Researchers test and validate platform for potential PPE tracking across U.S. hospitals

A multidisciplinary team that includes a Vanderbilt computer science professor has established the foundation for an automated, up-to-date assessment of personal protective equipment across U.S. hospitals—work that got its start before the COVID-19 pandemic but took on greater urgency.

Significantly, the team developed a secure, third-party system to operate independent of federal and state governments as well as protect the proprietary information of hospitals and hospital systems. No universal database of U.S. hospital PPE data currently exists.

Kelly Aldrich

“This could be extremely impactful and important,” said Kelly Aldrich, associate professor of nursing informatics “This will direct the nation’s work in this area.”

Dana Zhang, assistant professor of the practice of computer science, supervised a group of four Data Science Institute master’s degree students who wrangled the data: Sarah Torrence, Logan King, Donnie Sengstack, and Li Yuan.

“The team did a lot of research to understand and standardize the PPE information,” said Zhang, PhD’18. “In this project, students were given the opportunity to practice their data science skills on real-world data that has such a large national impact.”

Researchers reported on the proof-of-concept phase, in the journal Health Security in November 2021, in “Lessons Learned from the Development and Demonstration of a PPE Inventory Monitoring System for U.S. Hospitals.”

Dana Zhang

The 2009 H1N1, 2014 Ebola, and 2015 Zika epidemics had already highlighted vulnerabilities in PPE management and supply. With that backdrop, the CDC’s National Personal Protective Technology Laboratory tapped Aldrich and The Center for Medical Interoperability, based in Nashville, to look at hospital PPE data with an eye toward planning for a pandemic. That was early in 2019, before the first reports of COVID-19 in China. By spring 2020, as COVID swept the U.S. and hospitals scrambled for masks and face screens, the plan had changed—the team had a much shorter timeline.

The team analyzing the data leveraged work done for the CDC by C4MI, which created a system that can not only provide a real-time snapshot of surgical masks, N95 masks and face shields but also predict where shortages were likely. The collaboration was a continued contract with CDC and Aldrich, who worked with Zhang and the graduate students to accomplish the analysis.

Development, deployment, and verification involved 78 hospitals that had agreed to participate with C4MI, a non-profit organization based in Nashville, and share their PPE data: 66 hospitals in a nationwide health care system, 11 in a regional network, and a stand-alone rural hospital worked.

The team developed the Healthcare Trust Data Platform and a mobile- friendly app with which hospitals reported near-daily inventory for N95 respirators, surgical masks, and face shields. In all, they provided 159 different PPE model numbers, and the team verified three of every four. Data was cross-checked and manually corrected.

With cleaned-up data, researchers developed algorithms to calculate individual PPE units replaced and hospital burn rate estimates for the three types of PPE. Reporting was automated, with data points converted to a common data model and results transferred to the app for access by study participants. Aggregated weekly data was reported to NIOSH for each of the 3 types of PPE.

The hospitals worked closely with the research teams to demonstrate how attributes of the platform would help in making strategic decisions during a pandemic.

The crunch of time and condition of the raw data dictated putting aside other inquiries, including whether and how on-hand PPE data correlated with COVID hospital admissions and infection rates. That could be addressed in a future study.

However, the team is analyzing data collected over the 15-week study to glean potential insights, which will be the subject of a subsequent journal article.

“In analyzing the data with advanced analytics, we will be able to find patterns that were not seen before,” Aldrich said. “I think because of that, it will have a true impact on supply chain management for the country.”