434. Tracking COVID-19 in Real Time: Leveraging Public Data Sources to Inform Infection Prevention Practices

BACKGROUND: Global spread of SARS-CoV-2 led to an urgent need for data on national and regional prevalence to inform public health policy. Healthcare systems were also in need of data to develop best practices around defining patient risk. We describe a data analytics tool developed at our institution which uses public data sources to track county-level prevalence of COVID-19 so as to delineate risk for individual patients. METHODS: We investigated a number of data sources tracking COVID-19 case counts, assessing for (1) frequency of updates, (2) granularity of geographic detail (optimally to zip-code or county) and (3) completeness of the data. We chose the Johns Hopkins University CSSE COVID-19 data set. This contains counts of new diagnoses per day by county using Federal Information Processing System (FIPS) codes. The dataset is updated daily with adjustments made for backdated corrections. We developed a data analytics tool which allowed for direct comparison of county period prevalence. We developed a metric of 10-day rolling period prevalence calculated as a total case count from the preceding 10 days, divided by county population from 2018 American Community Survey (ACS) estimates. RESULTS: Benchmarking against local (peak of 3.12 cases per 1,000 persons) and regional prevalence, we set 6 cases/1,000 persons as the threshold for a Geographic Region with Widespread Community Transmission (GReWCoT). Counties have to reach this threshold for at least 4 out of 7 days within the period 3 to 10 days prior to the evaluation, to adjust for bulking of test results and delayed reporting. We used the analytics tool to support a semimonthly review of geographic regions, and made specific recommendations for patients from qualifying regions including use of modified enhanced precautions (including surgical mask and eye protection), as well as restricted visitation of caregivers. Figure 1. Epidemic curves for 10-day rolling period prevalence of COVID-19 in the Mid-Atlantic Region: Philadelphia County, PA [Image: see text] Figure 2. Epidemic curves for 10-day rolling period prevalence of COVID-19 in the Mid-Atlantic Region: Westchester County, NY [Image: see text] Figure 3. Epidemic curves for 10-day rolling period prevalence of COVID-19 in the Mid-Atlantic Region: Bergen County, NJ [Image: see text] CONCLUSION: This approach allowed for a nuanced investigation of COVID-19 prevalence in real-time, and provided support for risk stratification of patients throughout our large catchment area. The dashboard was shared on an inward-facing site to support staff messaging about regions of increased risk. Next steps include leveraging international data to inform a similar approach to international travel for our patients and staff. DISCLOSURES: All Authors: No reported disclosures