithms with the highest security standards. As the business is growing, its infrastructure also needs to scale and satisfy regulatory requirements across global markets around how data is stored, managed, and transferred.

Savana built its infrastructure using AWS because of its ability to scale and its breadth of services. It’s created a reliable and scalable solution while reducing costs by up to 90 percent and data processing times by 35 percent, so it can quickly deliver results for customers’ pioneering medical research.

Savana uses natural-language processing (NLP) to obtain the most relevant information about different diseases, conditions, and drugs from the medical language used by clinicians in EMRs. It then combines this information with other data layers, such as genetics and biomolecular data, and information sources such as clinical trials. It identifies patterns and trends that can inform predictive analysis—for example, detecting which types of patients would benefit the most from a specific treatment. It can also allow for biomarkers identification, which shortens drug discovery processes.

The company supports research studies conducted by healthcare providers and life science companies that want to use their own data to improve medical knowledge around a wide range of conditions and diseases, including onco-immunology, neurological, and cardiovascular disorders. 

Using AWS, Savana has implemented the world’s only multilingual NLP-based research network. The network’s data harmonization process involves making sense of information located in the sections of electronic medical records where clinicians add notes. The notes are written in natural language and follow no predetermined structure, and must be contextualized with structured data sources.

Savana has reduced costs and processing times on AWS. Savana can run fault-tolerant workloads for up to 90 percent less compared to on-demand pricing, using Amazon Elastic Compute Cloud (Amazon EC2) Spot Instances, which allows it to access spare Amazon EC2 compute capacity. It has also reduced data processing time by 25 percent, with a typical study now being processed in 9 weeks, compared to 12 weeks using previous technology.

This rapid turnaround means customers get their results for medical research more quickly, and Savana can service more studies and customers. “We can adapt the instances that we use at a really low cost and we can maintain the time the processing takes,” says Tello. “It’s really useful to have this predictability about how much money we will spend in the next month, or the next year.”

Savana has seen a 37 percent saving in total IT costs by using AWS services such as Amazon EMR, Amazon Relational Database Service (RDS), and AWS Fargate. In addition, it has achieved a 73 percent reduction in the overall time needed to execute and complete tasks across all Savana teams.