Multistate Animal-Contact-Related Nontyphoidal Salmonella enterica Outbreaks in the United States, 2009-2022: Network and Machine Learning Analyses of Exposure Sources, Settings, and Serovars

This study analyzes 177 multistate animal-contact Nontyphoidal Salmonella outbreaks in the U.S. from 2009 to 2022 using network and machine learning methods to identify distinct transmission clusters driven by specific serovars, animal sources (particularly birds and reptiles), and settings, thereby highlighting the need for region-specific One Health prevention strategies.

The Gist

Imagine the United States as a giant, bustling neighborhood where people and animals live side-by-side. Sometimes, a sneaky germ called Salmonella (specifically the non-typhoidal kind) jumps from animals to humans, causing sickness. While we often think of Salmonella coming from bad food (like undercooked chicken), this paper investigates a different, trickier route: direct contact with animals.

The researchers wanted to know: When a Salmonella outbreak hits multiple states at once, where does it come from, and how does it spread?

Here is the story of their investigation, broken down into simple concepts.

1. The Detective Work: Tracking the "Super-Spreaders"

The team looked at data from 2009 to 2022, like a detective reviewing 14 years of police reports. They found 177 major outbreaks that crossed state lines.

• The Trend: Imagine a hill. From 2009 to 2013, the number of outbreaks slid down the hill very quickly (a big drop in sickness). But after 2013, the hill flattened out. The number of outbreaks stopped dropping and just stayed steady. It seems the initial fixes worked, but the problem didn't disappear entirely.

2. The Two "Gangs" of Outbreaks

The most interesting part of the study is how they used a computer to sort these outbreaks into two distinct "gangs" or groups. Think of it like sorting a deck of cards into two different suits.

Group A: The "Backyard Poultry" Crew

• The Stars: This group is all about birds (chicks, ducks, chickens) and agricultural feed stores.
• The Vibe: These outbreaks happen mostly in the Midwest, Northeast, and South.
• The Culprits: Specific types of Salmonella (called Enteritidis and Infantis) are the main troublemakers here.
• The Story: People buy baby chicks or ducks from mail-order hatcheries or local feed stores. These birds are shipped across the country. If one bird is sick, it can infect a whole flock, which then spreads the germ to people in many different states. It's like a "chain letter" of germs traveling on trucks.

Group B: The "Exotic Pet" Crew

• The Stars: This group is all about reptiles (turtles, lizards), mammals, and residential homes.
• The Vibe: These outbreaks are concentrated in the West and South.
• The Culprits: Different types of Salmonella (like Hadar, Typhimurium, and Braenderup) are the main troublemakers.
• The Story: This is about people keeping exotic pets in their living rooms. Even though there are rules about selling certain reptiles, they still get sold through underground or informal channels. A sick turtle in a pet store in California can end up in a home in Texas, spreading the germ.

3. The "Social Network" of Germs

The researchers built a giant digital map (a network) to see how everything connects.

• The Hubs: Just like a popular person at a party who knows everyone, Virginia and Pennsylvania were the most connected states in the bird/feed-store group. Washington and Texas were the hubs for the reptile group.
• The Bridges: Some states acted like bridges, connecting different parts of the country. Kansas and Georgia were key bridges for the bird outbreaks.

4. What Does This Mean for You? (The Takeaway)

The study tells us that you can't fight these outbreaks with a "one-size-fits-all" plan. You need a tailored approach:

• If you live in the Midwest/South: The risk is likely from backyard chickens or feed stores. The solution is better hygiene when handling birds and stricter rules for hatcheries.
• If you live in the West/South: The risk is likely from reptiles and exotic pets. The solution is better education for pet owners and cracking down on illegal reptile sales.

The Big Picture Analogy

Imagine Salmonella outbreaks as forest fires.

• Group A is like a fire spreading through a farm via a strong wind (the mail-order bird supply chain). You stop it by fixing the wind (better hatchery safety).
• Group B is like a fire spreading through backyards via people carrying burning embers (exotic pets). You stop it by telling people not to carry embers (better pet safety education).

The Conclusion:
The germ is smart and adapts to different environments. To keep people safe, we need to stop treating all animal-contact outbreaks the same. We need to target the specific "gang" causing the trouble in your specific region, using a "One Health" approach that connects human health, animal health, and the environment.

Technical Summary

1. Problem Statement

Nontyphoidal Salmonella enterica (NTS) is a leading cause of foodborne illness in the United States. While foodborne transmission is well-studied, illnesses resulting from direct animal contact (zoonotic transmission) represent a significant and often under-recognized public health burden. These outbreaks are particularly challenging because:

• They frequently cross state lines due to the commercial distribution of live animals (e.g., day-old chicks from hatcheries, reptiles from breeders).
• They involve diverse exposure sources (birds, reptiles, mammals) and settings (farms, feed stores, homes).
• Previous surveillance often focused on single serovars or single states, failing to capture the complex, interconnected patterns of multistate transmission.
• There is a lack of long-term, serovar-level analysis linking specific animal contact sources to geographic clusters and outbreak profiles.

2. Methodology

The study analyzed data from the CDC National Outbreak Reporting System (NORS) covering the period 2009–2022. The analysis integrated epidemiological statistics, machine learning, and network science.

• Data Scope: 177 multistate animal-contact NTS outbreaks were identified. The dataset included outbreak dates, locations (states), exposure sources, settings, and serovars. Population data were sourced from the U.S. Census Bureau.
• Incidence Calculation: Outbreak incidence rates (IR) were calculated per 10 million population-years (MPY).
• Temporal Trend Analysis: Joinpoint Regression was used to identify significant changes in outbreak trends over time, calculating Annual Percent Change (APC) and Average Annual Percent Change (AAPC).
• Machine Learning (Random Forest): A Random Forest classifier was applied to a binary predictor matrix containing state indicators, exposure sources, and settings. The model used Mean Decrease in Gini and Permutation-based Mean Decrease in Accuracy to rank variables by their importance in distinguishing outbreak profiles.
• Network Analysis: A weighted, undirected co-occurrence network was constructed where nodes represented U.S. states, serovars, exposure sources, and settings. Edges represented shared participation in the same outbreak.
  • Metrics: Degree, strength, betweenness, closeness, and clustering coefficients were calculated.
  • Community Detection: The Louvain algorithm was used to identify clusters (communities) of densely connected nodes, revealing distinct transmission pathways.

3. Key Contributions

• First Long-Term Multistate Analysis: This is the first study to comprehensively analyze long-term (14-year), serovar-level animal-contact NTS multistate outbreaks in the U.S.
• Integrated Analytical Framework: The study uniquely combines epidemiological trend analysis with machine learning (for feature importance) and network analysis (for structural clustering) to map transmission dynamics.
• Identification of Distinct Transmission Pathways: The research moves beyond general associations to define specific "outbreak profiles" characterized by unique combinations of serovars, animal sources, settings, and geographic regions.
• Quantification of Drivers: It quantitatively identifies birds and reptiles as the primary drivers of multistate spread, distinguishing them from mammal-associated outbreaks which tend to be more localized.

4. Key Results

A. Epidemiological Trends

• Total Burden: 177 multistate outbreaks were identified (0.06 per 10 MPY) involving 40 serovars.
• Temporal Trends: Joinpoint regression revealed a significant decline in incidence from 2009 to 2013 (APC = -10.4%, p=0.001), likely due to improved hatchery biosecurity and traceback capabilities. From 2013 to 2022, the trend stabilized with no significant change.

B. Machine Learning Findings (Random Forest)

• Top Exposure Sources: Birds and Reptiles were the most influential variables for distinguishing outbreak profiles.
• Top Exposure Settings: Agricultural feed stores and Residential homes were the most significant settings.
• Geographic Importance: States such as Ohio, Indiana, Nebraska, Vermont, Pennsylvania, and Texas showed the highest variable importance scores in classifying outbreak patterns.

C. Network Analysis & Community Detection

The network analysis revealed two distinct Louvain communities representing different transmission ecosystems:

• Community 1 (Midwest, Northeast, South):

  • Serovars: Enteritidis, Infantis, Altona, Mbandaka, Muenster, Agbeni, Johannesburg.
  • Sources: Primarily Birds.
  • Settings: Agricultural feed stores, farms/dairies, schools, and daycares.
  • Geography: High connectivity in Virginia, Pennsylvania, Ohio, Kentucky, and surrounding states.
  • Interpretation: Reflects outbreaks driven by the commercial distribution of live poultry (e.g., mail-order hatcheries) affecting backyard flocks and agricultural settings.

• Community 2 (West, Mountain West, South):

  • Serovars: Hadar, Typhimurium, Braenderup, Berta, Pomona, Montevideo, Newport, Indiana, I 4,[5],12:i:-, Thompson, Poona, Uganda, Cotham.
  • Sources: Reptiles, Mammals, and Amphibians.
  • Settings: Primarily Residential homes.
  • Geography: Concentrated in Washington, Texas, California, Oklahoma, and surrounding western/southern states.
  • Interpretation: Reflects outbreaks driven by the pet trade (exotic pets, reptiles) and illegal/unregulated distribution channels, often linked to household transmission.

• Network Metrics:

  • Birds had the highest degree (59) and strength (3733), indicating they are the most central node in the network.
  • Virginia had the highest state degree (67) and strength (3465).
  • Amphibians and Mammals showed high betweenness centrality, acting as bridges in the network, though less frequently than birds.

5. Significance and Public Health Implications

• One Health Approach: The study underscores that NTS control requires a region-specific One Health strategy. A uniform national approach is insufficient because transmission pathways differ significantly by region (e.g., poultry in the Midwest vs. reptiles in the West).
• Targeted Interventions:
  • Upstream Control: Strengthening biosecurity at mail-order hatcheries and reptile breeding facilities is critical to stop interstate spread before it reaches consumers.
  • Traceback: Enhanced tracing capabilities are needed for live animal distribution chains, particularly for reptiles and amphibians where illegal trade complicates investigations.
  • Education: Public health messaging should be tailored: focusing on agricultural feed stores and backyard poultry in the Midwest/Northeast, and on pet handling and exotic pet sourcing in the West/South.
• Surveillance Gaps: The study highlights the need for better reporting of exposure sources and settings in NORS to reduce underreporting and improve source attribution.

Conclusion: Multistate animal-contact NTS outbreaks are not random; they cluster into distinct, geographically and biologically defined communities. Understanding these specific pathways allows public health authorities to move from reactive outbreak response to proactive, targeted prevention strategies.