Lead/Presenter: Vanessa Stevens, COIN - Salt Lake City
All Authors: Malone C (VA Salt Lake City Healthcare Center), Lumby C (VA Office of Healthcare Transformation (OHT)) Richards G (VA Office of Healthcare Transformation (OHT)) Pham RH (VA Business Intelligence Service Line (BISL)) Stevens V (Informatics, Decision-Enhancement and Analytic Sciences (IDEAS) Center, Veterans Affairs (VA) Salt Lake City Health Care System; Division of Epidemiology, Department of Internal Medicine, University of Utah; VA Office of Clinical Systems Development and Evaluation (CSDE)) Peterson K (VA Office of Analytics and Performance Integration (API); Division of Epidemiology, Department of Internal Medicine, University of Utah) Jones K (VA Office of Health Equity (OHE)) Hardan D (VA Business Intelligence Service Line (BISL)) Matthew Wollner (VA Office of Intormation and Technology (OIT)) Augie Turano (VA Office of Information and Technology (OIT)) Makoto Jones (VA Office of Clinical Systems Development and Evaluation (CSDE); Informatics, Decision-Enhancement and Analytic Sciences (IDEAS) Center, Veterans Affairs (VA) Salt Lake City Health Care System; Division of Epidemiology, Department of Internal Medicine, University of Utah) Ernest Moy (VA Office of Health Equity (OHE)) Gebhard G (VA Business Intelligence Service Line (BISL)) Gamage S (VHA National Infectious Diseases Service (NIDS)) Gary Roselle (VHA National Infectious Diseases Service (NIDS) Plomondon M (VA Office of Clinical Systems Development and Evaluation (CSDE)) Hubert T (VA Healthcare Operations Center (HOC)) Bartlett A (VA Healthcare Operations Center (HOC)) Box T (VA Office of Analytics and Performance Integration (API)) Wallace K (VA Office of Biosurveillance) Francis J (VA Office of Analytics and Performance Integration (API))
Objectives:
The COVID-19 pandemic created an urgent need for surveillance data at unprecedented scale. In response, a multi-office team constructed the VA National Surveillance Tool (NST), a centralized source for VA COVID-19-related data. We describe the development of the NST, including how we used unstructured data and new data models when initial schemas for pandemic surveillance proved insufficient in representing our evolving understanding of the disease or its disparate impacts on certain groups.
Methods:
Initial NST surveillance efforts relied primarily on laboratory results, with time series data displays contextualized with clinical disposition and age breakdowns. Ultimately, laboratory data were augmented by community case capture using artificial intelligence methods confirmed by clinical review. As the pandemic intensified, resource availability data (e.g., ICU capacity, PPE, ventilator use) were centralized to coordinate an enterprise-wide response. New data needs (e.g., vaccination) and definition changes (e.g., CDC COVID-19 episode definitions) required further adaptations and new data views and tools. As evidence of disparities emerged, race and ethnicity were classified as per guidance from Office of Health Equity and added as a feature across the NST. To explore quality, disparities, and understand patient stories, a utility was created to analyze the administration of therapies and preventative measures across demographic groups in clinical context.
Results:
As of July 8, 2022, the NST has identified over 688,000 COVID-19 cases that can be analyzed in VA- and public-facing tools. NLP case identification now leverages increased use of standardized data elements to reduce annotator workload. Data visualizations have expanded to include facility- and patient-level views, and tools to support risk assessment in CLCs and in the community. The NST supplies data to the public through the VA Access to Care website. As an example of analyses of equitable therapy enabled by NST, we found that, since July 1, 2021, out of all acute inpatient VA COVID-19 hospitalizations, 37% of Black or African American Veterans hospitalized with COVID-19 received dexamethasone, compared to 42% among all Veterans. Ongoing development seeks to meet emerging challenges of vaccination, reinfections, and COVID-19 variant monitoring.
Implications:
The NST had to be continually adapted to meet the needs of our healthcare system as it increased in scale and impacted every service. Artificial intelligence methods were used but they had to be continually adjusted and human experts were needed to maintain high surveillance data quality. NST reports facilitate exploration of disparities in an environment where standards of care can change rapidly.
Impacts:
The NST has served as a centralized data hub and authoritative source for COVID-19 data across the enterprise. NST features expand on traditional surveillance paradigms by facilitating investigation of disparities in treatments, outcomes, disease phenotypes, and resource allocation. Data models, extraction methods, and analytic tools developed for the NST may generalize to other efforts that must also grow, be sustained, and integrate equity while managing a flood of data.