Socio-economic conditions were associated with substantial differences in state-level vulnerability to influenza-like illness across the US, with the District of Columbia identified as the most vulnerable jurisdiction in a machine learning–based analysis published in PLOS Computational Biology.
Researchers developed a vulnerability index integrating socio-economic and health indicators to estimate susceptibility to influenza-like illness (ILI) during 2022. The team combined influenza surveillance data from the Centers for Disease Control and Prevention with socio-economic indicators from US Census Bureau data and applied a Random Forest regression model to evaluate how these factors contributed to peak infection levels.
The researchers initially selected 39 socio-economic indicators and applied a variance inflation factor threshold of 10 to remove highly collinear variables, resulting in 22 indicators included in the final analysis.
Using the Random Forest model, the team generated feature-importance weights for each indicator and used these values to calculate a vulnerability index for every state. States were then classified into five vulnerability categories ranging from very low to very high.
The analysis identified 11 jurisdictions with very high vulnerability to influenza-like illness:
States With the Highest Flu Vulnerability
District of Columbia
Massachusetts
Hawaii
New Mexico
Rhode Island
Connecticut
Maryland
Oregon
Washington
Michigan
Arizona
Several indicators contributed strongly to the vulnerability index, including recent migration from abroad, lack of health insurance among native-born and foreign-born residents, the proportion of female residents, adult influenza vaccination coverage, and the percentage of the population aged 65 years or older.
Example analyses also showed that different factors influenced vulnerability across jurisdictions. In densely populated areas such as the District of Columbia, higher vulnerability was associated with population density, mobility, and uninsured populations. In states such as New Mexico, vulnerability was more closely linked to demographic factors including older populations, higher poverty levels, and limited access to digital technologies.
The researchers noted that some indicators showed counterintuitive relationships with reported influenza-like illness rates. For example, higher proportions of uninsured residents were associated with lower reported infection rates, which the researchers suggested may reflect reduced testing or underdiagnosis rather than lower disease burden.
The analysis relied on state-level data and a relatively small sample of about 50 state-level jurisdictions, which may limit model performance and obscure variation within states, the researchers noted.
The results underscore the need for tailored public health measures, including expanded health care access in rural regions, targeted vaccination campaigns, improved insurance coverage, and initiatives addressing gender-specific health care disparities, wrote lead study author Shrabani S. Tripathy, PhD, of Washington University in St. Louis, Missouri, and colleagues.
The researchers reported no conflicts of interest.
Source: PLOS Computational Biology
