Household transmission of COVID‐19 among the earliest cases in Antananarivo, Madagascar

Background Households are among the highest risk for the transmission of SARS‐CoV‐2. In sub‐Saharan Africa, very few studies have described household transmission during the COVID‐19 pandemic. Our work aimed to describe the epidemiologic parameters and analyze the secondary attack rate (SAR) in Antananarivo, Madagascar, following the introduction of SARS‐CoV‐2 in the country in March 2020. Methods A prospective case‐ascertained study of all identified close contacts of laboratory‐confirmed COVID‐19 infections was conducted in Antananarivo from March to June 2020. Cases and household contacts were followed for 21 days. We estimated epidemic parameters of disease transmission by fitting parametric distributions based on infector‐infected paired data. We assessed factors influencing transmission risk by analyzing the SAR. Findings Overall, we included 96 index cases and 179 household contacts. Adjusted with the best‐fit normal distribution, the incubation period was 4.1 days (95% CI 0.7–7.5]). The serial interval was 6.0 days (95% CI [2.4–9.6]) after adjusting with the best‐fit Weibull distribution. On average, each index case infected 1.6 family members (95%CI [0.9–2.3]). The mean SAR among close contacts was 38.8% (95% CI [19.5–58.2]) with the best‐fit gamma distribution. Contacts older than 35 years old were more likely to be infected, and the highest SAR was found among them. Conclusion The results of our study provide key insights into the epidemiology of the first wave of SARS‐CoV‐2 in Madagascar. High rates of household transmission were found in Antananarivo, emphasizing the need for preventive measures to reduce community transmission.

Understanding of COVID-19 comes largely from disease surveillance and epidemiologic studies undertaken in China (2, 3) and highincome countries (4)(5)(6). However, confirmed cases of COVID-19 have also occurred in low-and middle-income countries (7,8). The first confirmed case of COVID-19 in Africa was reported in Egypt on February 14, followed by Algeria (9). By March, COVID-19 cases were reported across most of the continent. The first three confirmed COVID-19 cases were imported to Antananarivo, the capital city of Madagascar, on March 19-20, 2020. Following this notification and with the aim of stopping or slowing down the rate of transmission of SARS-CoV-2, the Malagasy government introduced stringent NPIs, such as physical distancing (school closures, work-from-home arrangements for civil servants, closing bars and restaurants, and suspension of public leisure). The country exceeded 17 000 confirmed cases and more than 240 total deaths in early November 2020 (10).
During the early phases, testing, contact tracing, quarantine, and isolation were carried out when cases were identified. Uninfected and asymptomatic contacts were often closely tracked, providing information about transmission and natural history of the disease.
Here, we analyzed data from the earliest cases detected in Antananarivo, Madagascar, and their intra domiciliary contacts to characterize epidemiological parameters of COVID-19 during the first wave that affected the capital city of Madagascar. Using data from contact tracing, we evaluated SARS-CoV-2 transmission by estimating the serial interval, the household secondary attack rate (SAR), and the average number of family members infected by each index case. Furthermore, we describe risk factors for transmission and infection.

| METHODS
We used an adaptation of generic protocols already in place in some countries, such as "The First Few Hundred (FF100)" enhanced case and contact protocol for pandemic influenza in the United Kingdom of Great Britain and Northern Ireland (11).

Case identification
On March 12-20, 2020, the Malagasy Ministry of Public Health tested for SARS-CoV-2 all travelers coming from Europe and China on international flights. Nasopharyngeal and oropharyngeal specimens (NP/OP) were sent to the virology unit at the Institut Pasteur de Madagascar, where they were tested for SARS-CoV-2 using real-time RT-PCR (RT-qPCR) as previously described (12).
The population included in this study was among the first confirmed cases of COVID-19. NP/OP and blood specimens were collected from laboratoryconfirmed cases and household contacts as soon as possible after laboratory confirmation. For all laboratory-confirmed index cases and household contacts, data were collected during the first visit and every 7 days until 21 days. NP/OP specimens were tested using RT-qPCR within 24 h following collection, while sera were tested retrospectively at the end of the study.

Epidemiological parameters
• The incubation period refers to the delay between exposure or contact with confirmed cases and symptom onset. We determined the left and right boundaries of the possible exposure and symptom onset times. Imported cases were assumed to have been exposed within 14 days prior to symptom onset. Cases without recent international travel history but with exposure to a confirmed case were assumed to be exposed from the time of earliest to latest possible contact with the case. Only cases for which it is possible to identify the earliest and latest time of exposure and who had a date of symptom onset were included in the estimation of the incubation period. The earliest exposure time was assumed to be within 15 days, and the latest exposure time was assumed to be within 30 days. Its distribution was calculated by fitting a parametric distribution (normal, gamma, lognormal, Weibull).
• Transmission was analyzed by examining the relationship between index cases and their close contacts.
• The serial interval is the average time expressed in days between the time of symptom onset of a primary case and that of a secondary case (13). Only pairs of symptomatic primary and secondary cases are included in the estimation of this indicator.
• The household SAR is calculated by dividing the household contacts who were later confirmed to have SARS-CoV-2 infection by the total number of household contacts included in the study.
• The distribution of the average number of family members infected by each index case was calculated from the number of secondary infections observed among close contacts of each index case.
All distributions of those parameters were also calculated by fitting parametric distributions (Normal, Lognormal, Gamma, Weibull) based on infector-infected paired data.
-SAR was estimated for a range of factors using univariate analysis and multivariate mixed effects logistic regression models with a random intercept for households. Households with one or more household members were included. The following potential explanatory variables were examined: characteristics of the index case, including age, gender, comorbidities, and whether the case was symptomatic; characteristics of the contact, including gender and age group; and household size. A stepwise backward selection variable (less than 0.20) was used in univariate analyses to choose the final model in the multivariate analysis.
All analyses were conducted with R software (14).

Ethical statement
Written informed consent was obtained from participants before enrolment in this study. It was approved by the Ethics Committee of

| Household transmission dynamics
The empirical SAR among close contacts was 29.6%, and after gamma distribution adjustment, the SAR was 38.8% (95% CI [19.5-  Our estimated SAR in Antananarivo was higher than those reported in Asia, mainly in Singapore (15) and China (16,17) and was similar to the household SAR found in the United Kingdom (18) and in some US states, such as Tennessee and Wisconsin (19). The heterogeneity in SAR across different regions might be explained by differences in control measures and crowdedness in households (17). This high SAR we reported in the current study reflects the existence of high transmission within households. In Madagascar, at the beginning   Our results suggested a higher transmission among household contacts aged 35 years and older compared to children. A systematic review and meta-analysis by Madewell ZJ et al. (22) reported that the SAR of SARS-CoV-2 to adults was higher than that to children, assuming that adults might be more susceptible to SARS-CoV-2 than children when they exposed themselves to the same sources of infection.

58.2]). Unadjusted SAR odds ratios and multivariate analysis of sec-
Data collected in Madagascar from March to September 2020 confirmed the same finding and suggested that individuals aged 50 years and older had a higher probability of having a positive RT-qPCR for SARS-CoV-2 (12).
We found an incubation period of 4.1 days (95% CI [0.7-7.5]), which is similar to those reported elsewhere (3,23). This estimate provides evidence to support a 14-day period of quarantine for infected and exposed persons. We reported a wider serial interval, which is similar to that found in China (3,24). Two hypotheses could explain these data: the memorization bias of the date of symptom onset and intrahousehold contamination from asymptomatic cases due to the lack of respect for protective/distancing measures. contacts which may lead to selection bias and biases in the estimation of the incubation period. As the inclusion in the study was limited to the early phase of the outbreak when the epidemic was rapidly growing, we might have missed other infected people with longer incubation periods.
Our results confirm that the household is an important venue for transmission and could explain the intensity and rapid spread of the virus during the second wave that started in early March 2021. To avoid transmission in the community, control measures such as appropriate isolation of cases and their household contacts should be adopted.