Skip to end of metadata
Go to start of metadata

David Raths


During the Precision Medicine World Conference, informaticians describe the rapid progress they made harnessing EHR data as well as their hopes for improved public health infrastructure


Informatics executives are making a valuable contribution to the pandemic response through the collection, dissemination and analysis of EHR and clinical registry data. During a recent panel discussion, leaders from several health systems also described the challenges and shortcomings they faced and a wish list for the future.

The June 15 Precision Medicine World Conference panel was hosted by Atul Butte, M.D., Ph.D., director of UCSF’s Institute for Computational Health Sciences. He asked his panelists to describe some of their accomplishments during the COVID era, and each respondent described significant changes their teams were able to make in short order.

For instance, Melissa Haendel, Ph.D., chief research informatics officer at the University of Colorado, who leads the National Center for Data to Health, described the number of people who came together to harmonize and aggregate data and create the National COVID Cohort Collaborative (N3C), which aims to take EHR data and harmonize it and bring it together and make it broadly accessible. “One of the overarching goals of this initiative was to create a fully transparent, reproducible, and broadly accessible electronic health record data repository of COVID patients and matched controls being drawn from a variety of different clinical institutions,” she said. “We're up to almost 90 institutions now that have signed on to the initiative. We now have over 2,000 people working on it. It demonstrates the sort of commitment and partnership between the community members, the research networks from the different common data models, the government centers for translational science award sites, and commercial entities all working collectively together in a rather unprecedented governance structure to enable the creation of this limited data set, which to our knowledge is now the largest publicly available, limited data set in U.S. history.”

Philip Payne, founding director of the Institute for Informatics at Washington University in St. Louis, also works on N3C. He described his organization’s work with partners at BJC Healthcare. “The real lesson learned for us was all about how we realign our priorities — how do we harness our capabilities in the informatics and data science research arena, and use them to tackle what was largely an operational problem? These included activities, such as bringing together all of the data across our regional health systems and brokering that data such that it would be available for broad use for hospital capacity planning, response planning, and then later in the pandemic, to help manage our public vaccine campaigns, all while at the same time making sure that we're able to do our core work of research around the pandemic and make that data quickly available to our investigators.”

Payne added that this wasn't just about providing data, but also about pilot funding. “We launched a series of just-in-time pilot funding mechanisms to bring together different disciplinary teams that could tackle fundamental problems such as one project that led to a predictive model to identify patients coming into our emergency departments that would benefit from palliative care as opposed to admission to the ICU, especially for those with multiple comorbid conditions and a high probability of mortality.”

In addition, Chris Longhurst, CIO at UC San Diego Health, and Jessie Tenenbaum, Ph.D., chief data officer for the North Carolina Department of Health and Human Services, described some of the challenges they overcame in tracking and reporting on COVID cases and vaccination rates.

Despite all the impressive work described, Butte asked the panelists what tools they wish they had and how we could best prepare for future challenges, including the next pandemic.

Longhurst noted that UCSD has all its students and employees on the same electronic health record that it uses for its patients. “That really helped to support us during the pandemic, because we had all that data about our testing and our vaccine administration in one place. However, I'd say that the vaccine data on our populations has been a real challenge. The State of California said if you're going to administer COVID vaccines, you need to report it to the registry. But, of course, these registries were built primarily for pediatric immunizations — low volume, and in many cases, not bidirectional interfaces. We really stressed those systems, and the public health infrastructure broke when we stressed them. So there was a period of time for a month or more when they were unable to send us that bidirectional data, and we couldn't integrate that, to understand the first- and second-dose gaps and things of that nature.”

Although they have done a lot of hardening of the system in California, they still have significant infrastructure issues. “For example, here in San Diego County, we have our own county registry that then reports data to the state,” Longhurst said. “And, of course, they use different identifiers so it's very hard to reconcile that. The one perhaps provocative suggestion I might make is that we shouldn't be doing anything at the county level. I don't think that our county IT colleagues really have the wherewithal from a security standpoint or data expertise standpoint. And I think that if we can harden that and centralize it at the state level, at least, that's going to make it easier to get the right data to the county public health leaders, but also make it easier for the health systems, which in most cases, span multiple counties.”

After describing some of the efforts the State of North Carolina has made to automate aspects of public health data reporting on COVID, Tenenbaum noted that although they now have all this data, they have a lot of trouble with linking it. “We have huge efforts right now at creating a master patient index. We don't have one identifier for vaccination data versus case data. And so for breakthrough cases, we're having to do a pretty manual matching probabilistic process. We're working with our state HIE, which is really good at this, to get that master patient index.”

Payne noted that his organization operates a health system that spans multiple states, about a 300-mile catchment area. “We see an extreme degree of variability in the vaccine registries and the ways in which vaccination information is being documented in EHRs, in the occupational health context or in student health contexts relative to our universities, not to mention what happens when we have large public vaccination events run either by the National Guard or various public health departments.” They work with five different public health departments within just the St. Louis metropolitan area.

“The timeliness of that data is very challenging based upon how it's collected and how it's submitted,” Payne stressed. “The way in which we can query it from the registries is highly variable. That leads to some very challenging situations where we're trying to understand what is our real state of vaccination, and in particular, in communities that are highly underserved or at risk. We are a microcosm of the U.S. healthcare system, where we have a lot of access, a lot of equity and a lot of disparity issues to navigate, particularly amplified in this campaign. We can't get to that data quickly enough. And I think it speaks volumes to the lack of a real interoperable healthcare data fabric at a regional or national level, which we all knew was the case. And we've just sort of amplified our understanding of that.”

Haendel closed by noting that a few of the themes that were mentioned included centralization and identifiers. “It's all about being able to harmonize information for asking important questions. In the face of a new disease, we really want to take advantage of all the different data that we can about an individual and a population. In healthcare i we take data that is about a patient. Now we have EHR data, we have imaging data, we have viral sequencing data, we have survey data. We ship those data to different places, never to be reconnected again. And so my hope in the future is that we have a system where we can put the patient back together again — my very technical term — to do the multimodal analytics, get the imaging machine learning working together with the clinical data at the scale of the whole nation. And that really does take that management of centralization into the state or the national level, and really good identifier management for security as well as for data analytics.”