New York City was hit hard and fast by COVID-19 in early 2020. At NYU Langone Health, the crisis was met with a swift, multi-pronged, analytics response. Clinical informaticists and analytics teams worked closely with hospital operations and institutional leadership to build a source-of-truth COVID-19 dataset for the institution, to identify and define key metrics, and build near-real-time dashboards.

Additionally, the teams constructed a de-identified COVID-19 data repository to support the many research projects related to COVID-19 and enhance education around informatics and data use.

Communicating data and analytics needs

"NYU Langone Health has existing processes in place that allow for operational leadership to communicate data and analytics needs to our medical center IT department," said Dr. Eduardo Iturrate, health IT safety officer and medical director for enterprise data and analytics at NYU Langone Medical Center.

"These processes include several committees that allow leadership to identify and describe needs to MCIT, as well as a reporting and metrics committee, which serves as the venue for governance and to rigorously define metrics to a degree that can be operationalized by reporting analysts to be built and then flow to the health system's existing dashboard infrastructure," he added.

The dashboard infrastructure is another essential piece of the puzzle that allows for rapid development of consumable business intelligence for health system leadership, he said.

"The dashboard system is a flexible, extensible system that has permission schemes built in to allow for the creation of publicly accessible dashboards as well as dashboards with a limited distribution list as defined by operations," he said.

"Finally, our underlying enterprise data warehouse is the backbone data infrastructure that underpins our systems ability to use data and analytics to inform leadership."

Several institutional challenges

The development of the de-identified COVID-19 data repository required surmounting several institutional challenges, some of them technical and others related to getting bureaucratic approval to move forward.

"In order to proceed, we needed to present the idea of developing the dataset, and in particular we needed to define the scope of use that we were proposing," Iturrate explained. "This was reviewed by our architecture governance committees, as well as by our legal and compliance departments.

"We developed a data use agreement for this effort that formalized the scope of use and, in collaboration with these groups, determined that we would restrict sharing of the dataset to only members of the NYU Langone Health community.

"This limitation guided our technical approach to de-identification and allowed us to define a degree of de-identification along the spectrum toward anonymization that satisfied institutional concerns related to exposing patient health information," he added.

"We also then needed to build tools and infrastructure to allow users to have secure access to the dataset, and to that end we created an approach to use a virtual desktop with pre-installed data analysis tools."

Finally, staff need to build out the support and systems to accommodate requests for the addition of data elements to the system and for the development and sharing of derived data elements, he said.

A big impact

In the end, these dashboards and de-identified data repositories could have quite an impact.

"Having operational dashboards with timely and accurate data about the patient care conditions that affect the organization allows operational leadership to make informed decisions regarding resource allocation and strategic planning," Iturrate said.

"This was crucial during the rapidly changing landscape of the evolving COVID-19 crisis, which forced our health system (and others) to reinvent itself rapidly to respond to the crisis.

"The de-identified data repository expands access to clinical data to the research community by providing a degree of self-service access to clinical data," he continued. "This approach improves the bottleneck that can exist related to requests for clinical data that are limited by the number of available data analysts."

Additionally, the de-identified data repository allowed staff to encourage collaboration and make connections between researchers and data scientists across all campuses and many academic departments of the large health system. The data repository allowed for the generation of several research manuscripts that may have been held up by usual research request processes.

Iturrate will offer more detail during his HIMSS21 session, "Case Study: Rapid Analytics Response to the COVID-19 Crisis." It's scheduled for August 10, 10-11 a.m., in Venetian Marco Polo 701.