The Representativeness of Regional Influenza Virus Genomic Surveillance for National Trends in the United States

This study demonstrates that intensive genomic sequencing of influenza viruses in a single U.S. state (Michigan) effectively captured nearly all national haplotype diversity from 2021 to 2025, indicating that regional surveillance can serve as a broadly representative proxy for national trends when sequencing effort is sufficient.

The Gist

The Big Question: Can One State See the Whole Picture?

Imagine the United States is a giant, bustling city where millions of people are constantly changing their outfits. These outfits represent different versions of the Flu Virus. Every year, health officials need to know exactly what "outfits" are popular to design the right vaccine for the next season.

The problem is that the U.S. is huge. It's hard to check everyone's outfit in every single state. Usually, different states check different people, and some states check way more people than others. This creates a worry: If we only look closely at one state (Michigan), will we miss the weird, rare outfits circulating in other states?

This paper asks: Is a deep, detailed look at the flu in Michigan enough to tell us what the whole country is dealing with?

The Detective Work: How They Did It

The researchers acted like super-sleuths. They gathered genetic "photos" (sequences) of the flu virus from Michigan and compared them to photos from all over the U.S. over four flu seasons (2021–2025).

Instead of looking at every single tiny detail of the virus, they grouped them into "Haplotypes."

• The Analogy: Think of a haplotype as a specific "style" of outfit. Maybe it's a red shirt with blue jeans and a hat. Even if the shirt has a tiny stain, it's still the same "style." The researchers found that in any given season, only a few specific "styles" (haplotypes) were actually worn by the vast majority of people.

The Big Discoveries

Here is what they found, broken down into simple concepts:

1. The "Headline News" Effect
Just like in the news, a few big stories dominate the headlines. The researchers found that a tiny number of virus "styles" made up most of the flu cases.

• The Metaphor: Imagine a fashion show where 90% of the models are wearing variations of the same three outfits. Even if you only watch the front row (Michigan), you are still seeing almost every outfit that is actually being worn on the runway.

2. Michigan Was a Super-Scanner
Michigan sequenced (photographed) way more viruses than any other state. Because they looked so hard, they caught almost every major virus style that was circulating anywhere in the U.S.

• The Result: If a dangerous new "outfit" appeared in California or New York, Michigan usually found it too, often within a few months. They didn't miss the big players.

3. It's About Effort, Not Geography
The researchers wondered: "Did Michigan find these viruses just because they are lucky, or because they looked harder?"

• The Finding: It was all about effort. The more samples a state tested, the faster and more completely they found the virus styles.
• The Analogy: If you are looking for a specific type of fish in a lake, you won't find it just because you are standing in the "right" spot. You find it because you cast your net many, many times. Michigan cast a huge net.

4. The "Rare" Stuff Doesn't Matter Much
They tried to see what would happen if Michigan stopped testing so many people. Even if they tested only 5% of the people they usually do, they would still catch the vast majority of the common virus styles.

• The Takeaway: You don't need to interview every single person in the country to know what the majority is wearing. You just need a good sample size.

Why This Matters

This is great news for public health.

• The Old Fear: We worried that because states test differently, we might miss a dangerous virus strain hiding in a state that doesn't test much.
• The New Reality: As long as some places (like Michigan, California, or New York) are doing a lot of testing, we can trust that data to represent the whole country. We don't need every single state to be a super-scanner to get a clear picture of the national flu trends.

The Bottom Line

Think of the U.S. flu situation like a massive library. You don't need to read every single book in the library to know what the most popular genres are. If one branch (Michigan) reads a huge number of books, they can tell you exactly what the whole library is reading.

In short: Intensive testing in one well-sampled location is enough to capture the diversity of the flu virus across the entire United States. We don't need to be everywhere to see the whole picture; we just need to look very closely in a few key places.

Technical Summary

1. Problem Statement

Influenza genomic surveillance is critical for vaccine strain selection, evolutionary forecasting, and tracking antigenic drift. In the United States, surveillance is decentralized: while the CDC coordinates national efforts, individual states collect and sequence viral samples.

• The Core Issue: Sequencing efforts vary drastically across states in terms of volume, sampling strategy (random vs. convenience), and timeliness. This creates a risk of spatial sampling bias, raising the question: Can a single, well-sampled region (like Michigan) accurately represent the genetic diversity of the entire nation?
• The Gap: There is a lack of empirical evidence validating whether dense sequencing in one location can substitute for fragmented national data in downstream modeling and evolutionary analysis.

2. Methodology

The study analyzed influenza virus genomes from four seasons (2021/22 to 2024/25) to evaluate the representativeness of Michigan's surveillance data against national trends.

• Data Sources:
  • Michigan Dataset: 10,958 genomes generated by the University of Michigan (UOM) from three major health systems (covering >4.5 million residents). Sampling was described as "deep" and "random."
  • National Dataset: 34,743 genomes from all U.S. states (H1 and H3 subtypes) sourced from GISAID and GenBank.
• Haplotype Definition:
  • Instead of using coarse WHO clades or analyzing single mutations in isolation, the authors defined seasonal haplotypes based on amino acid combinations at specific mutation sites.
  • Sites were identified using frequency thresholds (tested 0.005 to 0.2); a threshold of 0.1 was used for primary analysis, meaning a site was considered a mutation if at least two amino acids appeared with frequencies $\ge$ 10%.
• Analytical Approaches:
  • Detection Delay Analysis: Modeled the time lag between a haplotype's first national appearance and its first detection in Michigan using logistic regression (detection probability) and general linear models (delay duration).
  • Downsampling Simulation: The Michigan dataset was randomly reduced to 75%, 50%, 25%, 10%, and 5% of its original size to test the robustness of diversity capture at lower sequencing volumes.
  • Cross-State Comparison: Michigan's performance was compared to other states by downsampling Michigan to match the sample size of each state and calculating an Area Under the Curve (AUC) metric. This metric combined coverage (proportion of isolates captured) and speed (time to detection).
  • Rarefaction Analysis: Used the iNext R package to estimate the minimum number of sequences required to capture 95% of national haplotype diversity.

3. Key Contributions

• Empirical Validation of Regional Surveillance: The study provides quantitative evidence that a single, high-volume regional surveillance system can capture nearly all national genetic diversity, challenging the assumption that fragmented national sampling is strictly necessary for broad evolutionary insights.
• Haplotype-Centric Framework: The authors propose a flexible haplotype definition based on seasonal mutation frequencies, which balances the granularity of single-mutation analysis with the stability of traditional clade systems.
• Quantification of Diminishing Returns: The research identifies the specific point where increased sequencing effort yields diminishing returns in terms of new haplotype discovery.

4. Key Results

• High Representativeness: Michigan detected >90% of circulating U.S. isolates within one year of their national emergence. The mean detection delay was 107 days (SD=31).
• Dominance of Few Haplotypes: National diversity is driven by a small number of haplotypes.
  • 3 or fewer haplotypes accounted for 50% of circulating isolates in every season.
  • <20 haplotypes accounted for 90% of isolates.
• Drivers of Detection:
  • Haplotype Frequency: The primary predictor of detection was the frequency of the haplotype in the national population. High-frequency haplotypes were detected rapidly and consistently.
  • Sequencing Effort: Detection timeliness was strongly correlated with sequencing volume. Higher effort reduced delays, but with diminishing returns at high volumes.
  • Geography: The state of origin or geographic distance to Michigan did not significantly impact detection probability or delay once sequencing effort was accounted for.
• Downsampling Findings: Even when Michigan's sequencing was reduced to 5% of its original volume, the majority of circulating isolates still had a haplotype match in the dataset, though detection delays doubled.
• Rarefaction Analysis: Only 227 sequences (for H1 in 2023) were theoretically required to capture 95% of national diversity. Michigan consistently captured 99% of national diversity, outperforming all other states except California and New York (which also exceeded 95% coverage).
• Regression Model: A model including log-transformed sequence count, season, and subtype explained 52.4% of the variance in coverage (AUC). Adding specific state variables did not improve the model, suggesting that volume matters more than location.

5. Significance and Implications

• Policy and Resource Allocation: The findings suggest that public health agencies do not need to force random sampling in every state to get a representative national picture. Instead, concentrating resources on a few high-volume, well-sampled locations (like Michigan) may be more efficient for capturing national diversity.
• Vaccine Strain Selection: Since dominant haplotypes are captured quickly and consistently by high-volume regional surveillance, these data are sufficient for informing vaccine strain selection and evolutionary forecasting.
• Surveillance Strategy: While random sampling improves coverage at low sequencing volumes, once a threshold of sequencing effort is met, the specific sampling strategy (random vs. convenience) becomes less critical than the sheer volume of sequences generated.
• Future Directions: The study highlights the need to monitor low-sequencing states to ensure they do not miss rare, emerging variants, but confirms that for dominant circulating strains, regional hubs are highly effective proxies for national trends.

Conclusion: The paper concludes that intensive genomic sequencing in a single, well-sampled U.S. location is broadly representative of national influenza diversity, driven primarily by sequencing volume and haplotype frequency rather than geographic factors.