A Pandemic Vulnerability Index

The Pandemic Vulnerability Index is a tool for policymakers to assess county-level vulnerability to public health shocks, providing a machine-learning-based, predictive roadmap to make preventative policy choices before the next crisis. Specifically, we believe that a predictive, data-driven approach to measuring vulnerability can provide granular and actionable insights into future funding decisions.

Our team used a wide variety of data sources to model how county-level factors had an impact on Covid-19 deaths, which serves as a proxy for susceptibility to any major public health disruption. We used ensemble models based on Random Forest and Gradient Boosted Tree models to generate our PVI scores.

The PVI is the first index of its kind to attempt to predict vulnerability, using broad descriptions of counties that include everything from census data to political sentiment. Other indices with similar aims tend to combine and display data that is assumed to correlate with vulnerability to crises, with poverty and demographic trends being the most common inputs. We believe that these approaches have several shortcomings, the most important of which is that they describe the past and leave the user to draw conclusions on their own. In contrast, the PVI is purpose built to provide decision makers with the actionable intelligence required to better the situation of the populations of vulnerable counties.