Prediction of severe COVID-19 cases requiring a ventilator in Tokyo, Japan

BackgroundTo avoid exhaustion of medical resources by COVID-19, policy-makers must predict care needs, specifically the proportion of severe cases likely to need ventilator care. ObjectiveThis study was designed to use a statistical model to elucidate dynamics of severe cases in Tokyo and to discuss the timing of effective policy activation. MethodsThe study extended from April 27 through October 18, 2020 in Japans Tokyo Metropolitan area. Medical exhaustion was defined as use of more than half of the ventilator capacity available before the COVID-19 outbreak. We regressed the number of severe cases on the newly onset patients of more than 14 days prior. As earlier research indicated, the COVID-19 severity changed at the end of May. Therefore, we added dummy variables to reflect changing severity from June and its product with newly onset patients as the explanatory variable. Then we calculated the threshold using R(t): R(t)=0.99 for the number of patients 14 days prior was used as a threshold at which strong countermeasures should be activated to keep to avoid medical exhaustion. ResultsThe critical number signaling medical exhaustion in Tokyo was defined as 655 cases. We selected 15, 30, 60 and 90 days prior as explanatory variables for explaining the number of severe cases. A coefficient of determination larger than 0.95 was inferred as indicating good fit. The threshold was estimated as more than 4500 cases for R(t)=1.1 and monotonically decreasing by R(t) to be 600 cases for R(t)=2.5. Discussion and ConclusionResults showed that newly onset patients reported more than 14 days prior can explain the number of severe cases very well. We also confirmed the threshold number of patients 14 days prior by R(t) for which strong countermeasures should be activated to avoid medical exhaustion with R(t)=0.99. This method is expected to be useful for countermeasure activation policies for Tokyo.