Other scholars think the overall pattern of greater male death is more meaningful. “You still see this one phenomenon make it through all that variability,” says Stefanick. “It’s passing my consistency test.” Of course, a consistent state-to-state pattern could still be primarily caused by social factors, since the states are culturally similar in many ways.
But public health data collection is a decentralized process run by individual state health departments, so comparing states can be difficult. “States follow radically different systems of how data is collected and reported,” Shattuck-Heidorn says. For Schubert, these inconsistencies make it impossible to say whether the variation among states is meaningful. “At the end of the day, if we’re comparing apples to oranges, it’s not going to make a difference,” she says. “It has to all be done in a collective, standardized way.”
And the data isn’t just incomparable from state to state: It is also insufficiently detailed. Researchers like Richardson and Shattuck-Heidorn desperately want data that not only tracks deaths by sex, age, comorbidities, and other factors separately, but also records how those variables overlap. To test whether heart disease might explain the sex difference in Covid-19 deaths, a scientist would want data about how many men and women who died of Covid-19 also had heart disease. Just knowing the numbers of men and women who died of Covid-19 and, separately, how many people with and without heart disease died of Covid-19 wouldn’t be enough to link the sex difference in deaths to heart disease.
But while some states do track comorbidities, they generally don’t divide up that data by sex. Similarly, it is extremely difficult to evaluate whether race, occupation, and age play into the sex differences in Covid-19 case and death rates because most states also don’t record how those attributes interact with sex. “Our current frustration is that we cannot see sex-disaggregated data by age, by race/ethnicity, by comorbidity, and by occupation,” Richardson says. “We want that so much. And we cannot see it.”
Sharma has similar concerns. “I truly want to see—once you control for hypertension, once you control for chronic lung disease, once you control for diabetes, once you control for obesity, once you control for cardiovascular disease or myocardial infarction, once you control for heart failure, once you truly control for smoking and all other high-risk behaviors—if you control for all of that, is there still a sex-specific or a sex-dependent case-fatality rate?” she wonders.
Even the labels of “male” and “female” conceal a great deal of complexity. A 2016 report estimated that 0.6 percent of the US population identifies as transgender, and arguments about hormones and gendered behavior that apply to cisgender people may not be as relevant to that group. But no state has offered more detailed information about the gender identities of their Covid-19 patients than a “male/female” binary label. “We are talking about people ascribed male or female, attributed male or female, by public health agencies,” Richardson says.
Interestingly, “about 5 percent of individuals are categorized as unknown” in some state data sets, according to Richardson. “We don’t know how to interpret that,” she says. Perhaps these individuals are nonbinary—or perhaps that data is simply missing, too.
The data is getting better, albeit slowly; New Jersery just started reporting cases and deaths by sex last week, and Georgia now provides an anonymized spreadsheet that includes the age, sex, race, and county of each Covid-19 death, as well as whether the person had a preexisting chronic condition. This week, the GenderSci Lab will release a Covid-19 “data report card” that will rate the quality of the data coming from each state and, with any luck, convince lagging states to improve their data.