Evaluating the Ascertainment Bias of SARS-CoV-2 Using Spatial Backcalculation of Airport Screening Data in Japan

Despite global efforts to strengthen surveillance systems, ascertainment bias persisted. We evaluated the extent of this bias by comparing Japan's airport entry screening data with local surveillance data.

Entry screening data of non-Japanese passengers arriving from 25 countries were analyzed, and countries were categorized by income level: high-income countries (HICs), upper-middle-income countries (UMICs), and lower-middle-income countries (LMICs). A statistical model was developed to describe entry screening positivity using surveillance data of the country of origin.

Ascertainment bias was lowest in HICs (positivity = 1.76 times that of local surveillance data, 95% confidence interval [CI]: 1.42, 2.13). Bias estimates in the remaining income levels were substantially higher: positivity was 25.71 times (95% CI: 21.72, 30.09) and 24.38 times (95% CI: 21.29, 27.72) than that of surveillance data in UMICs and LMICs, respectively. The denial of new entry visas was associated with a relative risk reduction of 0.78 (95% CI: 0.72, 0.83) in UMICs. However, no significant effect was observed in LMICs (-0.13; 95% CI: -0.33, 0.03). Vaccination did not substantially alter post-arrival positivity.

A comparison of entry screening data with local surveillance data indicated that ascertainment bias depends on income level.