From case counts to probability sampling: Simulation insights into pandemic surveillance

To identify data strategies that ensure valid epidemic surveillance across different prevalence levels. A simulation study using realistic microdata from the German MikroSim project, reflecting the demographic structure of two districts. Epidemic dynamics were modeled over one year with an SIR framework, varying prevalence (>0% to 12%), test accuracy (80% to 98%), and sample size (5,000 to 30,000). Passive case-based surveillance relying on reported infections was compared with probability-based population sampling for estimating weekly prevalence levels and changes. Case-based surveillance was reliable only at very low prevalence. Once prevalence exceeded 3–5%, estimates became unstable and systematically biased, reflecting testing patterns rather than true infection dynamics. Probability sampling, in contrast, produced unbiased, precise estimates and enabled timely integration of individual-level social and health data. Surveillance systems should be adaptive. While passive reporting may suffice at low prevalence by practicability and costs, probability-based sampling becomes essential once moderate prevalence thresholds are crossed (3-5%). Such thresholds vary by disease and are shaped by symptom profile and transmission dynamics. Embedding predefined prevalence-based switch points that trigger representative sampling would ensure valid estimates, strengthen preparedness, and support timely, evidence-based public health decision-making