Gold Coast University Hospital Optimizes ICU Efficiency Through Datarwe Solutions Running on AWS and Intel

Every day, Australia’s Gold Coast University Hospital (Gold Coast) collects data from thousands of devices, including bedside blood pressure and electrocardiogram (ECG) monitors and imaging equipment. As the hospital streamed and aggregated more data, researchers wanted to better analyze information in ICUs. “We had integrated regular sampling of this data into our electronic medical records system, but we wanted to go further by linking in full near real-time streaming data,” says Dr. Brent Richards, medical director of innovation and director of critical care research for Gold Coast University Hospital. Specifically, the hospital wanted more data about how patients are treated so it could gain better visibility into specific illnesses and treatment plans and enable better outcomes.

Gold Coast also wanted to use bedside medical data to create a new application that could provide increased visibility into supply and demand for beds (as a pilot for a state-wide system) so the hospital system could more efficiently treat patients in critical care. Dr. Richards explains, “Our clinical system only offered a single-patient view, and we needed a dashboard that showed a multi-patient view.” 

Gold Coast engaged AWS Partner Datarwe to solve its data challenges. Datarwe offers an acute care medical research data platform as a service based on AWS and powered by Intel Xeon processors. Datarwe’s Clinical Data Nexus (CDN) is a precision medical platform that integrates data from multiple sources, including electronic medical records and high-frequency Internet of Things (IoT) data from bedside devices. “The key to the platform is linking together IoT data from the devices, analyzing it, and presenting it in dashboard form,” says Kelvin Ross, PhD, chief technology officer for Datarwe. “This provides Gold Coast with a larger, more connected picture than just the device operating on its own.” 

The CDN solution is based on a data lake and machine learning (ML) pipeline engine running on Amazon Elastic Compute Cloud (Amazon EC2) and Intel-optimized containers in Amazon Elastic Container Service (Amazon ECS). Datarwe will eventually use the ML pipeline for data cleaning and enrichment, and plans to take advantage of Intel processors featuring Gaudi accelerators from Habana Labs, which are designed to support training deep learning models and can deliver up to 40 percent better price performance. 

The Datarwe CDN streams data from Amazon Kinesis and stores it in Amazon Simple Storage Service (Amazon S3) buckets for processing and reshaping. Data scientists use Amazon Athena to query the data within Amazon S3. 

Datarwe built ML pipelines on top of Amazon SageMaker, which helps researchers and data scientists train predictive models. The company also employs Amazon SageMaker Ground Truth, a data labeling service that makes it easy to build accurate training datasets. In addition, Datarwe developed a new clinical dashboard application that relies on ML insights to display updated ICU bed inventory data based on patient discharge or patient condition information. 

Using the Datarwe CDN platform powered by Intel-based Amazon EC2 instances, Gold Coast has the scalability and performance required to collect and analyze terabytes of data each week. “As we grow our medical research data platform from one ICU at Gold Coast to 13 ICUs across Queensland, the data requirements will grow dramatically,” says Ross. “We collect about 150,000 hours of ECG data just from one hospital’s ICU. We can easily support this because of the performance and scalability we get from AWS and Intel.” 

Taking advantage of the Datarwe solution, the hospital can gain new insights such as patient outcome or risk prediction and early warning of deteriorating patient conditions. Ross explains, “Through these insights, Gold Coast can more easily understand how illnesses respond to different treatments, or how to optimize drug dosage, for example.” 

Gold Coast is piloting the new clinical dashboard application to gain stronger visibility into supply and demand for beds in ICUs. For example, as ICU resources become constrained, clinicians anywhere in the health network can use the dashboard to see illness severity for patients and make faster, more accurate decisions on discharge timelines. “Because we can use the Datarwe dashboard to view illness severity, we can see that a unit might be full, but the patients in it are not all that ill. This will help us optimize supply and demand, which drives more efficiencies within the hospital, and ultimately offers the capability of ensuring the load is balanced between ICUs and hospitals,” says Dr. Richards. “The dashboard will eventually give us the opportunity to translate research into practice because we will have alerts in place. For instance, by using combinations of background and streaming physiology data we can in the future see if a patient is predicted to develop septic shock, and by alerting clinicians, give the opportunity for earlier assessment to potentially avoid further complications.” 

Next, Datarwe plans to expand the Clinical Data Nexus solution across other hospitals throughout Australia. “The world of medical IoT is in its infancy, and we have an enormous amount of equipment in hospitals that is not connected,” says Dr. Richards. “The Datarwe solution, because of the power of AWS and Intel, will help us start unlocking the potential of medical IoT data.”