Every U.S. state has a unique flu vulnerability fingerprint, and a new index maps them all

PLOS Computational Biology, 2026. DOI: 10.1371/journal.pcbi.1013839

The District of Columbia, New Mexico, and Michigan are all highly vulnerable to influenza outbreaks. But the reasons could hardly be more different. D.C. faces risk from population density and a sizable uninsured foreign-born population. New Mexico's vulnerability stems from aging demographics and rural healthcare gaps. Michigan contends with the compounding effects of urban transmission and rural poverty.

A new vulnerability index, published in PLOS Computational Biology by researchers at Washington University in St. Louis, maps these differences with a precision that broad national metrics have never achieved.

Beyond the CDC's Social Vulnerability Index

The Centers for Disease Control and Prevention maintains a Social Vulnerability Index (SVI) designed to assess how well communities can withstand disasters. But the SVI was built for natural disasters broadly, not infectious disease specifically. Rajan Chakrabarty, the Harold D. Jolley Professor of Engineering at Washington University's McKelvey School of Engineering, wanted something more targeted.

His team built a flu-specific vulnerability index that integrates 39 socioeconomic and health indicators drawn from census data. The indicators span migration patterns, insurance coverage, population density, demographic composition (including proportions of elderly, female, and Hispanic populations), commute times, and healthcare access measures.

The critical methodological choice was using machine learning algorithms that can identify non-linear relationships between these factors. Traditional models tend to assume that each factor contributes independently and proportionally. But vulnerability to infectious disease does not work that way. The interaction between factors, density combined with low insurance rates, or poverty combined with limited healthcare access, can produce risks that are greater than the sum of their parts.

State-level fingerprints

The index produces a unique vulnerability profile for each state, which the researchers describe as a "fingerprint." Each fingerprint identifies the specific combination of factors driving that state's flu risk.

D.C. ranks as the most at-risk region. High population density and mobility allow viruses to spread easily. A sizable uninsured foreign-born population faces barriers to healthcare access. Longer commute times increase interpersonal contact during transit. These factors converge to create a distinctive risk profile that differs from any individual state.

States with large rural areas, such as New Mexico and Arizona, show elevated vulnerability for entirely different reasons. Their risk factors center on aging populations, higher proportions of Hispanic residents (a population with elevated flu complication rates), and geographic barriers to healthcare access.

Michigan illustrates yet another pattern: a state with both densely populated cities and economically distressed rural areas faces the dual challenge of high urban transmission risk and rural healthcare deserts. Lead author Shrabani Sailaja Tripathy highlighted this mix as particularly difficult for policymakers to address with a single strategy.

From mapping to action

The practical value of the index lies in its specificity. Rather than telling policymakers that a state is "vulnerable," it identifies which factors are contributing most to that vulnerability. This allows for targeted interventions: connecting uninsured populations to coverage in one state, improving rural healthcare access in another, expanding flu vaccination outreach to elderly populations in a third.

Chakrabarty noted that while every state will need some version of all these interventions, the index helps prioritize where extra resources might make the biggest difference. A state whose vulnerability is driven primarily by low insurance rates among foreign-born residents needs a different response than one whose vulnerability stems from an aging population with limited mobility.

The researchers also suggest the index framework could be adapted for other infectious diseases, providing vulnerability assessments for any pathogen where socioeconomic and health factors influence spread and outcomes.

What the index does not capture

The index is built from census data, which means it captures structural vulnerability rather than real-time conditions. Actual flu activity in any given season depends on additional factors, including the circulating viral strains, vaccine match and uptake rates, and weather patterns, none of which the index incorporates.

The model also operates at the state level, which may mask significant within-state variation. A state's overall vulnerability score might obscure hotspots or pockets of resilience within its borders. County-level or even neighborhood-level analysis would provide more actionable granularity, though it would require correspondingly more granular data.

The machine learning approach identifies associations rather than causal mechanisms. The index reveals that certain factor combinations correlate with higher flu vulnerability, but it does not prove that modifying any single factor would reduce risk by a specific amount. Interventions based on the index will need to be evaluated through their own studies.