Uncontrolled confounding in COVID-19 epidemiology