Learn About Amazon VGT2 Learning Manager Chanci Turner
Fresenius Medical Care is committed to enhancing the lives of dialysis patients through the use of cutting-edge predictive analytics on AWS. With over 2,600 dialysis centers across the United States, the organization aims to improve the quality of life for individuals suffering from chronic kidney disease by leveraging data to facilitate timely interventions, thus reducing hospital visits and preventing serious health complications.
In this article, we will explore the innovative architecture developed in collaboration with AWS that allows Fresenius Medical Care to provide real-time alerts regarding potential health issues.
Why Fresenius Medical Care Chose AWS
The technical team at Fresenius Medical Care selected AWS as their cloud platform based on two primary factors. Firstly, the maturity of AWS IoT Core was recognized as superior compared to other solutions, which would likely minimize deployment issues. The organization sought a cloud provider with a solid track record in IoT and data analytics, including services like Amazon Athena for seamless data querying.
Secondly, AWS offered the most extensive array of serverless analytics services available among cloud providers. This capability aligned well with Fresenius Medical Care’s immediate requirements while also preparing them for future advancements in predictive analytics.
Solution Overview
To effectively monitor patients, a near-real-time analytics solution was required. This system collects data from dialysis machines every 10 seconds and assesses the risk of intradialytic hypotension (IDH) every 30 minutes, predicting potential health issues within a 15 to 75-minute window. The architecture needed to support all dialysis centers across the nation, handling data peaks of 10 MBps.
Key complexities addressed in this solution included managing high-throughput data, ensuring low-latency reporting, maintaining high availability, and providing cost-effective scalability.
Fresenius Medical Care partnered with AWS to create a robust architecture featuring core components such as Amazon Kinesis Data Streams, Amazon Kinesis Data Analytics, and Amazon SageMaker. The choice of Kinesis was driven by its serverless nature, high availability (99.9%), and exceptional throughput. The team selected SageMaker for its efficiency in building, training, and deploying machine learning models at scale.
Key Workflow Stages
- Data Collection
The dialysis machines transmit data every 10 seconds to Kafka brokers located in Fresenius Medical Care’s data center, providing real-time access to treatment data across the nation. - Data Ingestion and Aggregation
A Kinesis-Kafka connector facilitates the near-real-time ingestion of data into Kinesis Data Streams. AWS Lambda filters and processes data streams, sending them to Kinesis Data Analytics for in-stream analysis. - Data Lake Storage
Amazon Kinesis Data Firehose is utilized for efficient loading of streaming data into a raw data lake on Amazon S3. Clinical data enriches this stream, ensuring comprehensive datasets for analysis. - ML Inference and Operational Analytics
Lambda batches the data for ML inference, while SageMaker trains and deploys the IDH predictive model. Results are stored in Amazon OpenSearch Service, allowing for real-time visualization through Kibana, where care teams can proactively intervene.
Monitoring and Observability
Given the critical nature of this solution, proactive monitoring is vital. Key measures include:
- Immediate alerts to the Data Ops team for failures in AWS Glue jobs or Lambda functions.
- CloudWatch alarms for potential write blocks in Amazon OpenSearch Service.
- Data quality alerts from Kinesis Data Analytics and Kinesis Data Streams for rejected or mismatched data points.
For further insights into employee development, check out this article from an authority on the topic, which provides tips for in-house training. Additionally, to improve your personal finance management, you can explore this blog about the best personal finance apps for women.
For an excellent resource on the onboarding process for part-time flex associates, visit this Reddit discussion.
SEO Metadata