Data driven assessment of community needs across multiple axes; currently focused on COVID-19 but will be useful for other pandemics or disasters.
This project is an extension of work completed by Community Insight and Impact. The overarching goal is to update and expand upon an existing project including: expanding the existing dataset with new measures, fine tune existing indices & creating new ones, update & fine-tune machine learning models, validating indices, and fine tuning existing dashboard for user experience.
Expanding existing dataset: The Community Insight and Impact has selected new measures (from open source datasets) that we will be used to create new indices. These measures have been selected by the lit review team and will need to be incorporated into the existing dataset.
Refining existing metrics: Conduct research to identify new indicators for the economic harm metric.
Exploring advanced ML based models: Current supervised models predict ICU COVID admissions. We need optimize supervised methods to increase prediction accuracy of ICU COVID admissions. Furthermore we will use unsupervised methods to find commonalities between counties that score similarly on our 8 indices.
Longitudinal impact studies: As a further validation of the 8 indices, the goal of this analysis is to look for trends in the indices overtime and identify major shifts in the 8 indices. Once these shifts are identified, we will look for correlations between shifts and other changes in county characteristics.
Refining existing dashboard: The current indices are viewable through interactive ArcGIS and Dash dashboards. The dashboards will be updated with the additional 5 indices. Furthermore, two versions of the dashboard are already created and will need fine tuning. The first, is a pared down version intended for less technical user and will be designed to be as intuitive to use as possible. The ideal user of this pared down version will be using the information in how to allocate health resources or for researchers using the data in grant applications and research proposals. The second, will have more customization options and be able to present more information for the user, but will require more technical skill by the user. The intended user of this second version are researchers tying the 8 indices to their own research on community needs.
The organization is in process of securing partnerships with decision-makers.
Our current dataset draws from 10 different validated, open-source datasets from the CDC, FCC, Robert Wood Johnson Foundation, Johns-Hopkins University, and others. It contains socio-economic, demographic, health, and infrastructure information about every US county. We currently have over 50 variables and 8 constructed vulnerability indices. All data is open-source and county level.
Exploring advanced ML- based or statistical models: Use unsupervised models to better understand community similarities.
Time-series analysis: Longitudinal analysis that correlates major shifts in the 8 indices to changes in characteristics of county.
Longitudinal impact studies Replicate the vulnerability indices for extended time periods, look for major shifts in specific areas and correlate them with changes in county characteristics.
The updated data and metrics will be incorporated into our dashboard which is available free of charge to non-profits, health systems, and individual community members. We are currently developing partnerships with key non-profits and analytics organizations to understand how they can best use this tool. We use a phase structure to organize project work and are currently wrapping up Phase 2 which focuses on finalizing our three key metrics (COVID-19 case severity, economic harm, and mobile health). Phase 3 will begin August 1 and will focus on improving the other 5 metrics and transitioning to more advanced analyses.
Updates during project execution