A Mixed-Effects Model to Predict COVID-19 Hospitalizations Using Wastewater Surveillance (preprint)

The potential utility of wastewater-based epidemiology as an early warning tool has been explored widely across the globe during the COVID-19 pandemic. Early in the pandemic, methods were developed to detect the presence of SARS-CoV-2 RNA in wastewater. Since then, extensive research has been conducted to study the relationship between viral concentration in wastewater and COVID-19 cases in catchment areas of sewage treatment plants over time. However, few reports, to date, have attempted to develop predictive models for hospitalizations using SARS-CoV-2 RNA concentrations in wastewater. This study uses wastewater data to forecast hospitalizations using a linear mixed-effects model that allows for repeated measures and fixed and random effects. We use wastewater data from various treatment plants in California to predict hospitalizations at the county level assuming data from March 14, 2022, to May 21, 2023. The results suggest that wastewater data can serve as a dependable substitute for clinical data in creating robust models to predict hospitalizations. This approach can enhance our understanding of community-level transmission and its impact on hospital capacity.

Bayesian sequential approach to monitor COVID-19 variants through positivity rate from wastewate (preprint)

Trends in COVID-19 infection have changed throughout the pandemic due to myriad factors, including changes in transmission driven by social behavior, vaccine development and uptake, mutations in the virus genome, and public health policies. Mass testing was an essential control measure for curtailing the burden of COVID-19 and monitoring the magnitude of the pandemic during its multiple phases. However, as the pandemic progressed, new preventive and surveillance mechanisms emerged. Implementing vaccine programs, wastewater (WW) surveillance, and at-home COVID-19 tests reduced the demand for mass severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) testing. This paper proposes a sequential Bayesian approach to estimate the COVID-19 positivity rate (PR) using SARS-CoV-2 RNA concentrations measured in WW through an adaptive scheme incorporating changes in virus dynamics. PR estimates are used to compute thresholds for WW data using the CDC thresholds for low, substantial, and high transmission. The effective reproductive number estimates are calculated using PR estimates from the WW data. This approach provides insights into the dynamics of the virus evolution and an analytical framework that combines different data sources to continue monitoring the COVID-19 trends. These results can provide public health guidance to reduce the burden of future outbreaks as new variants continue to emerge. The proposed modeling framework was applied to the City of Davis and the campus of the University of California Davis.

Participant Demographics and Testing Trends: A Community Pandemic Response Program (preprint)

Objective: This study aimed to identify demographic characteristics of test participants and changes in testing participation over time in a community pandemic-response program launched in a college town in California, USA. Methods: We described overall testing participation, identified demographic characteristics of frequent testers, and evaluated changes in testing participation over four different periods of the COVID-19 pandemic. Results: A total of 770,165 tests were performed between November 18, 2020, and June 30, 2022, among 89,924 residents of Yolo County (41.1\% of population), with significant participation from racially/ethnically diverse participants and across age groups. Most positive cases (49.9\%) were captured during Omicron, which also corresponds to the period with the highest daily participation (895 per 100K population). The proportion of participants which we considered "frequent testers" (28.9\% vs. 39.7\%, p<0.0001) and individuals that tested once (39.5\% vs. 47.9\%, p<0.0001) increased significantly from Delta to Omicron. Women (58.8\%), participants of age 19-34 years (38.8\%), and White (53.2\%) tested more frequently throughout the program. The proportion of tests conducted among Latinos remained steady around 18\% over time, with the exception of the post-Omicron period (13\%). Conclusion: The unique features of a pandemic response program that supported community-wide access to free asymptomatic testing provides a unique opportunity to evaluate adherence to testing recommendations, and testing trends over time. Identification of individual and group-level factors associated with testing behaviors is essential to improve access and protect communities at-large.

Model training periods impact estimation of COVID-19 incidence from wastewater viral loads (preprint)

Background: Wastewater-based epidemiology (WBE) has been deployed broadly as an early warning tool for emerging COVID-19 outbreaks. WBE can inform targeted interventions and identify communities with high transmission, enabling quick and effective response. As wastewater becomes an increasingly important indicator for COVID-19 transmission, more robust methods and metrics are needed to guide public health decision making. Objectives: The aim of this research was to develop and implement a mathematical framework to infer incident cases of COVID-19 from SARS-CoV-2 levels measured in wastewater. We propose a classification scheme to assess the adequacy of model training periods based on clinical testing and assess the sensitivity of model predictions to training periods. Methods: We present a Bayesian deconvolution method and linear regression to estimate COVID-19 cases from wastewater data. We described an approach to characterize adequacy in testing during specific time periods and provided evidence to highlight the importance of model training periods on the projection of cases. We estimated the effective reproductive number (Re) directly from observed cases and from the reconstructed incidence of cases from wastewater. The proposed modeling framework was applied to three Northern California communities served by distinct wastewater treatment plants. Results: Both deconvolution and linear regression models consistently projected robust estimates of prevalent cases and Re from wastewater influent samples when assuming training periods with adequate testing. Case estimates from models that used poorer-quality training periods consistently underestimated observed cases. Discussion: Wastewater surveillance data requires robust statistical modeling methods to provide actionable insight for public health decision-making. We propose and validate a modeling framework that can provide estimates of prevalent COVID-19 cases and Re from wastewater data that can be used as tool for disease surveillance including quality assessment for potential training data.

The Role of SARS-CoV-2 Testing on Hospitalizations in California (preprint)

The rapid spread of the new SARS-CoV-2 virus triggered a global health crisis disproportionately impacting people with pre-existing health conditions and particular demographic and socioeconomic characteristics. One of the main concerns of governments has been to avoid the overwhelm of health systems. For this reason, they have implemented a series of non-pharmaceutical measures to control the spread of the virus, with mass tests being one of the most effective control. To date, public health officials continue to promote some of these measures, mainly due to delays in mass vaccination and the emergence of new virus strains. In this study, we studied the association between COVID-19 positivity rate and hospitalization rates at the county level in California using a mixed linear model. The analysis was performed in the three waves of confirmed COVID-19 cases registered in the state to September 2021. Our findings suggest that test positivity rate is consistently associated with hospitalization rates at the county level for all waves of study. Demographic factors that seem to be related with higher hospitalization rates changed over time, as the profile of the pandemic impacted different fractions of the population in counties across California.

COVID-19 Vaccination in California: Are We Equitable Yet? (preprint)

Background. By March 2021, California had one of the least equitable vaccine distribution programs in the US. To rectify this, Governor Newsom ordered 4 million vaccine doses be reserved for the census tracts in the lowest quartile of the Healthy Places Index (HPI). California plans to lift state-wide COVID-19 restrictions on June 15th, 2021, as long as test positivity and vaccine equity thresholds are met in the state's most vulnerable census tracts. This short investigation examines current vaccine equity and forecasts where California can expect to be on June 15th. Methods. Current vaccine equity was investigated with simple linear regression between the county mean HPI and both single and full-dose vaccination rate. Future vaccination coverage per county were forecast using a compartmental mathematical model based on the average rate over the previous 30 days with four different rate-change scenarios. Results. County mean HPI has a strong positive association with both single and full dose vaccination rates (R2: 0.716 and 0.737, respectively). We predict the overall state rate will exceed 50% fully vaccinated by June 15th if the current rates are maintained; however, the bulk of this coverage comes from the top 18 counties while the remaining 40 counties lag behind. Discussion. The clear association between county HPI and current vaccination rates shows that California is not initiating opening plans from an equitable foundation, despite previous equity programs. If nothing changes, many of the most vulnerable counties will not be prepared to open without consequences come June 15th.

Bayesian sequential data assimilation for COVID-19 forecasting (preprint)

We introduce a Bayesian sequential data assimilation method for COVID-19 forecasting. It is assumed that suitable transmission, epidemic and observation models are available and previously validated and the transmission and epidemic models are coded into a dynamical system. The observation model depends on the dynamical system state variables and parameters, and is cast as a likelihood function. We elicit prior distributions of the effective population size, the dynamical system initial conditions and infectious contact rate, and use Markov Chain Monte Carlo sampling to make inference and prediction of quantities of interest (QoI) at the onset of the epidemic outbreak. The forecast is sequentially updated over a sliding window of epidemic records as new data becomes available. Prior distributions for the state variables at the new forecasting time are assembled using the dynamical system, calibrated for the previous forecast. Moreover, changes in the contact rate and effective population size are naturally introduced through auto-regressive models on the corresponding parameters. We show our forecasting method's performance using a SEIR type model and COVID-19 data from several Mexican localities.