Read on Twitter
In July @ayesharshahid and @OsmSiddiqi wrote a fantastic short piece puzzling over the skewed gender breakdown of reported COVID-19 cases in Pakistan (**74%** male!) and suggested that a skew in testing might be responsible /1 https://www.theigc.org/blog/dangers-of-not-knowing-enough-about-covid-19-imapct-on-women-in-pakistan/
hard to know as we don't have gender disaggregated testing data. also natl. & prov. govt. websites no longer have updated gender disaggregated case data. they do however have mortality statistics: men account for 71% of fatalities (source: https://www.sindhhealth.gov.pk/upload/daily_status_report/Daily_Situation_Report_for_30th_November_2020.pdf) /2
this certainly seems to be in line with established global trends of higher mortality risk from COVID-19 for men. but if testing is really skewed, then these fatality statistics are likely biased as well, and we are possibly underestimating mortality risk for women /3
i think there's strong suggestive evidence that what looked like a highly male caseload could be due to male skewed testing. it's this study reporting on results of serosurveys in Khi conducted in April & June 2020: @AKUGlobal @FyezahJehan https://www.medrxiv.org/content/10.1101/2020.07.28.20163451v3 /4
over half of the 2004 participants across 2 waves in 2 neighborhoods (Distt. East & Malir) are female (see the paper for a full understanding of the sample e.g. they have really high refusal rates which is interesting & concerning in and of itself) /5
using this, authors model age and gender-stratified seroprevalence estimates and conclude that "Seropositivity rates were indistinguishable between male and female within each district" /6
obv. am not equipped to understand the technical details of the paper, but the takeaway for me was that a random gender-balanced sample reveals a completely different picture of COVID-19 prevalence than testing (which ppl self-select into, potentially in highly gendered ways) 7/7
