The study leveraged on an ongoing population-based infectious disease surveillance (PBIDS) platform. The PBIDS platform conducts surveillance within two well-characterized populations in Asembo (rural western Kenya in Siaya County) and in Kibera (the largest urban, densely populated, informal settlement in Nairobi) and is run by Kenya Medical Research Institute-Centre for Global Health Research (KEMRI-CGHR) with technical and financial support from US Centers for Disease Control and prevention (CDC) ^{6, 7} . The platform has been running since 2006 and the surveillance objectives and methods have been described previously ^{6, 7} . Briefly, the PBIDS platform aims to monitor burden and aetiology of common and (re)emerging acute infectious diseases and evaluate the impact of public health interventions in the two populations in Kenya, Figure 1. In Kibera, PBIDS covers two villages characterized by very poor water purification, supply, and waste disposal, and a large number of infectious diseases including an adult HIV prevalence of 15% ^{6, 10} . The surveillance area covers 0.40 km ², with a high population density (~77,000 individuals/km ²). The Asembo site covers 33 villages that are sparsely populated (~350 individuals/km ²) in ~80 km ². The area is culturally homogeneous (95% Luo ethnicity); subsistence farming and fishing constitute the principal economy. The area has perennial, high-level malaria transmission and has reported high prevalence of HIV (~15%) among adults aged 15–64 years. Consequently, both sites have mortality rates that reflect a high number of infectious diseases ^{11– 13} .

Enrolled households are followed regularly (once in 2020 due to COVID-19 interruptions from April to September and thrice in 2021) to collect data on recent illnesses, healthcare-seeking behaviour, and demographic characteristics. The households are within 5 km radius of St Elizabeth Lwak Mission Hospital (LMH) in Asembo and a 1 km radius of the Tabitha Medical Clinic in Kibera. The active PBIDS participants receive free medical care for acute illnesses at the centrally located surveillance clinics. All households within the defined radius are invited to participate in PBIDS and enrolment is continuous. As of December 2020, the Asembo PBIDS had 34,999 persons in 9,225 households, while Kibera PBIDS had 23,103 persons in 5,265 households under follow-up.

A longitudinal cohort of randomly selected households from the PBIDS database were enrolled and serial sampling of all available household members conducted in March and June 2021, Figure 2. The target was to enrol approximately 900 individuals in each site. For unavailable households and household members, three attempts for enrolment were made before randomly selected replacement households were considered. The survey size was anticipated to detect a seroprevalence of 45% with a precision of 5% and design effect of 2 at 95% confidence interval. An attrition of 20% was also incorporated. Our expected seroprevalence of 45% was informed by earlier serosurveys reporting a seroprevalence of 13% among blood bank specimens in Kenya ^{14, 15} and an observation of 3.6 times higher seroprevalence among informal settlement residents compared to non-informal settlement residents in India ¹⁶ .

The household members’ consenting, enrolment, data and blood collection were conducted through home visits by trained field teams consisting of a field worker and a phlebotomist. Venous blood samples (approximately 5 ml for persons aged >12 years; 2–3 ml for children 2–12 years and 1.5 ml for children <2 years) were collected from each participant and transported in a cool box at 2-8°C to the laboratory on the same day. Sera were separated from the whole blood specimen and stored at -80 ^oC before testing.

At the surveillance clinic in each site, PBIDS participants of all ages presenting with severe acute respiratory illness (SARI) were consented for a nasopharyngeal and oropharyngeal (NP/OP) specimen. Since 1 ^st May 2020, all the NP/OP specimens collected were tested for SARS-CoV-2 using real time Reverse Transcription-Polymerase Chain Reaction (rRT-PCR). From September 2020, the eligibility criterion for NP/OP collection was expanded to include any patients presenting with acute febrile illness (AFI) or acute respiratory illness (ARI) as well as known contacts of confirmed COVID-19 cases regardless of their symptom status. Table 1 presents the case definitions used in the enhanced COVID-19 surveillance.

All laboratory tests were performed in an international organization for standardization (ISO)15189 certified and Good Clinical Practice-accredited CDC-supported laboratories at KEMRI-CGHR in Kisumu and Nairobi, Kenya. The serum specimens were tested for anti- Spike IgG antibodies according to manufacturer’s instructions using the SCoV-2 Detect™ IgG ELISA kit (Catalogue #64824, InBios International, Inc, USA). The kit manufacturer reports of 92% and Specificty of 99% ^{17, 18} . The NP/OP swabs were tested by rRT-PCR for SARS-CoV-2 and select positives sequenced for variant detection using partial and full genome sequencing at the CDC-supported laboratories in Kisumu.

All data cleaning, management and analyses were performed using Stata 15.1 software (STATA Corp, Texas, USA) (free alternative, RStudio) in accordance with methods we previously published ⁹ . Briefly, individual seroprevalence of SARS-CoV-2 antibodies was defined as a percentage of the seropositive individuals among those tested accounting for household clustering and weighting by age and sex distribution of the PBIDS general population in each surveillance site as of March 2021 ( Figure 3). The standard errors for generating the 95% confidence intervals were computed using the Taylor linearized variance estimation method ¹⁹ . Pearson’s chi-squared test was used to assess the association of categorical variables with individual seropositivity. Household seroprevalence (defined as the percentage of households with at least one seropositive member) was estimated and stratified by household size (usual number of persons in the household), and number of persons enrolled in the serosurvey per household. Age, sex, relationship to head of the household, main occupation, household size, and underlying medical conditions associated with serious COVID-19 complications ( e.g., known hypertensive, asthmatic or diabetic) were considered in the univariable logistic regression model for determining the factors associated with individual seropositivity for each site and survey round. Age and sex were considered a priori for inclusion in the multivariable logistic regression. The final multivariable logistic regression model also included variables with p-value of ≤0.05 and accounted for sampling weights and clustering by household using the clustered sandwich estimator ^{20, 21} . Adjusted odds ratio (aOR) and 95% confidence intervals (CI) were presented and two-sided p-values <0.05 were considered statistically significant. In addition, weekly SARS-CoV-2 positivity (percentage of RT-PCR positive cases from the total tested in a week) by rRT-PCR from the health facility surveillance is presented to provide context to the seroprevalence surveys up to 31 ^st December 2021.

Individual written informed consent/verbal assent was obtained from all the study participants and/or their parents/guardian. Ethical approval for the study was provided by the KEMRI Scientific and Ethical Review Committee in Kenya (#4168, February 15 ^th, 2021). This activity was also reviewed by CDC and was conducted consistent with applicable federal law and CDC policy as provided for in the Code of Federal Regulations (45 C.F.R part 46 and 21 C.F.R. part 56). The PBIDS protocol is approved by KEMRI Scientific and Ethical Review Committee in Kenya (#2761), Washington State University reliance agreement and CDC institutional review board (#6775).

From 1 ^st May 2020 to 31 ^st December 2021, SARS-CoV-2 was detected at varying intensities at different time periods, as shown in Figure 4. In Asembo, three waves were observed by 31 ^st December 2021: first wave in September to December 2020, second wave from March to July 2021 and the last from early December 2021 and ongoing by the end of the reporting period. In comparison, Kibera had five waves: first wave was ongoing from the beginning of the COVID-19 surveillance in May 2020 and ended in August 2020, second wave from September 2020 to January 2021, third wave from February 2021 to May 2021, fourth from June to October 2021, and fifth wave began in November 2021 and was ongoing by the end of the reporting period. The second wave in Asembo and the fourth in Kibera were predominated by the Delta variant (unpublished reports, Clayton Onyango) and started earlier in Asembo in western Kenya before being detected in Kibera in Nairobi. The latest wave was associated with Omicron (BA.1) variant (unpublished reports, Clayton Onyango) and had the highest weekly RT-PCR positivity at the peak in December 2021 of 50% in Asembo and 87% in Kibera.

Of the 349 randomly selected households (with 1,265 members) in Asembo, household heads in 79.1% (276 households with 1,188 eligible members) gave permission for their households to participate ( Figure 2). Of the 469 randomly selected households (with 1,589 members), household heads in 64.6% (294 households with 1,415 eligible members) agreed to participate in Kibera. Of the 76 households in Asembo that did not participate, 37 (48.7%) declined participation, 13 (17.1%) did not have a household head available for consenting, 11 (14.5%) had moved, and 12 (15.8%) provided no reason for not participating. In Kibera, 172 households did not accept to participate; 44 (25.6%) declined participation, 56 (32.6%) did not have a household head available for consenting, 63 (36.6%) had moved, and 12 (0.6%) provided no reason. Individual written informed consent was obtained from 955 (67.5% of eligible members) and 882 (74.2%) participants in Asembo and Kibera, respectively. Of the 233 who didn’t consent in Asembo, 115 (49.4%) had out-migrated, 69 (29.6%) were not found at home, even after three attempts, 47 (20.2%) declined and two (0.9%) provided no reason. In Kibera, 533 did not consent; 347 (65.1%) were not found at home, 145 (27.2%) declined participation, 6 (1.1%) had out-migrated and 35 (6.6%) provided no reason.

The median number of individuals in a household was three (interquartile range (IQR), 1–5; range, 1–15) in Asembo, and three (IQR, 1–5; range, 1–19) in Kibera while the corresponding median number of consented individuals per household was three (IQR, 2–5; range, 1–13) in Asembo and three (IQR, 1–4; range, 1–11) in Kibera.

In Asembo, female participants aged 50 years and above, male participants over 60 years and both sexes aged 5–19 years old were overrepresented, while men aged 20–29 years were underrepresented in the two surveys, relative to the PBIDS population ( Figure 3). For Kibera, male children aged 5–9 years and women aged 30 years and above were overrepresented, while both female and male children below 5 years of age and men aged 30–59 years were underrepresented in both surveys, compared to the PBIDS general population ( Figure 3).

The first serosurvey (round one, R1) was conducted between 19 ^th February and 28 ^th March 2021 (median date, 11 ^th March 2021), and the second round (R2) between 4 ^th June to 4 ^th July 2021 (median date, 15 ^th June 2021) yielding 859 and 750 serum specimens, respectively, in Kibera ( Table 2). In Asembo, 935 and 888 sera specimens were collected in the first (12 ^th March to 8 ^th April 2021, median date, 25 ^th March 2021) and second (8 ^th July to 4 ^th August 2021, median date 19 ^th July 2021) serosurvey, respectively. The median number of sampled individuals per household was three (IQR, 2–5; range, 1–11) in Asembo and three (IQR, 1–4; range, 1–11) in Kibera. During the second round of serosurveys, consent was obtained for an additional 115 and 160 individuals and they were sampled from the participating households in Asembo and Kibera, respectively. These were mainly individuals who were not available in the households for enrolment during the first round of the serosurvey.

The majority of the sampled individuals during the two rounds in both sites were female (55.4% in Asembo and 57.5% in Kibera in the first round, Table 2). The average age in years for participants was higher for Asembo participants compared to Kibera participants (26.6 (95% CI: 25.3, 28.0) versus 22.7 (95% CI: 21.7, 23.8) years during the first serosurvey).

Of the tested individuals, 234 (25.0%) and 440 (49.5%) were seropositive in Asembo for the first and second serosurvey, respectively. Correspondingly, 478 (55.6%) and 483 (64.4%) individuals were seropositive in the respective rounds in Kibera. The overall weighted individual seroprevalence increased from 56.2% (95% CI: 52.1, 60.2%) in March 2021 to 63.9% (95% CI: 59.5, 68.0%) in June 2021 in Kibera. For Asembo, the weighted seroprevalence almost doubled from 26.0% (95% CI: 22.4, 30.0%) in March 2021 to 48.7% (95% CI: 44.3, 53.2%) in July 2021. At the respective time points for the serosurvey, Kibera had consistently higher seroprevalence than Asembo. In both sites and for each of the two rounds, the seroprevalence did not vary by sex, but was lower in children compared to adults ( Table 3). An appreciable increase in age-specific individual seroprevalence over the rounds was observed in all age groups in Asembo (14.3–28.5% increase) and among young participants in the 5–19 years age groups (11.8–17.7%) in Kibera.

In Kibera, the age effect was also observed for seroprevalence estimates by relationship to the household head in the first round with grandchildren (25.0%; 95% CI: 10.1, 51.2%) and children (52.1%; 95% CI: 47.1, 57.1%) registering lower estimates compared to the household head (59.6%; 95% CI: 51.1, 67.7 %) and other adults including spouses (70.9%; 95% CI: 62.9, 77.7%) and other relatives (54.1%; 95% CI: 30.1, 76.4%). However, these differences in seroprevalence by relationships were not statistically significant in the second serosurvey in Kibera as well as in both rounds in Asembo.

Seroprevalence in each round and site increased with the level of education, with the lowest among those with no education and highest among those with secondary or above level of education—again highlighting the age effect as the young children were coded to have no education. There was no obvious pattern in seroprevalence based on main occupation groups ( Table 3) for the two sites and over the two survey rounds. Participants with any underlying medical conditions had similar seroprevalence as those without in each round and site ( Table 3).

COVID-19 vaccine uptake was low at the time these serosurveys were conducted. None of the study participants had received the COVID-19 vaccine before round one of the serosurvey in both sites. By round two, 11 (1.2%) participants reported having received the COVID-19 vaccine in Asembo, while in Kibera, 12 (1.6%) reported receipt of the vaccine—only one participant in Asembo and three participants in Kibera reported receipt of two doses. All the vaccinated participants had received Oxford-AstraZeneca vaccine except for one participant from Kibera who reported receiving Johnson & Johnson COVID-19 vaccine. Of the 23 individuals who reported having received the COVID-19 vaccine in round two of the serosurvey, 18 (78.3%) were seropositive—nine from each site. Among those reporting receipt of COVID-19 vaccine, seropositivity was 81.8% in Asembo and 75.0% in Kibera.

Samples tested were from 276 and 262 households in Asembo and 294 and 252 in Kibera for round one and two, respectively ( Table 4). Among sampled households in Asembo, 145 (52.5%; 95% CI: 46.5–58.6%) and 208 (79.4%; 95% CI: 74.0–84.1%) had at least one seropositive participant in the first and second surveys, respectively. In Kibera, 237 households (80.6%; 95% CI: 75.6–85.0%) and 220 households (87.3%; 95% CI: 82.5–91.1%) had at least one seropositive participant in round one and two, respectively. Within households with at least one seropositive individual, the median (IQR) percentage of seropositive individuals in the household was 40.0% (25.0–60.0%) and 63.3% (40.0–100%) in Asembo for round one and two, respectively. The corresponding median (IQR) estimates for Kibera were higher at 75.0% (50.0–100%) and 80.0% (50.0–100%). Of 223 and 210 households with two or more members enrolled in Asembo, 133 (59.6%) and 176 (83.8%) had at least one person seropositive in the respective survey rounds. On the other hand, the mean seropositivity among households with at least one seropositive person and two or more participants enrolled in Asembo was 42.3% (95% CI: 38.4–46.1%) and 58.6% (95% CI: 54.7–62.6%) in round one and two of the serosurvey, respectively. For Kibera, the corresponding mean seropositivity in the similar households was 62.7% (95% CI: 58.9– 66.6%) and 68.1% (95% CI: 64.3–71.9%) in round one and two of the serosurvey, respectively.

Age was a predictor of individual seropositivity. Relative to adults aged 20–29 years, young age groups (<20 years) had lower odds of being seropositive in the respective sites and rounds. The odds of being seropositive were similar for older adults (age groups ≥30 years) compared to the reference group, 20–29 years. The odds of being seropositive among the elderly groups (≥60 years) were not different from the referent 20–29 years age group across the sites and by survey round. Relative to household heads, spouses had higher odds of being seropositive in Asembo round two (adjusted Odds Ratio, aOR, 2.47; 95% CI: 1.35–4.54) and Kibera round one (aOR; 1.89 (95% CI: 1.08–3.31). During round two of serosurvey in Asembo, the odds of seropositivity were higher in the employed participants compared to those not employed, aOR, 3.37 (95% CI: 1.76–6.45). This association of employment status and seropositivity was not observed in Kibera. Sex and highest level of education were not significantly associated with individual seropositivity for any of the sites or rounds ( Table 5).

Of the individuals enrolled in round one, 754 (80.6%) in Asembo and 606 (70.6%) in Kibera participated in the second serosurvey. The median duration between the serosurveys was longer in Asembo (3.9 months; IQR, 3.6–4.0 months) than in Kibera (3.2 months; IQR, 2.9–3.4) months). Between the two serosurveys, 215/564 (38.1%) of the previously seronegative participants seroconverted in Asembo (median age, 18.1 (IQR, 11.8–40.1 years) and 98/273 (35.9%) in Kibera (median age, 12.1 (IQR, 6.7–25.1 years). Conversely, 29/190 (15.3%) of the previously seropositive participants returned a negative result in the second serosurvey in Asembo (median age, 23.2 (IQR, 12.5–38.3 years), whereas in Kibera 40/333 (12.0%) sero-reverted (median age, 32.8 (IQR, 23.7–40.6 years). There was no change in serostatus in 161/190 (84.7%) with median age of 21.7 (IQR, 14.2–42.9 years), and 293/333 (88.0%) with median age of 22.3 (IQR, 12.4–36.3 years), of participants who were previously seropositive, and 349/564 (61.9%) with median age, 19.1 (IQR, 9.8–45.4 years), and 175/273 (64.1%) with median age, 12.1 (IQR, 6.7–25.1 years), of those who were previously seronegative, in Asembo and Kibera, respectively. There were statistically significant differences in median ages by change or no change in the serostatus in Asembo. In Kibera, there was age dependent pattern with median age increasing from lowest in those not seroconverting, followed by those seroconverting and those non sero-reverting, to highest in those sero-reverting.