In addition to not addressing confounding by indication, adjusting for the wrong indicators of severity, not accounting for clustering, a supplemental table that appears to have fitted data, the Australian death numbers don’t add up: bit.ly/2TKQmnA. @walidgellad
We also note that the North American hospitalized case numbers exceed publicly reported cases by a log when considering Canada is reporting only 4K, and we estimate 60-100k in the U.S. by 4/13. They quoted 60k from only 550 hospitals: bit.ly/2yJlBIw @dvgbiostat
The @WHO should not have stopped Soldiarity based on this study. bit.ly/2ZHGChP, well-done observational study which performed chart review and adjusted for disease severity via multiple lab/vital signs, did show such a large signal for mortality.
We should not be stopping RCTs as a result of the @TheLancet study. Balance of benefits and harms for #HCQ still unclear. I would not take it outside of a clinical trial, but I would not stop trials for the same reason n.pr/2XbdJZE
#qSOFA is not appropriate for disease severity. designed for bacterial sepsis not a #COVID. GCS and qSOFA scores not routinely collected at many hospitals SpO2<94 strange cut-off. bit.ly/36BbaTI Statement that imputation not needed very strange
Missing data very common in these types of analyses, we have strategies such as multiple imputation to address. I trust a carefully done chart review with a tenth of the n which acknowledges missing data much more than this #bigdata study that says it doesn’t exist @JuliaLMarcus
Tweet number 3 above should say well-done observational studies did not see such a large signal for mortality. My apologies
