Quantifying selection bias due to unobserved patients in pharmacoepidemiologic studies of severe COVID-19 cohorts. — SciRadar