*Limitations of the Study*

There were several limitations in our study. One major limitation was that the original sampling design was aimed at evaluating the age-specific seroprevalence in Guangdong at a city-prefecture level. However, the overall low seropositivity precluded this and we instead derived the seroprevalence estimates according to region of epidemic activity. Another limitation was the sampling bias that occurred between urban centers and small cities. It was easier to sample in urban centers due to the larger number of medical institutions available. Sampling was also particularly difficult for persons 10–19 years old, which resulted in the smaller sampling size of younger age groups than that of other age groups in our study. One of the reasons might be that younger individuals were less likely to seek non-emergency medical attention, especially during the period when epidemic control measures were in place. This resulted in wider confidence interval for the estimate

of seroprevalence in this age group. Females were slightly oversampled compared to males (Figure S1), and therefore we also weighted by sex in addition to age to provide a more representative estimate. A meta-analysis of >3 millions COVID-19 global cases suggested there is no major difference in the risk of infection between sex [20], and therefore we expect that there would be minimal effect on the overall weighted estimate of seroprevalence due to the oversampling. In addition, our sampling period coincided with the gradual resumption of economic activity within the province as well as the easing of interprovincial travel restrictions. By April 8, the lockdown on Wuhan was lifted, signifying the last major travel restriction within China. As the cities with seropositive samples, including the three cities in the low-risks region, also had high numbers of travelers going to Wuhan in January, we are unable to determine if the positive samples were from a local resident or a returning migrant from Wuhan, or any other province. We also did not conduct IgM-testing, as we were interested in exposure history, for which IgG-antibody titers are more reliable [16]. Finally, as no data were available on asymptomatic infections prior to 1 April 2020 the number of infections in the early days that were reported will likely be underestimated.
