乌鸦传媒

If small in-person social gatherings tend to promote COVID-19 transmission, then wouldn't an identifiable and predictable reason to hold them -- like birthdays -- also be tied to increased risk?

That was the hypothesis for a study published in , and that is indeed what the authors found.

With insurance data on 2.9 million households including members' birth dates, a team at RAND Corporation in Santa Monica, California, and Harvard Medical School in Boston found a 31% bump in rates of COVID-19 diagnoses in households having a birthday in the previous 2 weeks in counties with high rates of transmission. Child birthdays seemed to increase the risk far more strongly than adult birthdays.

But the authors, led by RAND's Christopher Whaley, PhD, offered no suggestion as to how this knowledge could inform public health policies, other than as indirect support for social gatherings' already recognized role in perpetuating COVID transmission.

Instead, Whaley and colleagues conceived the study as a way to eliminate "the problem of confounding in who does and does not socially gather" in research intended to quantify the benefits of policies seeking to limit gatherings. Tracking birthdays and subsequent COVID diagnoses should do this, they argued, because birthdays should carry no intrinsic infection risk but rather serve only as a reason to assemble in close quarters.

The 2.9 million households included in the study, with some 6.5 million individual members, had private insurance from about 200 U.S. employers. Summary data indicated nothing very unusual in their demographics, although race/ethnicity was not included, and households in the Northeast were underrepresented (just over 10% of the sample), while the South was overrepresented (about 41%).

As expected, households were more likely to have a member develop COVID as the county-level prevalence increased, but this risk was magnified in the 2 weeks following a member's birthday. For example, in the first and second decile of county prevalence, households with birthdays had no greater risk than did households with no birthday; but in 10th-decile counties -- where rates averaged 27.8 per 10,000 -- the rate was higher by 8.6 per 10,000 (95% CI 6.6-10.7) in households with versus without birthdays.

Child birthdays were particularly risky in these high-prevalence counties: COVID diagnoses occurred at 15.8 per 10,000 (95% CI 11.7-19.9) in the 2 weeks after a child birthday versus 5.8 per 10,000 (95% CI 3.7-7.9) following an adult birthday.

Whaley and colleagues also sought to determine if other factors might affect these apparent birthday-associated risks: counties' political leanings (as determined by votes for Donald Trump or Hillary Clinton in 2016), stay-at-home policies, and bad weather. They also looked at "milestone" birthdays to see if they had a stronger effect. Surprisingly, perhaps, no independent associations turned up for any of these factors that reached statistical significance (there were weak trends, however, toward higher risk in high-prevalence counties for milestone birthdays and Trump having won in 2016).

But the study came with a host of limitations. Of course, the researchers had no data on actual gatherings that may have accompanied the birthdays. Moreover, they noted, asymptomatic COVID cases were not tracked, and there remained a number of sources of potential confounding, such as the possibility that households having birthday parties were more likely to seek testing afterward.

In addition, households with public insurance -- for whom COVID risk is known to be especially high -- were not part of the study, although that may lead to an underestimate of the risk associated with birthdays, Whaley and colleagues indicated. Another potential source of underestimation is the possibility that people in high-prevalence locations were aware of the risk from birthday parties and refrained from having them.

Comment