A Deep Learning Approach to Forecast Short-Term COVID-19 Cases and Deaths in the US

Since the US reported its first COVID-19 case on January 21, 2020, the science community has been applying various techniques to forecast incident cases and deaths. To date, providing an accurate and robust forecast at a high spatial resolution has proved challenging, even in the short term. Here we present a novel multi-stage deep learning model to forecast the number of COVID-19 cases and deaths for each US state at a weekly level for a forecast horizon of 1 to 4 weeks. The model is heavily data driven, and relies on epidemiological, mobility, survey, climate, and demographic. We further present results from a case study that incorporates SARS-CoV-2 genomic data (i.e. variant cases) to demonstrate the value of incorporating variant cases data into model forecast tools. We implement a rigorous and robust evaluation of our model - specifically we report on weekly performance over a one-year period based on multiple error metrics, and explicitly assess how our model performance varies over space, chronological time, and different outbreak phases. The proposed model is shown to consistently outperform the CDC ensemble model for all evaluation metrics in multiple spatiotemporal settings, especially for the longer-term (3 and 4 weeks ahead) forecast horizon. Our case study also highlights the potential value of virus genomic data for use in short-term forecasting to identify forthcoming surges driven by new variants. Based on our findings, the proposed forecasting framework improves upon the available forecasting tools currently used to support public health decision making with respect to COVID-19 risk. Research in contextO_ST_ABSEvidence before this studyC_ST_ABSA systematic review of the COVID-19 forecasting and the EPIFORGE 2020 guidelines reveal the lack of consistency, reproducibility, comparability, and quality in the current COVID-19 forecasting literature. To provide an updated survey of the literature, we carried out our literature search on Google Scholar, PubMed, and medRxi, using the terms "Covid-19," "SARS-CoV-2," "coronavirus," "short-term," "forecasting," and "genomic surveillance." Although the literature includes a significant number of papers, it remains lacking with respect to rigorous model evaluation, interpretability and translation. Furthermore, while SARS-CoV-2 genomic surveillance is emerging as a vital necessity to fight COVID-19 (i.e. wastewater sampling and airport screening), to our knowledge, no published forecasting model has illustrated the value of virus genomic data for informing future outbreaks. Added value of this studyWe propose a multi-stage deep learning model to forecast COVID-19 cases and deaths with a horizon window of four weeks. The data driven model relies on a comprehensive set of input features, including epidemiological, mobility, behavioral survey, climate, and demographic. We present a robust evaluation framework to systematically assess the model performance over a one-year time span, and using multiple error metrics. This rigorous evaluation framework reveals how the predictive accuracy varies over chronological time, space, and outbreak phase. Further, a comparative analysis against the CDC ensemble, the best performing model in the COVID-19 ForecastHub, shows the model to consistently outperform the CDC ensemble for all evaluation metrics in multiple spatiotemporal settings, especially for the longer forecasting windows. We also conduct a feature analysis, and show that the role of explanatory features changes over time. Specifically, we note a changing role of climate variables on model performance in the latter half of the study period. Lastly, we present a case study that reveals how incorporating SARS-CoV-2 genomic surveillance data may improve forecasting accuracy compared to a model without variant cases data. Implications of all the available evidenceResults from the robust evaluation analysis highlight extreme model performance variability over time and space, and suggest that forecasting models should be accompanied with specifications on the conditions under which they perform best (and worst), in order to maximize their value and utility in aiding public health decision making. The feature analysis reveals the complex and changing role of factors contributing to COVID-19 transmission over time, and suggests a possible seasonality effect of climate on COVID-19 spread, but only after August 2021. Finally, the case study highlights the added value of using genomic surveillance data in short-term epidemiological forecasting models, especially during the early stage of new variant introductions.

Method versatility in RNA extraction-free PCR detection of SARS-CoV-2 in saliva samples

Early in the pandemic, a simple, open-source, RNA extraction-free RT-qPCR protocol for SARS-CoV-2 detection in saliva was developed and made widely available. This simplified approach (SalivaDirect) requires only sample treatment with proteinase K prior to PCR testing. However, feedback from clinical laboratories highlighted a need for a flexible workflow that can be seamlessly integrated into their current health and safety requirements for the receiving and handling of potentially infectious samples. To address these varying needs, we explored additional pre-PCR workflows. We built upon the original SalivaDirect workflow to include an initial incubation step (95{degrees}C for 30 minutes, 95{degrees}C for 5 minutes or 65{degrees}C for 15 minutes) with or without addition of proteinase K. The limit of detection for the workflows tested did not significantly differ from that of the original SalivaDirect workflow. When tested on de-identified saliva samples from confirmed COVID-19 individuals, these workflows also produced comparable virus detection and assay sensitivities, as determined by RT-qPCR analysis. Exclusion of proteinase K did not negatively affect the sensitivity of the assay. The addition of multiple heat pretreatment options to the SalivaDirect protocol increases the accessibility of this cost-effective SARS-CoV-2 test as it gives diagnostic laboratories the flexibility to implement the workflow which best suits their safety protocols.

Sequencing SARS-CoV-2 Genomes from Saliva

Genomic sequencing is crucial to understanding the epidemiology and evolution of SARS-CoV-2. Often, genomic studies rely on remnant diagnostic material, typically nasopharyngeal swabs, as input into whole genome SARS-CoV-2 next-generation sequencing pipelines. Saliva has proven to be a safe and stable specimen for the detection of SARS-CoV-2 RNA via traditional diagnostic assays, however saliva is not commonly used for SARS-CoV-2 sequencing. Using the ARTIC Network amplicon-generation approach with sequencing on the Oxford Nanopore MinION, we demonstrate that sequencing SARS-CoV-2 from saliva produces genomes comparable to those from nasopharyngeal swabs, and that RNA extraction is necessary to generate complete genomes from saliva. In this study, we show that saliva is a useful specimen type for genomic studies of SARS-CoV-2.

COVID-19 Outcomes and Sequencing of SARS-CoV-2 isolated from Veterans in New England

In an older cohort of veterans with a high comorbidity burden, age was the largest predictor of hospitalization, peak disease severity, and mortality. Most infections in six New England states until September, 2020, were from SARS-CoV-2 B.1 lineage, dominated by spike protein D614G substitution in 97.3% of samples.

Saliva viral load is a dynamic unifying correlate of COVID-19 severity and mortality

While several clinical and immunological parameters correlate with disease severity and mortality in SARS-CoV-2 infection, work remains in identifying unifying correlates of coronavirus disease 2019 (COVID-19) that can be used to guide clinical practice. Here, we examine saliva and nasopharyngeal (NP) viral load over time and correlate them with patient demographics, and cellular and immune profiling. We found that saliva viral load was significantly higher in those with COVID-19 risk factors; that it correlated with increasing levels of disease severity and showed a superior ability over nasopharyngeal viral load as a predictor of mortality over time (AUC=0.90). A comprehensive analysis of immune factors and cell subsets revealed strong predictors of high and low saliva viral load, which were associated with increased disease severity or better overall outcomes, respectively. Saliva viral load was positively associated with many known COVID-19 inflammatory markers such as IL-6, IL-18, IL-10, and CXCL10, as well as type 1 immune response cytokines. Higher saliva viral loads strongly correlated with the progressive depletion of platelets, lymphocytes, and effector T cell subsets including circulating follicular CD4 T cells (cTfh). Anti-spike (S) and anti-receptor binding domain (RBD) IgG levels were negatively correlated with saliva viral load showing a strong temporal association that could help distinguish severity and mortality in COVID-19. Finally, patients with fatal COVID-19 exhibited higher viral loads, which correlated with the depletion of cTfh cells, and lower production of anti-RBD and anti-S IgG levels. Together these results demonstrated that viral load - as measured by saliva but not nasopharyngeal -- is a dynamic unifying correlate of disease presentation, severity, and mortality over time.

Detection of SARS-CoV-2 RNA by multiplex RT-qPCR

The current RT-qPCR assay recommended for SARS-CoV-2 testing in the United States requires analysis of three genomic targets per sample: two viral and one host. To simplify testing and reduce the volume of required reagents, we developed a multiplex RT-qPCR assay to detect SARS-CoV-2 in a single reaction. We used existing N1, N2, and RP primer and probe sets by the CDC, but substituted fluorophores to allow multiplexing of the assay. The cycle threshold (Ct) values of our multiplex RT-qPCR were comparable to those obtained by the singleplex assay adapted for research purposes. Low copies (>500 copies / reaction) of SARS-CoV-2 RNA were consistently detected by the multiplex RT-qPCR. Our novel multiplex RT-qPCR improves upon current singleplex diagnostics by saving reagents, costs, time and labor.