Cross‑country pooling in epidemic models can make different policies look uniformly effective by averaging over local timing, subnational heterogeneity and reporting biases. When models assume immediate step changes in transmission from named interventions and borrow strength across countries, they risk overattributing causality to the recorded policy while smoothing divergent behavioural or importation effects.
— Policymakers and the public need to treat pooled, cross‑country NPI estimates as conditional on modeling choices because those choices can change which interventions are credited for control and thus guide future policy incorrectly.
2020.06.08
100% relevant
Flaxman et al. (Nature 2020) explicitly used partial pooling across 11 European countries and assumed immediate Rt responses to interventions—choices the paper notes as limitations and that later methodological critiques focused on.
← Back to All Ideas