Using Bayes' rule to define the value of evidence from syndromic surveillance.
Authors: Andersson Mats Gunnar, Faverjon Céline, Vial Flavie, Legrand Loïc, Leblond Agnès
Journal: PloS one
Summary
# Editorial Summary: Bayesian Approaches to Syndromic Surveillance in Equine Practice Determining whether a cluster of clinical cases represents a genuine disease outbreak or normal background variation is a persistent challenge in equine health surveillance. Andersson and colleagues adapted statistical methods from forensic evidence evaluation to quantify the diagnostic value of syndromic surveillance data, using Bayes' rule to calculate likelihood ratios that translate observed case numbers into posterior probabilities of an actual outbreak occurring. Applied to historical time series data for equine respiratory and neurological presentations, their framework generated a "Value of Evidence" metric—essentially a numerical strength-of-evidence rating that separates prior assumptions about outbreak likelihood from the actual signal in the data itself. This transparent, quantifiable approach offers substantial advantages for decision-makers: rather than relying on arbitrary thresholds or subjective judgement, clinicians and authorities can now assign verbal certainty statements (such as those familiar from forensic contexts) to surveillance alerts and produce risk maps that integrate case reports with modelling predictions and epidemiological risk assessments. For equine practitioners, particularly those involved in herd health or working with competition or breeding populations, this methodology could refine outbreak detection protocols and improve the timeliness and accuracy of early intervention decisions.
Read the full abstract on PubMed
Practical Takeaways
- •Syndromic surveillance data (respiratory and neurological cases) can be statistically evaluated using likelihood ratios to support objective outbreak detection decisions rather than relying on subjective interpretation
- •This framework helps distinguish between natural variation in disease reporting and evidence of a true outbreak, reducing false alarms and improving resource allocation
- •Practitioners can integrate multiple data sources (surveillance, models, risk assessments) transparently into disease outbreak response decisions
Key Findings
- •Bayes' rule framework using likelihood ratios can quantify the value of evidence from syndromic surveillance data for disease outbreak detection
- •The approach separates prior beliefs about outbreak probability from the strength of surveillance evidence, providing transparent reasoning for decision-makers
- •Value of evidence can be translated into verbal statements and risk maps, enabling integration with predictive modeling and risk assessments