Epidemiology and contact tracing assessment of COVID-19 and potential risk of transmission at different exposure settings: A prospective cohort study

Background: Transmission risk of coronavirus disease 2019 (COVID-19) to close contacts and at different exposure settings are yet to be fully understood for the evaluation of effective control measures. Methods: We traced 1171 close contact cases who were linked to 291 index cases between July 3, 2020 and September 3, 2020. Clinical and epidemiological characteristics of all index cases, close contacts, and secondary contact cases were collected and analyzed the secondary attack rate and risk of transmission at different exposure settings. Results: Median age of 291 index cases were 43.0 years (range 18.5-82.3) including 213 male and 78 females. Among all 1171 close contact cases, 39(3.3%) cases were identied as secondary infected cases. Among 39 secondary cases, 33(84.62%) cases were symptomatic and 3 (7.69%) cases were asymptomatic. Of the 33 symptomatic cases, 31(86.1%) male and 5(13.9%) female. Of these 36 symptomatic cases, 24(66.7%) cases between age 20-59 and remaining 12(33.3%) cases were age 60 and over. Of the 36 symptomatic cases, 11(30.6%) cases were identied as severe, 19(52.8%) as moderate and 6(16.7%) as mild. The overall secondary clinical attack rate was 3.07% (95% CI 2.49-3.64). The attack rate was higher among those aged between 50 to 69 years and shows higher risk of transmission than age below 50 years. The attack rate was higher among household contact (6.17% (95%CI 4.7-7.6; risk ratio 2.44[95%CI1.5-3.4]), and lower in hospital facility (2.29%,95%CI0.58-3.40; [risk ratio 0.91,95%CI 0.17-1.9]), funeral ceremony (2.53%,95%CI 0.32-4.73), work places (3.95%,95% CI2.5-5.42 [risk ratio 1.56,95%CI 0.63-2.5]), contacts

The attack rate was higher among those aged between 50 to 69 years and shows higher risk of transmission than age below 50 years. The attack rate was higher among household contact (6.17% Contact tracing is a useful and cost-effective tool for assessing and controlling the transmission of many infectious diseases such as Ebola, SARS, MERS and tuberculosis. If the pre-symptomatic cases from a community are well traced and monitored it becomes easier to manage the transmission as opposed to costly entire lockdowns. Therefore, a well-established tracing and monitoring program increases the probability of rapid identi cation and isolation of cases and quarantine to restrict further secondary transmission of the virus. Secondary transmission through social interaction and other type of close contacts may change the overall transmission dynamics. Several recent literature highlighted that many secondary cases through different exposure settings play a signi cant role in the outbreak of COVID-19 [4][5][6] . The reports on epidemiological characteristics, transmission dynamics and the risk factors of COVID-19 in various exposure settings is rapidly growing. Tracing and identifying the close contacts, analyzing secondary transmission and potential risk of transmission in different exposure settings is essential to control transmission of the virus and to develop targeted control and prevention. On March 8, 2020, three COVID-19 cases were con rmed in Bangladesh, where two individuals were repatriated from Italy, and a third individual was a female relative of one of the two aged between 20 and 35 [IEDCR, 2020]. These individuals had no visible signi cant signs of illness and passed the airport surveillance. During this same time period, other returnees arrived from the USA, several middle east countries and India in March 2020. As of April 13, 2021 (at the time of the manuscript submission), there were 698,000 laboratory con rmed COVID-19 cases and 9,891 deaths were recorded [IEDCR, 2021]. Using contact tracing data, we examined the epidemiological characteristics and secondary attack rate and risk factors at different exposure settings in a cohort study of 1171 close contacts of 291 index cases of COVID-19 from Bangladesh.

Characteristics of index cases
Between July 3, 2020 and September 3, 2020, 291 laboratory-con rmed SARS-CoV-2 infected cases were identi ed in a densely populated district Chattogram, Bangladesh. Table 1 shows the demographic characteristics of all index cases. Among all cases, 103 (35.4%) were detected through contact-based surveillance while remaining 188 (64.6%) were found after symptom-based surveillance (Table 1). Among all reported cases, 213 (73.2%) were male and 78 (26.8%) were female. The mean age of all cases was 43.0 years (ranges from 19 to 83), and 29 (9.97%) cases were in the age group between 0-29, 224 (76.98%) cases were in the age group between 30-69 and remaining 38 (13.06%) cases were in the age group 70 and above. When patients were classi ed based on the disease severity, 145 (49.83%) cases were mild, 87 (29.9%) cases were moderate and 59 (20.3%) cases were severe/critical in nature. Of 291 cases, 258 (88.7%) were symptomatic, while 33 (11.34%) did not show any symptom (Table 1). Going by the growth of the epidemic and based on patients with well-de ned time of exposure and time of symptom onset, we estimated the onset to symptom period, median days in hospital stay, median onset to recovery and median onset to death for all index cases. In both surveillance modes, the majority of cases showed symptoms of the disease within 2 to 5 days, with the highest number of cases showing symptoms at day 4 ( Fig. 1A). The estimated incubation period for COVID-19 cases was found to be 3.9 days and 4.  Table S2).

Secondary infection rate and risk of transmission of close contacts in major exposure settings
We collected 1171 close contact cases who were linked to 291 index cases between the same time period. Of the 1171 close contacts, 79 (6.75%) were identi ed as funeral ceremonies contacts, 205 (17.1%) household contacts, 131 (11.19%) hospital facilities contacts, 177 (15.12%) workplaces contacts, 181 (15.56%) were family close contacts and remaining 398 (33.99%) cases belongs to others category (social gathering, food court, market places, public transportations). Table 2 summarizes detailed epidemiological characteristics of all close contact cases in different exposure settings. We identi ed six (6) clusters and the distribution of clusters in each exposure according to the number of index cases is presented and summarized in Fig. 2A (Supplementary Fig. S2; Table 3). Through contact tracing, we identi ed 39 secondary cases from 1171 close contacts in different exposure settings (Fig. 2B). Six clusters of exposure settings with one or more secondary cases (not including tertiary cases or subsequent cases) were identi ed ( Table 3). All cases in these clusters were locally transmitted. Three asymptomatic of 39 cases were excluded from the subsequent transmission analysis because none of the three asymptomatic cases transmitted to secondary cases. Of the 33 secondary cases, 11 (33.33%) were in household setting, 3 (9.09%) in the hospital, 2 (6.08%) in funeral, 7 (21.21%) in work places, 7 (21.21%) in family and 6 (18.8%) in others (public transportations, market places, food court etc.) exposure settings ( Fig. 2B; Table 4). We closely monitored these secondary contacts when they were under quarantine or isolated at least for two weeks and during this time all 36 out of 39 of the close contacts showed all the major symptoms of COVID-19 and tested positive for COVID-19 infection. For symptomatic secondary cases with con rmed COVID-19 diagnosis, the distribution of days of symptom to RT-PCR diagnosis, symptom onset and isolation start time is presented in Supplementary Figure S3.    (Table 4).

Discussion
This is the rst comprehensive contact tracing and secondary transmission dynamics analysis of COVID-19 in different exposure settings from Bangladesh. Our analysis of con rmed COVID-19 cases among close contacts revealed an insight into the demographic characteristics, sign of symptoms, infection periods and secondary transmission risk at different exposure settings. We found that the overall secondary attack rate was lower than 4% (actual attack rate was 3.07%). However, the secondary attack rate was higher among household contact followed by workplace and family close contacts. Moreover, the secondary attack rate was higher in higher age groups and more severe cases and likely to contribute in increasing the risk of infecting other close contacts.
The incubation period of COVID-19 is critical in deciding the types of measures to be considered for the safe and exact isolation period who are likely to be exposed. Most recent studies reported an incubation period that is fairly shorter (3.0 to 6.4 days), but with a longer detrimental clinical course [7][8][9][10][11][12][13] . The time to recovery or death from this disease is comparatively longer and may take several days to weeks 13 . Our data from index cases estimated the incubation period notably shorter (3.9 days) and whereas 2.5 days for secondary contact cases which is shorter than most reported COVID-19 cases 4,12 . Cheng et al 5 reported the median incubation period was 4.1 days for close contact cases from Taiwan reinforcing our assumption that close contact cases may have a shorter incubation period than primary cases. On the other hand, our estimate of the serial interval was shorter than the recently published report (3.3 [our case] vs 4.0 vs 7.5 days) 5,10,14 . In our cases, the estimated distribution of incubation and serial interval distribution for household contact cases is shorter than other previously reported settings. This differences in the incubation period and serial interval in household settings perhaps due to repeated and longer time interaction/exposure with the symptomatic cases. As we did not analyze the incubation periods between sex and different age groups, we suggest that determining the sex and age speci c incubation periods would provide additional intervention measures to conduct the epidemiological study.
Analyzing secondary transmission dynamics may provide valuable insights for characterizing the epidemiological modeling parameters, risk factors and disease transmission estimation 1,6 . Most studies assessed the transmissibility of COVID-19 primarily using a mathematical model at the population level [15][16][17] . A handful of studies investigated the other type of transmissibility such as household or through other types of close contacts 13,18 . Data obtained from contact tracing provides the most accurate and up to date information about person-to-person transmission possibility by obtaining and accounting individual-level exposure history. In Bangladesh, the COVID-19 transmission occurred most likely amongst close contacts, such as in the household setting, hospital facility, family members, funeral ceremonies, workplaces and public transportations. To ascertain these possibilities, we assessed and estimated the secondary attack rate of COVID-19 in Bangladesh among several exposure settings. The secondary attack rate of COVID-19 among household contacts was 6.07% in our study which is higher than in Taiwan 5 but lower than China [11.2%] 4 and 10.5% in the United States 19 and Guangzhou, China 20 .
Although, the overall secondary clinical attack rate was relatively low, adult cases and particularly the age group between 50 and 70 are at higher risk of being infected in general. Surprisingly, we observed that other exposure settings such as hospital facilities, funeral ceremonies and family contacts and public transportations had lesser risk of secondary transmission. Our results showed that patients with severe COVID-19 cases are more likely to have higher transmission possibilities due to higher viral loads than those of moderate and mild type cases supporting previously reported results 1,13,21 . The high transmission of COVID-19 amongst household contacts during the study period was most likely due to the mass movement of people during the government lockdown announcements as well too many household members living and sharing in under the same household. Other reasons could be people did not follow the guidance of proper personal protective measures and not following the government's strict containment measures in the country.
Our present work has several notable limitations. First, reliance on estimates from discrete but a small cohort of cases, and on cases with incomplete information of clinical symptoms, age or both. However, given the di culties of tracing of cases and the urgent timeline, we had to rely on the small cohort to provide real-time information on the COVID-19 pandemic situation in Bangladesh. In several cases we did not reach the patients to obtain additional missing information. Second, in some cases, we were unable to thoroughly examine the contacts before onset of symptoms in the contact-based surveillance and symptom-based surveillance cases, therefore limiting the analysis of early transmission. This limitation might contribute to underestimation of new infections from early transmission. Third, we were unable to obtain actual data to quantify the infectivity of the asymptomatic infection cases and silent transmission from asymptomatic cases from different exposure settings.
In summary, our results showed that overall secondary attack rate of COVID-19 in Bangladesh was comparatively low, but household contacts as well as severe cases are at higher risk of both infection and transmission and need more coordinated attention to stop the spread of virus. At the time of the preparation of this report the country is still experiencing the growth and transmission of the pandemic. As the cases continue to grow, it is fairly important that we need to obtain more knowledge about the pattern of transmission in different exposure settings.

Methods
Cases from individual labels and study population On April 3, 2020, BITID (Bangladesh Institute of Tropical and Infectious Diseases) identi ed the rst COVID-19 case in Chattogram, Bangladesh. The government since then expanded the testing and surveillance of all close contacts regardless of symptoms in the neighborhood. All suspected cases and close contacts were tested for COVID-19 and followed up until 14 days after last exposure to the index cases. All close contacts were identi ed through contact tracing who lived in the same house or apartment or in the same building, and who attended any funeral ceremony or socially interacted before any COVID-19 like symptom onset. Detailed demographic and clinical data and exposure history were recorded and reported for all con rmed cases. This work was conducted as part of the ongoing public health response, therefore institutional or regional review board approval and individual cases consent approval was waived. This study was approved by the central ethics committee of Bangladesh Medical Research Council (study # 2021-2023/62(1-20) -a waiver of informed consent from patients was also approved. All research was performed in accordance with the relevant ethical guidelines and regulations.

Epidemiological And Clinical Features Of All Contact Tracing Cases
All RT-PCR testing was done by IEDCR [IEDCR 2020] and its approved a liated eld centers across the country. The symptom-based surveillance was de ned as cases who were screened in the government certi ed test centers (not in the airports or any borders entry point) such as testing of patients admitted to hospitals, and other test centers. We de ned contact-based surveillance as cases who were screened through monitoring and testing of close contacts of any con rmed cases irrespective of the showing any symptoms. In this analysis, close contact was de ned as individuals who were closely linked by contact tracing and were considered a close contact group provided that no personal protective equipment (PPE) was worn having direct face to face contacts. Household contacts were de ned as individuals who lived and were sharing the same room and same apartment in the same household with the index cases. Family contacts were those who are the members of the same family not living in the same household.
Once a close contact is identi ed, a thorough clinical assessment was conducted for common symptoms as presented in Table 1, and at the same time patients' disease severity was assessed. We have classi ed disease severity in 3 main categories. (a) mild cases if the patients do not show any serious symptoms described in moderate or severe categories; (b) moderate cases if the patients show fever, some respiratory symptoms and any evidence of pneumonia by radiography; and (c) severe cases de ned as cases experiencing any of the following symptoms: breathing rate 30 or higher breaths/minute, the oxygen saturation level below 93%, one or multiple organ failures, requiring mechanical ventilation, a ratio of the partial pressure of arterial oxygen to the fraction of inspired oxygen (PaO2:FiO2) of less than 300 mm Hg, or in ltrates in more than 50% of the lung eld within 24 to 48 hours.
Statistical analysis, data processing and analysis of timing of key events Determining of serial interval and incubation period (symptom onset).
All summary analyses were done using R statistical software (version 4.0.2). We used R-packages "ggplot2" for generating graphs. To compare the characteristics between demographic groups and patients' underlying disorders we used non-parametric (Fisher's Exact test). Distributions were tted to the key events in all con rmed cases. For our primary analysis, we assumed that the incubation time (infection to symptom onset) follows the log-normal distribution as observed in other acute respiratory viral infection 5,22 . We estimated the prior and posterior boundaries on the symptom onset time. Time between symptom onset, days in hospital, onset to recovery and onset to deaths were estimated by tting gamma distribution using Bayesian methods. All con dence intervals were analyzed with bootstrapping or standard parametric estimators described previously 13,23 . Estimation of Case Fatality Ratio (CFR) was determined using recently reports by Verity et al 2020 24 and Mizumoto and Chowell 2020 25 which follows a gamma distribution. Following the WHO guidelines, we de ned the secondary clinical attack rate as the ratio of symptomatic con rmed cases among the close contacts 18,26 . Incubation period and serial interval were estimated using the contact tracing data in Bangladesh. We used similar methods as mentioned above to increase the stability of the small estimation for contact tracing cases. We collected symptom onset dates of COVID-19 cases with con rmed transmission links in IEDCR and from local testing centers or directly contacting con rmed COVID-19 cases. The exposure settings were categorized as (