SARS-Cov-2 Seroprevalence in Chattogram, Bangladesh Before a National Lockdown, March-April 2021 Supplemental Material
Total Page:16
File Type:pdf, Size:1020Kb
SARS-CoV-2 seroprevalence in Chattogram, Bangladesh before a National Lockdown, March-April 2021 Supplemental Material Supplemental Methods .................................................................................................................2 Study population .............................................................................................................................2 Supplemental Figure 1. Map of the study population in the Sitakunda Upazila (in green) in the Chattogram District of Bangladesh. The 169 enrolled households sampled in the serosurvey and the two healthcare facilities in Sitakunda (BITID and the Sitakunda UHC) are indicated on the right side of the figure .....................................................................................................................2 Survey methods ...............................................................................................................................3 Analysis ...........................................................................................................................................3 Supplemental Results ....................................................................................................................5 Study participation rates ................................................................................................................5 Test for trend ..................................................................................................................................5 Supplemental Table 1. Descriptive statistics for serosurvey participants (n=664) in Sitakunda Upazila by sociodemographic factors, measures of urbanicity, COVID-like symptoms, COVID testing and vaccination, and COVID-related behaviors .................................................................5 Supplemental Table 2. Descriptive statistics for unvaccinated, serosurvey participants (n=639) in Sitakunda Upazila by sociodemographic factors, measures of urbanicity, COVID-like symptoms, COVID testing and vaccination, and COVID-related behaviors ..................................................10 Supplemental Figure 2. Average seropositivity among all participants in each sampled cluster (1 km2) by population density and the minutes required to travel one minute .................................15 Supplemental Table 4. Estimated seroprevalence of SARS-CoV-2 in Sitakunda Upazila adjusted for sex, age (<15 vs 15+), household clustering, and test performance among all individuals, both vaccinated and unvaccinated .........................................................................................................17 References ....................................................................................................................................18 Supplemental Methods Study population The study population was the population of the Sitakunda Upazila (subdistrict) within the Chattogram district in Southeastern Bangladesh. This study was originally designed to estimate seroincidence of Vibrio cholerae infection through serial serosurveys. Participant enrollment was focused in a population that is known to primarily seek care for acute watery diarrhea at the Bangladesh Institute for Tropical and Infectious Diseases (BITID) and the Sitakunda Upazila Health Complex (UHC), two healthcare facilities in the Sitakunda subdistrict. This area was additionally selected to minimize loss to follow-up in subsequent survey rounds as it is less urban than Chattogram City (bordering the southern end of Sitakunda) and the population is less transient. Individuals living in Sitakunda have few options for accessing traditional healthcare facilities by road due to the road networks in the Chattogram district, except for BITID and the Sitakunda UHC. Assuming a SARS-CoV-2 seroprevalance of 25%, we estimated that a sample size of 1,632 individuals (~408 households) would lead to a 95% confidence interval half-width of 1.8 percentage points. This assumed a design effect of 3, given a household-level ICC of 0.676 and an average household size of 4.1 Supplemental Figure 1. Map of the study population in the Sitakunda Upazila (in green) in the Chattogram District of Bangladesh. The 169 enrolled households sampled in the serosurvey and the two healthcare facilities in Sitakunda (BITID and the Sitakunda UHC) are indicated on the right side of the figure. Survey methods We conducted this household survey between March 26–April 13, 2021. The sampling frame was constructed using satellite imagery of the Sitakunda subdistrict (collected in the fourth quarter of 2020) and applying a deep-learning algorithm to identify single- and multistory structures, based on sizes of the structures and their shadows and validated by manual review using an external agency.2 Prior to the survey, a dwelling assessment was conducted on a random selection of satellite identified households to determine the proportion or identified dwellings that are residential and the distribution of the number of households in multistory residential structures. We used a two-stage cluster sampling approach to select 574 structures identified from the satellite imagery. The Sitakunda subdistrict was divided into clusters one square kilometer in size and clusters were first randomly selected with replacement. Weights accounted for the estimated number of households in multistory units using data from the dwelling assessment of the area. We oversampled the number of dwellings to account for non-residential structures (~40%). Data collection efforts were terminated after enrolling 169 households due to the national lockdown imposed on April 5th, 2021 in response to the rising number of COVID-19 cases. Study staff explained the study to the head of each household (or designee) and first asked for verbal consent from the household head to discuss the study with other members of the household. Household members were eligible if they were at least one years old, if they had resided in that village for the past year, intended to continue living there for the next year, and did not plan to spend more than two consecutive months away from the village in the coming year. Each eligible individual was asked for written informed consent to participate if home during one of up to three repeated visits. This study was approved by the icddr,b Research and Ethics Review Committees and the Johns Hopkins Bloomberg School of Public Health Institutional Review Board. Analysis We estimated seroprevalence using a Bayesian regression model with fixed effects for age group and sex and a random effect for household clustering and adjusting for immunoassay performance.1 Within this model we also estimate the immunoassay performance (sensitivity and specificity) using data collected in Dhaka from positive and negative controls. We sampled from the posterior distribution using Hamiltonian MCMC as implemented in Stan (used through rstan 2.21.1) and ran four separate chains for 1,500 iterations each.3 We used the R-hat statistic to evaluate convergence. We post-stratified model results using 2011 Census data on the age and sex disaggregation of the Sitakunda Upazila population.4 Measures of urbanicity, such as population density and time required to travel one meter, for each sampled cluster were categorized by quartiles and assigned to participants. We used a chi- square test for trend to determine whether seropositivity was associated with increased levels of urbanicity. We also evaluated differences in seropositivity among those reporting less transportation use using Pearson’s chi-squared test. All analyses were performed in R. Data and source code to reproduce analyses are available at https://github.com/HopkinsIDD/sitakunda-sarscov2-round1. Supplemental Results Study participation rates We visited 250 households between March 27 and April 13 and enrolled 169 of 170 residential households. We enrolled 71.2% of all household members (n=665/935) and tested 99.8% (n=664/665) of collected samples for SARS-CoV-2. Test for trend Seropositivity was significantly associated with levels of population density (p<0.001) and time required to travel one meter (p<0.001) within the 1 km2 cluster grid. Supplemental Table 1. Descriptive statistics for serosurvey participants (n=664) in Sitakunda Upazila by sociodemographic factors, measures of urbanicity, COVID-like symptoms, COVID testing and vaccination, and COVID-related behaviors. Characteristic Negative, N = 271a Positive, N = 393a Sociodemographic Age (median, range) 23 (2, 90) 30 (1, 85) Age Group (years) 1-4 12 (55%) 10 (45%) 5-9 28 (57%) 21 (43%) 10-14 44 (58%) 32 (42%) 15-24 60 (41%) 88 (59%) 25-34 45 (39%) 69 (61%) 35-44 28 (29%) 68 (71%) 45-54 25 (35%) 46 (65%) 55-64 18 (33%) 36 (67%) 65+ 11 (32%) 23 (68%) Sex Female 155 (41%) 221 (59%) Male 116 (40%) 172 (60%) Main activity in last month Business Outside Home 38 (35%) 70 (65%) Child 15 (54%) 13 (46%) Farmer 11 (44%) 14 (56%) Homemaker 98 (37%) 165 (63%) Not Worked (Adult) 10 (53%) 9 (47%) Other 6 (20%) 24 (80%) Student 93 (49%) 97 (51%) Highest educational attainment No schooling 42 (39%) 66 (61%) Primary 96 (45%) 115 (55%) Lower Secondary 77 (40%) 115 (60%) Upper Secondary 45 (38%) 75 (62%) Bachelors 7 (28%) 18 (72%) Postgraduate 2 (33%) 4 (67%) Household monthly income (USD)b <12 0 (0%) 0 (0%) 12-35 0 (0%) 0 (0%) 35-59 4 (100%) 0 (0%) 59-83 5 (36%) 9 (64%) 83-118 28 (53%) 25 (47%) 118-236 108 (40%) 163 (60%) >236 126 (39%) 196 (61%) Measures of urbanicity “Friction” (Minutes required to travel