NYU Langone Leaders Share Their Rapid Analytics COVID Response

On Aug. 10 at HIMSS21, Jeff Shein, senior director, data and analytics and Eduardo Iturrate, M.D., medical director for enterprise data and analytics from New York-based NYU Langone Health (NYULH) virtually presented an educational session titled “Case Study: Rapid Analytics Response to the COVID-19 Crisis.”

Iturrate stated that the first confirmed cases of the virus cases across NYULH were on:

• March 5, 2020: NYU Langone Hospital, Long Island
• March 11, 2020: NYU Langone Hospital, Brooklyn
• March 13, 2020: Tisch Hospital & Kimmel Pavilion

Tracking the initial cases, hospital operations leveraged Epic Reporting Workbench reports for real time data. Numbers were then compiled and emailed to clinical leadership and reports were executed twice per day.

Shein then began explaining the COVID-19 Leadership Scorecard. He said that “A few reports turned into a few more reports.” The scorecard was for strategy, planning, and hospital operations. The data from Reporting Workbench (RWB) reports were stored in Excel and converted to PDF. Past reports were used for history and these reports were manually executed, complied, then distributed twice per day. Reports were run for each of the three campuses. This was an emergency situation, so it bypassed the Reporting and Metrics Subcommittee (RMS).

Shein then explained the COVID-19 Leadership Dashboard Project. He said that “We needed to conceptualize and automate a live interactive dashboard with patient data. We needed more up to date data.”

• The requirements for the dashboard were:
• Convert existing scorecard into automated, interactive dashboard
• Near real-time updates
• Filters for time and campus, drilling down to case level detail
• Email snapshot reports twice daily

The challenges for the project were:

• No documented metric definitions
• Unclear business contacts for project team
• Real time data cannot use existing data warehouse ETL processes
• COVID-19 workflows and criteria were new and frequently changing
• Manual data
• Time

Shein commented that “There’s a lot of challenges in this kind of project. I’ve been working with analytics for 30 years and data was changing so rapidly.”

He then went on to explain the Reporting & Metrics Subcommittee (RMS). RMS ensures the consistency of enterprise reporting metrics across NYULH and is responsible for the management of the governance review of issues and proposals for enterprise metric changes, additions, or deletions.

NYULH uses a centralized data governance approach to maintain standardization and consistency across the enterprise:

• Every dashboard metric has a published definition with a business owner
• Each metric can have only a single definition
• Eight hundred organizational metrics and over 200 business term definitions, reviewed by business owners
• Data management through Collibra
• Business definitions are linked to dashboards and available to all applicable staff through a portal

Shein then explained that NYU Langone COVID-19 Analytics 2.0 is in process. Leadership is rebuilding the backend to simplify architecture and integrate it into a historical data warehouse. Clinical leadership is also regrouping to update or retire metrics. Additionally, COVID-19 patients are being tracked long term, in both inpatient and outpatient settings.

Iturrate then came back on screen and introduced NYULH’s deidentified data repository. He said that “The research mission has slightly different needs, as it relates to the use of clinical data and how we get it.” COVID hit at a crushing speed, therefore, the mission of the research and academic community was to investigate why some patients were sick and dying and some patients were not. Also, what medications may help patients. Again, COVID hit hard and fast and the health system was reorganizing to deal with it and the research community asked questions.

“Our usual processes for exploring data was overwhelmed, part of the usual practice is governance and oversight around requests for clinical data,” Iturrate said. “Because this evolved so quickly, and COVID data was needed, we came up with the idea of a deidentified data repository that could be broadly shared with the community. We came up with the idea to democratize access to the file, protecting the patients’ confidentiality and use of data.”

The data use agreement is:

• Dataset is not shared outside of NYULH
• Does not require IRB review for use for research
• If one sees something in the dataset that one believes may be identifying, immediately inform the dataset administrators
• Cannot explicitly attempt to reidentify patients

Lastly, Iturrate explained the “NYULH COVID-19 Data Challenge.” This was an invitation to all clinicians, clinical researchers, data scientists, biostatisticians, and students to propose research questions that can be addressed using the COVID-19 deidentified dataset, or to devise novel data visualizations and data science techniques that can be applied to glean insights from the NYULH’s experience combatting COVID-19.

One hundred fifty participants took part in the challenge and submitted 15 projects. Two projects were named as winners and received project support to submit manuscripts to peer-reviewed journals. There are 100 ongoing users of the dataset.