Coronavirus Pandemic Secondary attack rates of COVID-19 in diverse contact settings, a meta- analysis

Introduction: The secondary attack rate (SAR) measures the transmissibility of an infectious agent. The reported SAR of COVID-19 varied in a broad range, and between different contact settings. Methodology: We conducted a meta-analysis on the SAR of COVID-19 with adherence to the PRISMA guideline. We searched published literatures and preprints in international databases of PubMed and medRxiv, and in five major Chinese databases as of 20 April 2020, using the following search terms: (‘COVID-19’ and ‘secondary attack rate’) or (“COVID-19” and “close contact”). The random effect model was chosen for pooled analyses, using R (version 3.6.3). Results: A total of 1,136 references were retrieved and 18 of them remained after screening. The pooled SAR of COVID-19 was 0.07 (95%: 0.03-0.12) in general. It differed significantly between contact settings, peaking in households (0.20, 95%: 0.15-0.28), followed by in social gatherings (0.06, 95%: 0.03-0.10). The point estimates of the pooled SARs in health facilities, transports, and work/study settings were all as low as 0.01. Among all the secondary cases, the proportion of asymptomatic infections was estimated to be 0.17 (95% CI: 0.09 – 0.34). The proportion was higher in households (0.26, 95% CI: 0.12-0.56), than in other contact settings. Conclusions: The transmission risk of SARS-CoV-2 is much higher in households than in other scenarios. Identification of asymptomatic secondary infections should be enhanced in households.


Introduction
COVID-19 emerged in late 2019 and rapidly caused a global pandemic [1]. Human to human transmission occurs mainly through respiratory droplets and fomite [2]. Cluster infections have been frequently reported [3][4][5]. However, the reported secondary attack rates (SARs) varied in a broad range, and between different contact settings. The SAR is defined as the proportion of secondary infections among susceptible persons within a reasonable incubation period following known contact with the primary case [6]. It is a key epidemiological parameter, indicating the transmissibility of a causative agent. It is necessary to know the SAR of COVID-19, and its variability between different contact settings, informing the adaptive implementation of public health measures. Asymptomatic COVID-19 cases can cause onward transmission [7], as silent spreaders. However, the proportion of the asymptomatic infection is still unclear, which limits our insight into their contribution to transmission.

Literature search, screening, and data extraction
We searched published literatures and preprints in English or in Chinese as of 20 April 2020, in two international databases of PubMed and medRxiv, and in five Chinese databases of Chinese National Knowledge Infrastructure (CNKI), WanFang databases, Chinese Journal of Epidemiology, Chinese Journal of Public Health and Chinese Journal of Preventive Medicine (Supplementary Table 1), using the following search terms: ("covid-19' and "secondary attack rate") or ("covid-19" and "close contact"). Literatures were firstly screened by title and abstract, and then assessed by the full text. Literature screening and data extraction were conducted by one investigator, and verified by another one. Any conflict was addressed through reassessment till a consensus was agreed. The literature quality was assessed (Supplementary Table 2).

Inclusion and exclusion criteria
Literatures that clearly indicated the number of the secondary infections and of the close contacts of origin were eligible for inclusion. Those using unverified media sourced data, or a subgroup data of another already-included study, were excluded.

Definitions
Confirmed cases: Laboratory confirmed COVID-19 cases, using PCR or equivalent nucleic acid amplification testing.
Close contacts: A close contact is defined as anyone with the following exposures to a COVID-19 case, from 2 days before to 14 days after the case's onset of illness (or the day the asymptomatic case was sampled): • Being within 1 metre of a COVID-19 case for >15 minutes; • Direct physical contact with a COVID-19 case; • Providing direct care for patients with COVID-19 disease without using proper personal protective equipment (PPE); • Other definitions, as indicated by local risk assessments. Asymptomatic cases: An asymptomatic case is a person with laboratory confirmed infection of COVID-19, who does not develop symptoms.

Statistical analyses
SAR was calculated as the proportion of secondary cases in close contacts of origin. The heterogeneity between studies was evaluated using the Higgins' I 2 test. Sensitivity analysis was conducted by omitting one study each time to evaluate its influence on the pooled estimate of SAR. Subgroup analyses were conducted by study regions and contact settings. Random effect metaregression was performed to identify significant moderators of SAR. Results are reported as summary point estimate and 95% confidence interval (CI). The Funnel and Egger's test were used to address publication bias. Statistical significance was set at p < 0.05 (two-tailed). All statistical analyses were performed using R (version 3.6.3 for Windows), including the "metafor" package [8].

Study characteristics
Most of the literatures (14/18) came from mainland China. No publication bias was indicated (Egger's test, p = 0.865) (Supplementary Figure 1). However, there was a significant heterogeneity between studies (I 2 = 99%, p < 0.0001). A total of 32,149 close contacts were documented. The reported SAR ranged from 0.00 (95% CI: 0.00 -0.02) to 0.80 (95% CI: 0.28 -0.99). Seven studies (all from mainland China) presented the number of asymptomatic secondary cases. For studies including multiple contact settings, the SAR and the proportion of asymptomatic secondary cases were extracted for each setting as well (Table 1).

SAR and proportion of asymptomatic secondary infections
We employed the random effects model for metaanalyses, considering the big heterogeneity between studies. The pooled SAR of COVID-19 was 0.07 (95%: 0.03-0.12) in general ( Figure 2). Sensitivity analyses showed that no single study had significant influence on the pooled estimate. The SAR differed significantly among contact settings. It peaked in households (0.20, 95% CI: 0.15-0.28), followed by in social gatherings (0.06, 95% CI: 0.03-0.10), and was low in healthcare facilities, transports and work/study settings (Figure 3  and 4). Meta-regression analyses indicated that household setting and social gathering setting were associated with significantly elevated SARs, and could jointly explain 51.0% of the total heterogeneity among studies. The pooled proportion of asymptomatic secondary cases was 0.17 (95% CI: 0.09 -0.34) in general and 0.26 (95%CI: 0.12 -0.56) in households ( Figure 5 and 6).

Discussion
Our pooled SAR of COVID-19 is similar with the SARs reported by later published studies [27,28]. Household SAR varied widely across literature in this study, which is also observed in another systematic review on SAR in household contacts [29]. The pooled SAR in household setting in this study is comparable     with that from a later published retrospective study from Guangzhou, China (17.1%, 95% CI: 13.3% -21.8%) [30], and those from other meta-analysis studies on SAR in households [31,32].
Household contacts were supposed to have a significantly elevated risk for being infected than other close contacts [18]. Our results indicate a significantly higher transmission risk in household setting than in social gathering, healthcare, public transport and work/study settings, which is echoed with recent studies from Singapore and China [28,33]. SARS-CoV-2 mainly transmits via respiratory droplets and fomite [2]. Close and prolonged contact could facilitate the viral transmission. Settings characterized with this kind of contact, such as households, bars and restaurants are expected to associate with high transmission risk. Less adherence to social distance, mask wearing, and hand hygiene, could also contribute to the higher SAR in households. In contrast, the comprehensive implementation of precautions could be attributed to the low SARs observed in healthcare facilities, transports and work/study settings.
The SAR of COVID-19 in households is higher than that of SARS (10.2%) [34] and pandemic influenza 2009 (13%) [35], which indicates a higher transmission capability of SARS-CoV-2. Infectiousness is suggested to peak on or before symptom onset of COVID-19 cases, while the peak is on 10 days and 1 day after symptom onset of SARS and influenza cases, respectively. It is estimated that 44% of secondary cases were infected during the index cases' pre-symptomatic stage [36]. This could jeopardize the effectiveness of symptom-based case isolation strategy badly.
Our pooled proportion of asymptomatic secondary cases is in line with the estimate from the Diamond Princess cruise ship [37], and the pooled estimate of another meta-analysis [38]. However, it is lower than the estimate (36%, 95% CI: 27% -45%) from a Singaporean study using a Bayesian model [33]. The serology testing used in that study contributed to finding more asymptomatic infections, compared with PCR testing used in other studies. The suggested high proportion of asymptomatic cases underscores the importance of placing targeted countermeasures to this population, although the transmission potential of them is lower than that of the symptomatic cases [38]. It is not clear why the proportion of asymptomatic secondary cases is higher in households compared with other contact settings. A possible explanation is that children usually get infected from family members in households at the early stage of the pandemic, and they are likely to present mild or no symptoms [39]. Active viral screening conducted in family contacts could also be responsible for finding more asymptomatic cases.
Given the high transmission risk within households, isolating mild COVID-19 patients and asymptomatic cases in designated isolation facilities is recommended rather than at homes. Aggressive move restriction could curb the inter-household transmission. Lift or ease of lock down measures should be cautious and step-wise, with essential social distancing measures maintained, to prevent the introduction of virus to unaffected households.

Conclusions
The SAR of COVID-19 in households is higher than that of SARS and pandemic influenza 2009. The transmission risk of SARS-CoV-2 is much higher in households than in other scenarios. Identification and management of asymptomatic secondary cases should be enhanced in households.

Supplementary
2) List inclusion and exclusion criteria for exposed and unexposed subjects (cases and controls) or refer to previous publications 5) Indicate if evaluators of subjective components of study were masked to other aspects of the status of the participants Describe any assessments undertaken for quality assurance purposes (e.g., test/retest of primary outcome measurements)