The 2007 nephrotoxicosis outbreak associated with melamine and cyanuric acid adulteration of pet foods in the United States sparked an urgent need for a nationwide companion animal surveillance program. In 2016, we introduced a syndromic surveillance system based on a novel epidemiological algorithm, the proportionate diagnostic outcome ratio (PDOR). The PDOR procedure was validated using simulated outbreaks of foodborne illness (i.e., aflatoxicosis and gastrointestinal illness) in dogs and cats. In this study, we further evaluated the PDOR procedure using the 2007 melamine-related outbreak of nephrotoxicosis. The performance of the PDOR procedure was assessed by the time to alert and positive predictive value (PPV). Electronic medical records of dogs and cats seen at networked primary care veterinary hospitals across the United States were retrieved from a centralized database. The data of four relevant syndromic components: elevated serum creatinine concentration, vomiting, anorexia, and lethargy from July 28, 2006 to May 31, 2007 were prospectively analyzed using the PDOR algorithm. The results showed that the alerts generated from the analysis of elevated serum creatinine concentration could have led to an early detection of this nephrotoxicosis foodborne outbreak and were well matched to the reported timeline of the outbreak. Additionally, we also observed variations in the performance of the PDOR procedure across age of animals and syndromic components, with the PPVs ranged from 0.61 to 1.0. Combined with the findings from previous evaluations using simulated outbreak scenarios, this study provided additional evidence that the PDOR procedure can be applied in syndromic surveillance to effectively and accurately detect various types of foodborne illness outbreaks in companion animals. However, the interpretations of and responses to alerts require an understanding of clinical veterinary medicine and relevant syndromic knowledge, and should not be based solely on quantitative measures.

In an effort to recognize and address communicable and point-source epidemics in dog and cat populations, this project created a near real-time syndromic surveillance system devoted to companion animal health in the United States. With over 150 million owned pets in the US, the development of such a system is timely in light of previous epidemics due to various causes that were only recognized in retrospect. The goal of this study was to develop epidemiologic and statistical methods for veterinary hospital-based surveillance, and to demonstrate its efficacy by detection of simulated foodborne outbreaks using a database of over 700 hospitals. Data transfer protocols were established via a secure file transfer protocol site, and a data repository was constructed predominantly utilizing open-source software. The daily proportion of patients with a given clinical or laboratory finding was contrasted with an equivalent average proportion from a historical comparison period, allowing construction of the proportionate diagnostic outcome ratio and its confidence interval for recognizing aberrant heath events. A five-tiered alert system was used to facilitate daily assessment of almost 2,000 statistical analyses. Two simulated outbreak scenarios were created by independent experts, blinded to study investigators, and embedded in the 2010 medical records. Both outbreaks were detected almost immediately by the alert system, accurately detecting species affected using relevant clinical and laboratory findings, and ages involved. Besides demonstrating proof-in-concept of using veterinary hospital databases to detect aberrant events in space and time, this research can be extended to conducting post-detection etiologic investigations utilizing exposure information in the medical record.