Inferring source attribution from a multiyear multisource data set of Salmonella in Minnesota

Area of Science:

• Foodborne illness
• Epidemiology
• Computational biology

Background:

• Salmonella enterica is a significant cause of global foodborne illness.
• Bayesian models are used to attribute human salmonellosis cases to specific sources for intervention prioritization.
• Evaluating data quality and model logic is crucial for accurate source attribution.

Purpose of the Study:

• To attribute human salmonellosis cases to specific sources in Minnesota using a modified Bayesian model.
• To understand pathogen population features, data gaps, and inform policy for Salmonella control.
• To develop a visual application for exploring large, multiyear datasets.

Main Methods:

• Analysis of over 12,000 non-typhoidal Salmonella isolates from human and animal sources in Minnesota.
• Application of a modified Bayesian source attribution model accounting for non-sampled sources.
• Utilized molecular epidemiological methods and developed a visual attribution application for data exploration.

Main Results:

• Chicken was attributed as the source for 60% of the 4,672 human Salmonella cases analyzed.
• A spike in cases attributed to a non-sampled source was observed in the latter half of the study period.
• High within-source diversity and low between-source similarity were noted, with visual exploration aiding interpretation.

Conclusions:

• This study provides the first Salmonella source attribution estimates for Minnesota.
• Findings highlight the importance of chicken as a primary source and identify data gaps for future research.
• Results will inform public health policies and management strategies to control Salmonella infections in the state.