Is a syndromic approach well suited to foodborne disease surveillance? Implication for salmonellosis surveillance and prevention in France from a “One Health” perspective

In response to the increasing emergence of zoonotic pathogens, flexible, multisectoral surveillance systems capable of generating alerts thanks to rapid, nonspecific detection, are crucial before pathogens reach human populations. Syndromic surveillance has proven to be a breakthrough for near real-time disease surveillance in the public health sector. It relies on existing nonspecific data, usually collected for other purposes. Salmonella is one of the leading causes of foodborne illness. Data from animal and human sectors can be used to help monitor it across the food chain and thus quickly detect outbreaks. For the first time, this work uses five datasets identified after mapping farm-to-fork surveillance systems in France to apply syndromic surveillance to foodborne Salmonella outbreaks. The datasets covered on-farm cattle mortality, laboratory Salmonella isolations, and entries to public emergency services from 2011 to 2018. Weekly time series were retrospectively analyzed with five detection algorithms (Holt-Winters, historical limits, exponentially weighted moving average, Shewhart, and cumulative sum) to identify abnormal excess events over the last three years. Our analysis revealed several simultaneous excess events reported across algorithms and datasets from both animal and human health sectors. This shows the potential of syndromic surveillance for monitoring foodborne disease using a One Health approach. Nevertheless, data quality and real-time data collection are key to reliable syndromic surveillance. Temporal and geographic resolutions can also affect anomaly detection. The added value of the different algorithms tested underline the importance of using statistics with complementary characteristics.