2.2. Setting and Subject
In Japan, the first COVID-19 patient, who had previously visited Wuhan, China, was identified on January 15, 2020 [4
]. Many of the COVID-19 patients at the earliest stage in Japan had been exposed to the virus in China, or on a cruise ship (Diamond Princess) and were quarantined off the Yokohama port in Japan [9
]. However, an increase in the number of COVID-19 patients who had not visited China nor were present on the cruise ship was reported in Japan in the latter half of February 2020. The number of cases rapidly increased by March 2020 [4
The Ministry of Health, Labor and Welfare regulated subjects for the SARS-CoV-2 PCR test on February 17 in the eighth week of 2020. Such patients included those suspected of infection through an unknown exposure by a physician, a close contact of a COVID-19 patient, or one who had migrated from foreign countries with the epidemic [4
Although the Governor of Hokkaido declared a situation of emergency and requested people to stay at home on February 28 (the ninth week), 2020 as a response against the increased number of COVID-19 patients [11
], no such interventions, except school closure in the whole nation, were implemented until the 11th week in other prefectures of Japan. However, as the number of COVID-19 cases increased in the 12th week, the Governor of Osaka Prefecture requested the populace not to commute between Osaka Prefecture and Hyogo Prefecture (12th week). The Governor of Tokyo also requested physical distancing on March 25 (13th week), one day after the joint statement rescheduling the 2020 Olympiad in Tokyo [12
]. Because of these responses, situations regarding diagnostic delay of COVID-19 might have changed after the 12th week in some prefectures. Therefore, we selected the symptomatic COVID-19 incident cases with dates of onset of up to the 11th week (March 15) in 2020. Patients with symptom onset in the 11th week were reported in the 12th week at most.
Japan consists of 47 prefectures. In this study, we included those prefectures with >30 COVID-19 patients reported as of March 22, 2020 (the end of the 12th week). Therefore, eight prefectures, Hokkaido, Saitama, Chiba, Tokyo, Kanagawa, Aichi, Osaka, and Hyogo, were eligible. However, we could not retrieve the data on the route of exposure of patients in Osaka and Tokyo. Finally, 6 prefectures (population), Hokkaido (5.2 million), Saitama (7.3 million), Chiba (6.3 million), Kanagawa (9.2 million), Aichi (7.6 million), and Hyogo (5.5 million), were eligible.
Although in 2 out of the 6 prefectures, the number of cases with symptom onset at the end of the eighth week was less than 5, the number of cases with symptom onset at the end of the ninth week in all the 6 prefectures were not less than 5. Therefore, we selected symptomatic COVID-19 cases with a date of onset in and after the ninth week (beginning on February 24, 2020), in each prefecture. Finally, we included study subjects whose date of onset was between the 9th and the 11th week, in 2020. This study included COVID-19 patients in the 6 prefectures whose symptoms onset was dated between February 24 and March 15 (the 9th and 11th week), 2020, and SARS-CoV-2 positive patients as confirmed by a PCR test. We excluded cases who were asymptomatic and those whose symptom onset dates were missing.
2.5. Statistical Analysis
To confirm that the obtained data represented the situation in each prefecture, we calculated the proportion of symptomatic cases with unknown onset date to all the SARS-CoV-2 positive cases, reported between March 2 (the 10th week) and March 22 (the 12th week) in each prefecture; since the incubation period is approximately a week.
We calculated the mean and 95% confidence interval (95% CI) of diagnostic delay, without adjusting for right truncation, by 1000 times bootstrapping and by fitting parametric distributions to diagnostic delay in a Bayesian framework.
We described the characteristics of subjects with COVID-19. We used the multivariable logistic regression model analyses to calculate the proportion of patients with LDD and the adjusted odds ratios with 95% CI. We compared them across the week of symptom onset, route of case detection, and 6 prefectures. We implemented similar analyses using ≥5 days and ≥7 days instead of ≥6 days, as the definition of LDD.
Statistical analyses were performed using R (version 3.6–2; The R Foundation for Statistical Computing, Vienna, Austria).