Testing participation was aggregated and subdivided into distinct periods according to the dominant variants Delta and Omicron: pre-Delta (November 18, 2020, to June 13, 2021), Delta (June 14, 2021, to December 20, 2021), Omicron (December 21, 2021, to March 15, 2022), and post-Omicron (March 16, 2022, to June 30, 2022). These periods were based on the most common variant reported by the California Department of Public Health [25] (Figure 1). Hereafter, the Delta B.1.617.2, Omicron B.1.1.529, and subsequent Omicron BA.2, BA.3, BA.4.266, and BA.5 variants will be referred to as Delta, Omicron, and post-Omicron, respectively. During the pre-Delta period, HDT primarily served the city of Davis. Subsequently, at the onset of the Delta variant in July 2021, testing access was extended to encompass Yolo County as a whole.

Yolo County has an estimated population of 218,774 individuals, with 106,002 males and 112,772 females. Among them, approximately 68,640 reside in the city of Davis. Throughout the program, 770,165 tests were administered to 89,924 unique individuals, including 53,869 residents of the city of Davis. A significant number of participants in Davis were linked to the University of California, Davis. The university enforced weekly testing mandates starting in the fall of 2020 for those who were physically present on campus.

HDT program participants were given a unique ID, which allowed us to distinguish between new and returning individuals. An individual can be classified as a unique participant for a specific period (eg, Delta) if they underwent testing at least once during that period, regardless of whether they were also tested during another period. However, for the overall program analysis, each participant was counted only once, regardless of how many times they were tested across multiple periods. New participants are individuals who underwent testing for the first time. Total tests refer to the overall number of tests conducted within a specific period or in total. Positive tests denote the total count of tests yielding a positive result, distinct from participants with a positive result, which indicates the number of individuals who had at least one positive test during a specified period.

This study used deidentified data and received an exemption from institutional review by the University of California Davis Office of Research.

Saliva tests were conducted using an innovative high-throughput quantitative reverse transcriptase–PCR method [24]. Most tests (98.4%) comprised of saliva tests while the remaining 1.6% consisted of BinaxNOW (Abbott) rapid antigen tests administered to symptomatic individuals. Employing saliva samples for quantitative reverse transcription PCR tests enhances the feasibility of widespread asymptomatic testing in the community, as it incurs lower monetary and labor costs with only a slight decrease in sensitivity [26].

We used regression modeling for recurrent events to analyze testing behavior, given that participants underwent multiple tests at irregular intervals. This modeling framework facilitates the analysis and interpretation of testing recurrence, providing insights into factors influencing testing adherence, frequency, and consistency among participants. Additionally, trends, fluctuations, and potential predictors of testing behavior can be identified, offering insights for optimizing testing strategies, resource allocation, and public health interventions to promote regular and sustained testing participation. The reReg library in R (R Foundation for Statistical Computing) was used to analyze the temporal dynamics of testing behavior [27,28].

The regression Cox model for recurrent events extends the traditional Cox proportional hazards model, which is commonly used for analyzing time-to-event data in survival analysis, to handle recurrent event data. Additionally, this modeling framework enables the examination of covariate effects on the rate of event occurrence while properly addressing the correlation among recurrent events within the same individual. This model, also known as the Andersen-Gill model, can be represented as follows:

where λ_i (t) represents the hazard rate for individual i at time t. λ₀ (t) is the baseline hazard rate at time t. β₁X_i1 + β₂X_i2 +...+ β_pX_ip denotes the hazard ratio (HR) associated with the covariates X_i1, X_i2,..., X_ip with β_1, β₂,…, β_p representing the regression coefficients for the covariates included in the model. This formula expresses the hazard rate for everyone as a function of the baseline hazard rate and the covariates, with the regression coefficients determining the impact of the covariates on the hazard rate. We assumed noninformative censoring, suggesting that the reasons people decided to take part in testing during the program are not related to why they might eventually stop testing altogether.

The Mean Cumulative Function (MCF) denotes the expected cumulative count of tests up to a specific time point, considering both the occurrence and timing of test events. The MCF was calculated by age and racial or ethnic groups. We also apply the terminal event process Cox model to examine the occurrence of complete cessation of testing while accounting for covariates.

For this study, individuals who undergo COVID-19 testing at least once a week are selected. Each selected individual is treated as a single event, even if they receive multiple tests within the same week. This study includes 61,363 participants who have undergone testing on at least 2 occasions (2 different weeks). The event for each participant begins with their first test and ends during the subsequent week of another test. A test occurrence is marked as 1, while the status remains 0 to signify ongoing participation. The final testing instance for each participant is identified, and the event of the subsequent week is marked as 0 to indicate a censoring event.

HDT conducted 770,165 tests among 89,924 residents of Yolo County (approximately 41.1% of the Yolo population, 218.793), of which 53,869 individuals reported a zip code in the city of Davis (approximately 78.4% of Davis population, 68.710). Testing coverage was most significant in the city of Davis, where the program originated, and efforts were first focused. Figure 1 illustrates the total daily tests per 100,000 population and positive test rates observed in Yolo County throughout this study. On June 28, 2021, the positivity rate surged, likely because there were fewer tests reported on that day.

Overall, 12,626 tests were positive among 11,545 unique individuals, for an aggregated test positivity rate of 1.64% and a case rate of 13%. The average daily number of tests conducted per 100,000 population was 422 (SD 210) during the pre-Delta period, 746 (SD 449) during Delta, 895 (SD 518) during Omicron, and 435 (SD 254) during the post-Omicron period. A summary of tests and positive cases per study period is reported in Table 1.

Total tests conducted, participation, recruitment, and positive test tallies across pandemic periods are displayed in Table 1. For instance, during pre-Delta, 191,904 tests were conducted, which corresponds to 24.9% (n=770,165) of the total tests conducted in the entire program. During Delta, 2570 positive cases were detected in new participants, corresponding to 22.3% (n=11,545) of the total of unique people who resulted as positive.

Changes in policy at the state and local level, social expectations and behavior, the emergence of new strains, and differences in the intensity of testing lead to differing participation rates across pandemic periods. In total, HDT or HYT served 89,924 unique individuals across the entire program. A total of 39,819 (44.3%) participated in testing during pre-Delta, 58,920 (65.5%) during Delta, 46,545 (51.8%) during Omicron, and 23,566 (26.2%) during the post-Omicron period. Further, 61,363 individuals, equivalent to 68.2% of all participants, took part in the testing process multiple times. A longer period of days did not mean a higher number of participants. The highest participation of people was during the Delta period.

Omicron had the highest number of positive cases despite being the shortest period in the program. Out of all positive cases, 6351 (50.3%) were observed during Omicron. On average, 1958 (SD 1333; ∼895, SD 518, tests per 100,000) daily tests were performed at a test positivity rate of 3.81% (Table 1). The enrollment of new participants decreased with each subsequent period: 39,819 individuals were engaged in the first stage of the program (pre-Delta), 33,814 during Delta, 13,503 during Omicron, and 2788 during post-Omicron. Most positive tests (a total of 2080, 97.1%) were from new participants (Table 1).

HDT or HYT collaborated with multiple school districts in Yolo County to provide free weekly testing to students. Excluding participants under 19, there were 67,083 participants from the community; among them, 8774 (13%) individuals tested positive, and 422 (4.8%) had positive results in two or more weeks. Further, 265,000 (34.4%) underwent testing once, while participation ranged from 1 to 83 weeks, with a median of 7.5 (IQR 7) weeks. Figure 2A illustrates uptake in testing by individuals, showing the weeks they underwent testing and information on when everyone was initially tested. It is noticeable that at the beginning of each testing wave, there is typically a higher flux of new participants signing up each week. It also enables the observation of participants’ ongoing engagement throughout the program. Figure 2B illustrates the weeks in which an individual received at least one positive test result, showing a notable increase in positive cases during the Omicron wave. Figure 2C describes key measures implemented during the Healthy Davis Program, with color bars representing the seasons: winter (blue), spring (green), summer (yellow), and fall (orange).

Testing rates between male and female residents varied across 4 distinct periods and overall. Female participants underwent more testing compared to males. Participants aged 18 or younger were less involved in testing during the pre-Delta period but increased their participation during the Delta period. Individuals aged 19-34 years underwent more testing in the pre-Delta period when compared to the Delta and post-Omicron periods. Participants who identified as Hispanics or Latinos had lower participation rates in post-Omicron (Table S1 in Multimedia Appendix 1 [29]).

A total of 48,670 participants identified as female, making up 45.9% of the 106,034 estimated female population in Yolo County. Meanwhile, 39,524 were male, representing 35% of the 112,759 estimated male population in the county. Testing in all race or ethnic groups and age groups younger than 85 years exceeded 54,698 (25%) of Yolo County’s population (Figure 3). The lowest testing participation rate was observed among individuals aged older than 85 years. This is not unexpected, given that testing for participants aged 85 years and older was compulsory at congregate living facilities funded by Medicare or Medicaid, as part of several state-level testing initiatives. The highest overall test positivity rates were observed among individuals of American Indian or Alaska Native descent (7.6%), Native Hawaiian or Other Pacific Islander (5.6%), Latinos (4.5%), and people aged 20-44 years (20-34 years, 6.5%; 35-44 years, 6.9%).

In a sample of 61,363 participants that tested during this study period, there were 652,232 tests, with an average number of 10.62 (SD 13) tests per individual, in a median follow-up time of 40 (IQR 41.5) weeks. The average number of tests observed per week (MCF) by age and racial or ethnic groups is described in Figure 4A. Participants between the ages of 0 and 19 years demonstrated more consistent engagement in testing compared to other age groups, maintaining a higher level of involvement throughout this study’s period. Participants identifying as multiracial exhibited higher levels of engagement in testing compared to those identifying as Hispanic, Latino, American Indian, or Alaska Native (Figure 4B). Throughout this study period, the average testing patterns remained consistent across various age and racial or ethnic groups.

Figure 5 illustrates the baseline cumulative rate function as the expected number of tests conducted up to a particular time point, independent of any covariates included in the model (Figure 5A). We also describe the baseline cumulative hazard function in recurrent tests, representing the expected cumulative number of tests occurring up to a specific time point, without considering any covariates or risk factors (Figure 5B). It provides a baseline estimate of testing occurrence over time.

The recurrent event analysis showed that male participants (HR 0.969, 95% CI 0.943-0.996) and individuals 85 years and older (HR 0.781, 95% CI 0.598-1.020) experienced lower adherence to testing compared to females and younger age groups (0-19 years), respectively (Table 2). Hispanic or Latino participants (HR 0.741, 95% CI 0.712-0.772) and American Indian or Alaska Native individuals (HR 0.735, 95% CI 0.567-0.953) experienced lower testing rates compared to individuals from other racial or ethnic groups.

Male participants showed a significantly higher risk of discontinuing testing compared to females (HR 1.10, 95% CI 1.08-1.13), see Table 3. Individuals aged 20-34 years (HR 2.62, 95% CI 2.54-2.70) exhibited the highest hazard of discontinuing testing, followed by those aged 35-44 years (HR 1.51, 95% CI 1.47-1.56) and 45-54 years (HR 1.31, 95% CI 1.27-1.35), compared to those aged 0-19 years. Participants identifying as multiracial (HR 1.52, 95% CI 1.20-1.94), American Indian or Alaska Native (HR 1.20, 95% CI 1.05-1.36), and Hispanic or Latino (HR 1.19, 95% CI 1.15-1.22) were more prone to discontinuing testing compared to White participants. These findings underscore the importance of targeted outreach strategies tailored to engage and retain the demographic groups at risk of stopping testing. Addressing barriers such as accessibility, convenience, and perceived need for testing among younger age groups and specific racial or ethnic groups may help mitigate the discontinuation of testing and continue surveillance of SARS-CoV-2 transmission within communities.