By: Louisa Gaylord
Over the past year, the University of Washington and UW Medicine have been at the heart of the city of Seattle’s Covid-19 response: researchers studying social distancing and how the virus spreads on our communities, volunteers collecting and distributing supplies, clinical vaccine trials, and a partnership with the city to process over 2 million test samples at no cost to residents – all of which have helped Seattle maintain one of the lowest Covid-19 rates of major US cities. And with so many departments and teams working together, often involving clinical patient data, an adaptable and secure info-sharing system is absolutely necessary.
Most people are familiar with cloud storage where photos, documents and other files can be stored on a remote server and retrieved in the future from any number of devices. On the other hand, cloud computing can be likened to the world’s largest research supercomputer, processing vast amounts of information with far more computing power, storage and network capacity than any traditional campus data center could dream of.
Cloud computing is based on “renting out” unused computing capacity in huge data centers operated by companies like Amazon, Google or, in this case, Microsoft. “The most notable difference is [the cloud] removes the need for a large upfront investment in hardware,” says Robert Fabiano, Senior Systems Engineer of TECHdesk, the IT services shared by the UW Medicine Departments of Anesthesiology & Pain Medicine (APM) and Surgery. “You don’t need to invest in the infrastructure that comes with owning or renting a data center, such as electricity, cooling or physical security.” Cloud computing also makes it possible to quickly construct complex Digital Research Environments (DREs), where researchers can generate customizable workspaces and do science with cloud-based secure data, which otherwise would be cost prohibitive.
Designing a HIPAA-Compliant DRE
In March 2019, UW Medicine began the Azure cloud computing implementation process when Fabiano and Roland Lai, IT Director of TECHdesk, connected with the Microsoft Research (MSR) Lab – Cambridge team about a perioperative hypotension study that had been collecting data on blood pressure during surgery. The Microsoft team didn’t want the encrypted dataset transferred, but instead wanted UW to create a secure environment where they could access the data and utilize it for science.
After negotiations about how this new environment would be built and funded, the Microsoft team began to meet with TECHdesk, UW Medicine Compliance and UW Medicine Health Information Management in October 2019 to start designing a perioperative hypotension DRE using Azure cloud computing. The UW-IT professionals had hoped to use existing “blueprints” for a HIPAA-compliant Azure environment to implement at TECHdesk. Because protected patient health information is involved, all data is required to be authenticated, encrypted and decrypted, have activity logged and audited, and accounts automatically logged out of all devices. In exchange for expanding the scope of the project, MSR compensated Lai and Fabiano for their time, provided $60,000 in Azure credits to fund the DRE infrastructure, and assigned Senior Cloud Solutions Architect Clayton Barlow as a technical resource for their project to provide invaluable support throughout the process.
The team of UW-IT professionals who built out the project’s cloud infrastructure consisted of Lai and Fabiano with TECHdesk at UW Medicine. They handled the technical design, implementation and maintenance of the cloud infrastructure. Jimmy Phuong with UW Medicine Research IT made sure patient data was stored and retrievable in the form needed by the research stakeholders.
While introducing the general Azure design within the UW and UW Medicine, the UW-IT team “encountered significant obstacles that required novel engineering and business processes,” said Lai. The DRE “blueprints” provided by Microsoft, while providing a valuable starting point for the team, turned out to be very generic and much of the environment needed to be created from the ground up. Thanks to the dedication of multiple teams across the UW and Microsoft, the MSR-UW Perioperative Hypotension DRE was online and in use by January 2020.
Adapting for Covid-19 Research
In April 2020 as the novel coronavirus began to accelerate globally, MSR and UW Medicine APM began discussing the possibility of creating a second DRE for Covid-19 research. “One thing we realized early on in the pandemic was that UW Medicine did not have the infrastructure for minimum-risk research with external collaborators on Azure cloud systems, which was something we really needed,” said Phuong. “The components were there, but just not yet ready.”
The project received a rapid response research grant from the UW’s Population Health Initiative, and after completing an IT Security Risk assessment with UW Medicine ITS Security, the Covid-19 Triage DRE became available for use by researchers in June. “The pandemic has put SARS-CoV-2 testing and patient information at the center of attention as we seek to learn about the differences in outcomes, severity, and the changing situation,” said Phuong.
The Covid-19 DRE assists with allocating limited resources across departments – such as diagnostic tests, ICU beds and PPE – as well as creating a comprehensive coronavirus dataset from health records of UW Medicine patients that can be jointly utilized by departments like perioperative care, biomedical informatics and epidemiology. Research resource requests are submitted to the UW Medicine Research IT to coordinate projects within the DRE like developing virtual environments, engineering of databases and data collection tools, and fulfilling data extraction requests on UW Medicine electronic health records. The latest Covid-19 research and updates are collected to help collaborate across teams with data scientists, project managers, and clinical and epidemiology experts to help define and prioritize their research questions.
The dataset can move between secure environments, where it’s analyzed using Python. Additionally, the Covid-19 DRE utilizes a digital security layer and machine learning models, two services provided by Azure that would otherwise have had to be developed in-house at great time and expense.
“The cloud is the ideal place to carry out collaborative, secure research,” says UW Director of Cloud & Data Solutions Rob Fatland, who helps to facilitate the use of the public cloud for medical research at the UW and other affiliated medical institutions. “But that isn’t standard practice just yet, so this study has been a fantastic demonstration of feasibility.”
Cloud Computing at UW and Beyond
Because the MSR-UW Perioperative Hypotension DRE and the Covid-19 Triage DRE overcame unforeseen obstacles and created a lot of HIPAA-complaint infrastructure from scratch, these projects can be immensely useful in developing new templates for other organizations who are interested in using Azure DREs for similar research purposes. The success of this case study can be attributed to the dedicated collaboration of many teams across UW and MSR – Cambridge.
The eScience Institute was an enthusiastic advocate for the use of cloud computing with these projects, as the university’s hub for data-driven science and research. They have also been instrumental in developing the NSF-funded CloudBank, a suite of managed services to simplify public cloud access for computer science research and education, which provides access to not only Microsoft Azure but Amazon AWS and Google GCP as well. Learn more here.