The Twitter account "CovidAnalysis" is promoting COVID-19 misinformation /conspiracy theory on HCQ. This includes promoting suspicious websites framed as if they were scientific articles.

@TwitterSafety @vijaya #MedTwitter

This is not an actual study and makes blatantly false claims: that it is a randomized control trial and that it used a sample size of two billion people.

@TwitterSafety @vijaya

Using existing data from public sources is not a form of assignment. It is also not random. "Entire countries were assigned" is a blatantly false statement intended to copy the style or lingo of actual research articles.

This is the same Twitter account that made the c19study website, a collection of research articles they claim support HCQ usage for COVD-19. I fact checked some of these elsewhere and found many did not say what the website claimed they said.

#MedTwitter

Spreading COVID-19 misinformation is a violation of Twitter policy. Regardless of the HCQ debate, everyone should agree that fabricating a study on an anon website counts as spreading COVID-19 misinformation. Highly malicious. Please report.

@TwitterSafety @vijaya

#MedTwitter

On a lighter note, having your Acknowledgments full of anonymous Twitter accounts is worth a chuckle.

Says they excluded countries that adopted early mask use, per Leffler et al. Yet here we see they included those countries: Greece, Morocco, Cuba, Israel + more. Clear cherry picking. Mask use likely explains the relationships.

#covid #disinformation

Here they misreport the results of one study, claiming HCQ is effective, when the authors said it was not. Mitja et al.

Poor statistical choice? (Deliberate to show an effect?) The average of individual country rates can skew toward outliers more easily, overestimating. Did they correct for this? We don't know since there is no real methodology section or explanation of what procedures they used.

They handwave away a limitation, but this is a serious problem. We can't know from this (lack of) data about how often HCQ is actually used within the populations of either group. HCQ govt policy may not line up with HCQ use or prescription in a population.

That they include 36 countries and conveniently exclude every other is a red flag. Is it really appropriate to exclude countries with small populations, for example, or is this done because inclusion would not give them the desired association?

Despite claiming this is a billion-subject sample (it is not, it's an association of country policy w/ death rate), using averaged death rates for 36 countries gives you 36 data points. This is your true sample size: 36. You now have a small to low-medium sized sample.

To reiterate: govt HCQ policy does not mean the population is using/ abstaining from HCQ. Yet they present it as such: entire populations as experimental or control groups. They are not. We don't know from this if govt policy is even associated with prevalence of HCQ use.

How did they define "early" and "late" treatment? This is quantifiable. Did they use a standardized measure (eg. date)? No. They said see Appendix 12. It's other Twitter accounts. No quantification of "early," "late." Can't know if "early" countries should be "late." Subjective.

Said they removed mask use as a confound (they didn't, see above). But not quarantine as a confound? Lockdown orders associated with 40-60% reduction in COVID spread/mortality.

Le et al. http://dx.doi.org/10.2471/BLT.20.262659

FOX News duped by this website. These two MDs also. Hard to know if they're spreading it while knowing it's fake or if they actually believe it's real.