Among 47 prefectures in Japan, 10 prefectures (Aomori, Akita, Gunma, Tochigi, Toyama, Shiga, Okayama, Kochi, Saga, and Kagoshima) (Appendix Figure 1) that showed a relatively low COVID-19 prevalence posted laboratory-confirmed cases and contact-tracing results on their websites (Appendix Table 1). In this study, we collected basic characteristics of cases from the reports issued during February 22–May 31, 2020, on those websites. The websites did not provide characteristics of uninfected close contacts or details of residence of family members. During the study period in Japan, doctors provided diagnoses of COVID-19 by using real-time PCR and reported cases to healthcare centers. These centers listed close contacts according to whether they spent >15 min in face-to-face contact and conducted follow-up by telephone for >14 days to monitor their symptoms.

Persons who had any COVID-19–related signs/symptoms, such as fever, cough, and fatigue, were categorized as having symptomatic cases. Asymptomatic cases were those without any symptoms at diagnosis. During this study period, all confirmed case-patients were hospitalized after they were given a diagnosis. Suspected case-patients and asymptomatic close contacts self-quarantined at home. Healthcare centers in 8 prefectures performed PCRs for close contacts regardless of their symptoms. One prefecture did not show the strategy of PCR testing for asymptomatic contacts, and 1 prefecture performed PCRs for symptomatic contacts, such as persons who had fever and respiratory symptoms.

In this study, we defined a primary case as the first case to show development of symptoms and to be diagnosed in a family or the first diagnosed asymptomatic case in a family who had an apparent history of contact with a nonfamilial COVID-19 case-patient. We defined secondary cases as laboratory-confirmed cases from the list of close family contacts of primary case-patients. Because websites provided only symptoms at diagnosis, we could not identify presymptomatic cases. We calculated SAR as the proportion of secondary cases of family close contacts among the total number of family close contacts and determined the SAR, risk ratio, and 95% CI, stratified by the characteristics of the primary case-patients. We compared the SAR before and after the declaration of the state of emergency on April 16. All statistical analyses were conducted by using Stata version 14.0 (StataCorp, https://www.stata.com).

During February 22–May 31, 2020, the 10 prefectures reported 306 primary cases and 775 family close contacts from 306 families. Eighty-seven primary cases were associated with 147 family secondary cases (Table 1; Appendix Figure 2). The overall SAR was 19.0%. Among 28 asymptomatic primary cases, 7 caused family clusters (Table 2; Appendix Table 2), and the SAR was 11.8%. Eight prefectures that tested for asymptomatic contacts showed an SAR that was 1.77 times higher than the SAR for 2 prefectures that used a nontesting strategy. The age-stratified SAR was higher for persons 60–69 years of age (36.5%) and persons <20 years of age (23.8%) than for persons 20–29 years of age (13.3%), persons 30–39 years of age (20.4%), persons 40–49 years of age (10.1%), and persons 50–59 years of age (16.1%) (Table 2).

With increasing time from symptom onset to diagnosis, the SARs in households increased from 11.6% (>2 days) to 40.0% (>14 days) (Table 2). When the data were stratified for analysis by the number of household contacts, 4 household contacts showed the highest SAR (25.7%). After a quarantine at home was requested from the government on April 16, the SAR increased from 17.4% to 21.0%, but the risk ratio did not reach statistical significance.

This family cluster analysis in the 10 prefectures of Japan showed that the overall SAR of the family cluster was estimated to be 19.0% in Japan. Meta-analysis from 43 household transmission studies estimated a SAR of 18.1% (3): 3.9% in Singapore (7), 4.6% in Taiwan (8), 10.3%–54.9% in China (9–12), and ≈30% in the United States (13) and Norway (14). In addition, the SAR of asymptomatic primary cases was 11.8% in our study, which was higher than the 0%–4.4% reported in a limited number of previous studies (6,15). The SAR heterogeneity might have been dependent on the surveillance protocol for asymptomatic contacts. The studies in the United States (13) and Norway (14), which had high SARs, detected secondary cases by using serologic tests. Our study also indicated that 8 prefectures that tested for asymptomatic contacts showed a 1.8 times higher SAR than did 2 prefectures that tested only for symptomatic contacts. A low proportion of diagnoses of asymptomatic cases might underestimate the SAR.

We showed that SAR was higher for persons <1–19 years of age and >60 years of age than for other age groups. High infectivity for the younger age group (6) and the older age group (4) was reported from South Korea and China, as in our study, but most other studies did not show significant differences in SAR by age of primary case-patients (9,13). Age-dependent infectivity might be associated with household lifestyles, family structure, and clinical conditions (9). Meta-analysis showed that the sex of the primary case-patient was not associated with transmission (5).

If primary cases were detected <2 days of symptom onset, the SAR was lower than that for primary cases detected >2 days after symptom onset. This finding was related to the low SAR for case-patients who had a contact history because they could receive PCRs, as close contacts did earlier, and might have had a short time of exposure to family members. Our results were concordant with previous studies showing an increased risk for transmission as the contact duration was prolonged (4), as well as the effect of quarantining index case-patients when symptoms were reported (10).

The first limitation of our study is that symptomatic cases diagnosed during the presymptomatic period might have been classified as asymptomatic cases. Second, the number of asymptomatic cases might have been underreported because of different testing protocols among prefectures. Third, we might have misclassified the primary cases if a coprimary case existed or the direction of transmission between asymptomatic cases and symptomatic cases was not clear.

In summary, our study results provide us with useful implications of the high SAR of asymptomatic primary case-patients and contacts with long exposure times to primary case-patients. Self-quarantine and rapid isolation of confirmed case-patients from households after symptom onset might be needed to reduce transmission in families.