Modeling the relative influence of socio-demographic variables on post-acute COVID-19 quality of life: an application to settings in Europe, Asia, Africa, and South America (preprint)

Long-term COVID-19 complications are a globally pervasive threat, but their plausible social drivers are often not prioritized. Here, we use data from a multinational consortium to quantify the relative contributions of social and clinical factors to differences in quality of life among participants experiencing long COVID and measure the extent to which social variables impacts can be attributed to clinical intermediates, across diverse contexts. In addition to age, neuropsychological and rheumatological comorbidities, educational attainment, employment status, and female sex were identified as important predictors of long COVID-associated quality of life days (long COVID QALDs). Furthermore, a great majority of their impacts on long COVID QALDs could not be tied to key long COVID-predicting comorbidities, such as asthma, diabetes, hypertension, psychological disorder, and obesity. In Norway, 90% (95% CI: 77%, 100%) of the effect of belonging to the highest versus lowest educational attainment quintile was not attributed to intermediate comorbidity impacts. The same was true for 86% (73%, 100%) of the protective effects of full-time employment versus all other employment status categories (excluding retirement) in the UK and 74% (46%,100%) of the protective effects of full-time employment versus all other employment status categories in a cohort of four middle-income countries (MIC). Of the effects of female sex on long COVID QALDs in Norway, UK, and the MIC cohort, 77% (46%,100%), 73% (52%, 94%), and 84% (62%, 100%) were unexplained by the clinical mediators, respectively. Our findings highlight that socio-economic proxies and sex may be as predictive of long COVID QALDs as commonly emphasized comorbidities and that broader structural determinants likely drive their impacts. Importantly, we outline a multi-method, adaptable causal machine learning approach for evaluating the isolated contributions of social disparities to long COVID quality of life experiences.

Mathematical assessment of the role of human behavior changes on SARS-CoV-2 transmission dynamics (preprint)

The COVID-19 pandemic has not only presented a major global public health and socio-economic crisis, but has also significantly impacted human behavior towards adherence (or lack thereof) to public health intervention and mitigation measures implemented in communities worldwide. The dynamic nature of the pandemic has prompted extensive changes in individual and collective behaviors towards the pandemic. This study is based on the use of mathematical modeling approaches to assess the extent to which SARS-CoV-2 transmission dynamics is impacted by population-level changes of human behavior due to factors such as (a) the severity of transmission (such as disease-induced mortality and level of symptomatic transmission), (b) fatigue due to the implementation of mitigation interventions measures (e.g., lockdowns) over a long (extended) period of time, (c) social peer-pressure, among others. A novel behavior-epidemiology model, which takes the form of a deterministic system of nonlinear differential equations, is developed and fitted using observed cumulative SARS-CoV-2 mortality data during the first wave in the United States. Rigorous analysis of the model shows that its disease-free equilibrium is locally-asymptotically stable whenever a certain epidemiological threshold, known as the control reproduction number (denoted by RC) is less than one, and the disease persists (i.e., causes significant outbreak or outbreaks) if the threshold exceeds one. The model fits the observed data, as well as makes a more accurate prediction of the observed daily SARS-CoV-2 mortality during the first wave (March 2020 -June 2020), in comparison to the equivalent model which does not explicitly account for changes in human behavior.  Of the various metrics for human behavior changes during the pandemic considered in this study, it is shown that behavior changes due to the level of SARS-CoV-2 mortality and symptomatic transmission were more influential (while behavioral changes due to the level of fatigue to interventions in the community was of marginal impact). It is shown that an increase in the proportion of exposed individuals who become asymptomatically-infectious at the end of the exposed period (represented by a parameter r) can lead to an increase (decrease) in the control reproduction number (RC) if the effective contact rate of asymptomatic individuals is higher (lower) than that of symptomatic individuals. The study identifies two threshold values of the parameter r that maximize the cumulative and daily SARS-CoV-2 mortality, respectively, during the first wave. Furthermore, it is shown that, as the value of the proportion r increases from 0 to 1, the rate at which susceptible non-adherent individuals change their behavior to strictly adhere to public health interventions decreases. Hence, this study suggests that, as more newly-infected individuals become asymptomatically-infectious, the level of positive behavior change, as well as disease severity, hospitalizations and disease-induced mortality in the community can be expected to significantly decrease (while new cases may rise, particularly if asymptomatic individuals have higher contact rate, in comparison to symptomatic individuals).

Mathematical assessment of the role of human behavior changes on SARS-CoV-2 transmission dynamics (preprint)

The COVID-19 pandemic has not only presented a major global public health and socio-economic crisis, but has also significantly impacted human behavior towards adherence (or lack thereof) to public health intervention and mitigation measures implemented in communities worldwide. The dynamic nature of the pandemic has prompted extensive changes in individual and collective behaviors towards the pandemic. This study is based on the use of mathematical modeling approaches to assess the extent to which SARS-CoV-2 transmission dynamics is impacted by population-level changes of human behavior due to factors such as (a) the severity of transmission (such as disease-induced mortality and level of symptomatic transmission), (b) fatigue due to the implementation of mitigation interventions measures (e.g., lockdowns) over a long (extended) period of time, (c) social peer-pressure, among others. A novel behavior-epidemiology model, which takes the form of a deterministic system of nonlinear differential equations, is developed and fitted using observed cumulative SARS-CoV-2 mortality data during the first wave in the United States. Rigorous analysis of the model shows that its disease-free equilibrium is locally-asymptotically stable whenever a certain epidemiological threshold, known as the control reproduction number (denoted by[R] C) is less than one, and the disease persists (i.e., causes significant outbreak or outbreaks) if the threshold exceeds one. The model fits the observed data, as well as makes a more accurate prediction of the observed daily SARS-CoV-2 mortality during the first wave (March 2020 -June 2020), in comparison to the equivalent model which does not explicitly account for changes in human behavior. Of the various metrics for human behavior changes during the pandemic considered in this study, it is shown that behavior changes due to the level of SARS-CoV-2 mortality and symptomatic transmission were more influential (while behavioral changes due to the level of fatigue to interventions in the community was of marginal impact). It is shown that an increase in the proportion of exposed individuals who become asymptomatically-infectious at the end of the exposed period (represented by a parameter r) can lead to an increase (decrease) in the control reproduction number ([R]C) if the effective contact rate of asymptomatic individuals is higher (lower) than that of symptomatic individuals. The study identifies two threshold values of the parameter r that maximize the cumulative and daily SARS-CoV-2 mortality, respectively, during the first wave. Furthermore, it is shown that, as the value of the proportion r increases from 0 to 1, the rate at which susceptible non-adherent individuals change their behavior to strictly adhere to public health interventions decreases. Hence, this study suggests that, as more newly-infected individuals become asymptomatically-infectious, the level of positive behavior change, as well as disease severity, hospitalizations and disease-induced mortality in the community can be expected to significantly decrease (while new cases may rise, particularly if asymptomatic individuals have higher contact rate, in comparison to symptomatic individuals).

Fine-Grained Forecasting of COVID-19 Trends at the County Level in the United States (preprint)

The coronavirus (COVID-19) pandemic has profoundly impacted various aspects of daily life, society, healthcare systems, and global health policies. This pandemic has resulted in more than one hundred million people being infected and, unfortunately, the loss of life for many individuals. Although treatment for the coronavirus is now available, effective forecasting of COVID-19 infection is the most importance to aid public health officials in making critical decisions. However, forecasting COVID-19 trends through time-series analysis poses significant challenges due to the datas inherently dynamic, transient, and noise-prone nature. In this study, we have developed the Fine-Grained Infection Forecast Network (FIGI-Net) model, which provides accurate forecasts of COVID-19 trends up to two weeks in advance. FIGI-Net addresses the current limitations in COVID-19 forecasting by leveraging fine-grained county-level data and a stacked bidirectional LSTM structure. We employ a pre-trained model to capture essential global infection patterns. Subsequently, these pre-trained parameters were transferred to train localized sub-models for county clusters exhibiting comparable infection dynamics. This model adeptly handles sudden changes and rapid fluctuations in data, frequently observed across various times and locations of county-level data, ultimately improving the accuracy of COVID-19 infection forecasting at the county, state, and national levels. FIGI-Net model demonstrated significant improvement over other deep learning-based models and state-of-the-art COVID-19 forecasting models, evident in various standard evaluation metrics. Notably, FIGI-Net model excels at forecasting the direction of infection trends, especially during the initial phases of different COVID-19 outbreak waves. Our study underscores the effectiveness and superiority of our time-series deep learning-based methods in addressing dynamic and sudden changes in infection numbers over short-term time periods. These capabilities facilitate efficient public health management and the early implementation of COVID-19 transmission prevention measures.

Correlates of symptomatic remission among individuals with post-COVID-19 condition (preprint)

Importance: Post-COVID-19 condition (PCC), or long COVID, has become prevalent. The course of this syndrome, and likelihood of remission, has not been characterized. Objective: To quantify the rates of remission of PCC, and the sociodemographic features associated with remission. Design: 16 waves of a 50-state U.S. non-probability internet survey conducted between August 2020 and November 2022 Setting: Population-based Participants: Survey respondents age 18 and older Main Outcome and Measure: PCC remission, defined as reporting full recovery from COVID-19 symptoms among individuals who on a prior survey wave reported experiencing continued COVID-19 symptoms beyond 2 months after the initial month of symptoms. Results: Among 423 survey respondents reporting continued symptoms more than 2 months after acute test-confirmed COVID-19 illness, who then completed at least 1 subsequent survey, mean age was 53.7 (SD 13.6) years; 293 (69%) identified as women, and 130 (31%) as men; 9 (2%) identified as Asian, 29 (7%) as Black, 13 (3%) as Hispanic, 15 (4%) as another category including Native American or Pacific Islander, and the remaining 357 (84%) as White. Overall, 131/423 (31%) of those who completed a subsequent survey reported no longer being symptomatic. In Cox regression models, male gender, younger age, lesser impact of PCC symptoms at initial visit, and infection when the Omicron strain predominated were all statistically significantly associated with greater likelihood of remission; presence of brain fog or shortness of breath were associated with lesser likelihood of remission. Conclusions and Relevance: A minority of individuals reported remission of PCC symptoms, highlighting the importance of efforts to identify treatments for this syndrome or means of preventing it.

Association between long COVID symptoms and employment status (preprint)

Background: Symptoms of Coronavirus-19 (COVID-19) infection persist beyond 2 months in a subset of individuals, a phenomenon referred to as long COVID, but little is known about its functional correlates and in particular the relevance of neurocognitive symptoms. Method: We analyzed a previously-reported cohort derived from 8 waves of a nonprobability-sample internet survey called the COVID States Project, conducted every 4-8 weeks between February 2021 and July 2022. Primary analyses examined associations between long COVID and lack of full employment or unemployment, adjusted for age, sex, race and ethnicity, education, urbanicity, and region, using multiple logistic regression with interlocking survey weights. Results: The cohort included 15,307 survey respondents ages 18-69 with test-confirmed COVID-19 at least 2 months prior, of whom 2,236 (14.6%) reported long COVID symptoms, including 1,027/2,236 (45.9%) reporting either 'brain fog' or impaired memory. Overall, 1,418/15,307 (9.3%) reported being unemployed, including 276/2,236 (12.3%) of those with long COVID and 1,142/13,071 (8.7%) of those without; 8,228 (53.8%) worked full-time, including 1,017 (45.5%) of those with long COVID and 7,211 (55.2%) without. In survey-weighted regression models, presence of long COVID was associated with being unemployed (crude OR 1.44, 95% CI 1.20-1.72; adjusted OR 1.23, 95% CI 1.02-1.48), and with lower likelihood of working full-time (crude OR 0.73, 95% CI 0.64-0.82; adjusted OR 0.79, 95% CI 0.70 -0.90). Among individuals with long COVID, the presence of cognitive symptoms -- either brain fog or impaired memory -- was associated with lower likelihood of working full time (crude OR 0.71, 95% CI 0.57-0.89, adjusted OR 0.77, 95% CI 0.61-0.97). Conclusion: Long COVID was associated with a greater likelihood of unemployment and lesser likelihood of working full time in adjusted models. Presence of cognitive symptoms was associated with diminished likelihood of working full time. These results underscore the importance of developing strategies to respond to long COVID, and particularly the associated neurocognitive symptoms.

Leveraging Serosurveillance and Postmortem Surveillance to Quantify the Impact of COVID-19 in Africa (preprint)

Background The COVID-19 pandemic has had a devastating impact on global health, the magnitude of which appears to differ intercontinentally: for example, reports suggest 271,900 per million people have been infected in Europe versus 8,800 per million people in Africa. While Africa is the second largest continent by population, its reported COVID-19 cases comprise <3% of global cases. Although social, environmental, and environmental explanations have been proposed to clarify this discrepancy, systematic infection underascertainment may be equally responsible. Methods We seek to quantify magnitude of underascertainment in COVID-19's cumulative incidence in Africa. Using serosurveillance and postmortem surveillance, we constructed multiplicative factors estimating ratios of true infections to reported cases in African nations since March 2020. Results Multiplicative factors derived from serology data - in a subset of 12 nations - suggested a range of COVID-19 reporting rates, from 1 in 630 infections reported in Kenya (May 2020) to 1 in 15 infections reported in South Africa (November 2021). The largest multiplicative factor, 3,795, corresponded to Malawi (June 2020), suggesting <0.05% of infections captured. A similar set of multiplicative factors for all nations derived from postmortem data points toward the same conclusion: reported COVID-19 cases are unrepresentative of true infections, suggesting a key reason for low case burden in many African nations is significant underdetection and underreporting. Conclusions While estimating COVID-19's exact burden is challenging, the multiplicative factors we present provide incidence curves reflecting likely-to-worst-case ranges of infection. Our results stress the need for expansive surveillance to allocate resources in areas experiencing severe discrepancies between reported cases, projected infections, and deaths.

Using digital traces to build prospective and real-time county-level early warning systems to anticipate COVID-19 outbreaks in the United States (preprint)

The ongoing COVID-19 pandemic continues to affect communities around the world. To date, almost 6 million people have died as a consequence of COVID-19, and more than one-quarter of a billion people are estimated to have been infected worldwide. The design of appropriate and timely mitigation strategies to curb the effects of this and future disease outbreaks requires close monitoring of their spatio-temporal trajectories. We present machine learning methods to anticipate sharp increases in COVID-19 activity in US counties in real-time. Our methods leverage Internet-based digital traces -- e.g., disease-related Internet search activity from the general population and clinicians, disease-relevant Twitter micro-blogs, and outbreak trajectories from neighboring locations -- to monitor potential changes in population-level health trends. Motivated by the need for finer spatial-resolution epidemiological insights to improve local decision-making, we build upon previous retrospective research efforts originally conceived at the state level and in the early months of the pandemic. Our methods -- tested in real-time and in an out-of-sample manner on a subset of 97 counties distributed across the US -- frequently anticipated sharp increases in COVID-19 activity 1-6 weeks before the onset of local outbreaks (defined as the time when the effective reproduction number $R_t$ becomes larger than 1 consistently). Given the continued emergence of COVID-19 variants of concern -- such as the most recent one, Omicron -- and the fact that multiple countries have not had full access to vaccines, the framework we present, while conceived for the county-level in the US, could be helpful in countries where similar data sources are available.

Characterizing features of outbreak duration for novel SARS-CoV-2 variants of concern (preprint)

Characterizing the dynamics of epidemic trajectories is critical to understanding the potential impacts of emerging outbreaks and to designing appropriate mitigation strategies. As the COVID-19 pandemic evolves, however, the emergence of SARS-CoV-2 variants of concern has complicated our ability to assess in real-time the potential effects of imminent outbreaks, such as those presently caused by the Omicron variant. Here, we report that SARS-CoV-2 outbreaks across regions exhibit strain-specific times from onset to peak, specifically for Delta and Omicron variants. Our findings may facilitate real-time identification of peak medical demand and may help fine-tune ongoing and future outbreak mitigation deployment efforts.

The Evolving Roles of US Political Partisanship and Social Vulnerability in the COVID-19 Pandemic from February 2020 - February 2021 (preprint)

The COVID-19 pandemic has had intense, heterogeneous impacts on different communities and geographies in the United States. We explore county level associations between COVID-19 attributed deaths and social, demographic, vulnerability, and political variables to develop a better understanding of the evolving roles these variables play in relation to mortality. We focus on the role of political variables, as captured by support for either the Republican or Democrat presidential candidates in the 2020 elections and the stringency of state-wide governor mandates, during three non-overlapping time periods between February 2020 and February 2021. We find that during the first three months of the pandemic, Democratic-leaning and internationally-connected urban counties were affected. During subsequent months (between May and September, 2020), Republican counties with high percentages of Hispanic and Black populations were most hardly hit. Importantly, in the time period between October 2020 and February 2021, when the effectiveness of non-pharmaceutical interventions, such as social distancing and wearing masks indoors, had been well-established. During this period, we find that Republican-leaning counties experienced up to 3 times higher death rates than Democratic-leaning counties, even after controlling for multiple social vulnerability factors. We also find that Republican-leaning counties in states with less strict mandates experienced the most severe outbreaks. Our findings suggest that ideologies promoted by prominent political actors may not align with efforts to mitigate the impact of the ongoing pandemic and prevent deaths.

High coverage COVID-19 mRNA vaccination rapidly controls SARS-CoV-2 transmission in Long-Term Care Facilities (preprint)

Residents of Long-Term Care Facilities (LTCFs) represent a major share of COVID-19 deaths worldwide. Measuring the vaccine effectiveness among the most vulnerable in these settings is essential to monitor and improve mitigation strategies. We evaluated the early effect of the administration of BNT162b2 mRNA vaccines to individuals older than 64 years residing in LTCFs in Catalonia, a region of Spain. We monitored all the SARS-CoV-2 documented infections and deaths among LTCFs residents from February 6th to March 28th, 2021, the subsequent time period after which 70% of them were fully vaccinated. We developed a modeling framework based on the relation between community and LTFCs transmission during the pre-vaccination period (July -December 2020) and compared the true observations with the counterfactual model predictions. As a measure of vaccine effectiveness, we computed the total reduction in SARS-CoV-2 documented infections and deaths among residents of LTCFs over time, as well as the reduction on the detected transmission for all the LTCFs. We estimated that once more than 70% of the LTCFs population were fully vaccinated, 74% (58%-81%, 90% CI) of COVID-19 deaths and 75% (36%-86%, 90% CI) of all expected documented infections among LTCFs residents were prevented. Further, detectable transmission among LTCFs residents was reduced up to 90% (76-93%, 90%CI) relative to that expected given transmission in the community. Our findings provide evidence that high-coverage vaccination is the most effective intervention to prevent SARS-CoV-2 transmission and death among LTCFs residents. Conditional on key factors such as vaccine roll out, escape and coverage --across age groups--, widespread vaccination could be a feasible avenue to control the COVID-19 pandemic.

Comparison of post-COVID depression and major depressive disorder (preprint)

BackgroundDuring the COVID-19 pandemic rates of depressive symptoms are markedly elevated, particularly among survivors of infection. Understanding whether such symptoms are distinct among those with prior SARS-CoV-2 infection, or simply a nonspecific reflection of elevated stress, could help target interventions. MethodWe analyzed data from multiple waves of a 50-state survey that included questions about COVID-19 infection as well as the Patient Health Questionnaire examining depressive and anxious symptoms. We utilized multiple logistic regression to examine whether sociodemographic features associated with depression liability differed for those with or without prior COVID-19, and then whether depressive symptoms differed among those with or without prior COVID-19. ResultsAmong 91,791 respondents, in regression models, age, gender, race, education, and income all exhibited an interaction with prior COVID-19 in risk for moderate or greater depressive symptoms (p<0.0001 in all cases), indicating differential risk in the two subgroups. Among those with such symptoms, levels of motoric symptoms and suicidality were significantly greater among those with prior COVID-19 illness. Depression risk increased with greater interval following acute infection. DiscussionOur results suggest that major depressive symptoms observed among individuals with prior COVID-19 illness may not reflect typical depressive episodes, and merit more focused neurobiological investigation.

High coverage COVID-19 mRNA vaccination rapidly controls SARS-CoV-2 transmission in Long-Term Care Facilities (preprint)

Residents of Long-Term Care Facilities (LTCFs) represent a major share of COVID-19 deaths worldwide. Information on vaccine effectiveness in these settings is essential to improve mitigation strategies, but evidence remains limited. To evaluate the early effect of the administration of BNT162b2 mRNA vaccines in LTCFs, we monitored subsequent SARS-CoV-2 documented infections and deaths in Catalonia, a region of Spain, and compared them to counterfactual model predictions from February 6th to March 28th, 2021, the subsequent time period after which 70% of residents were fully vaccinated. We calculated the reduction in SARS-CoV-2 documented infections and deaths as well as the detected county-level transmission. We estimated that once more than 70% of the LTCFs population were fully vaccinated, 74% (58%-81%, 90% CI) of COVID-19 deaths and 75% (36%-86%) of all documented infections were prevented. Further, detectable transmission was reduced up to 90% (76-93% 90%CI). Our findings provide evidence that high-coverage vaccination is the most effective intervention to prevent SARS-CoV-2 transmission and death. Widespread vaccination could be a feasible avenue to control the COVID-19 pandemic.

Near real-time surveillance of the SARS-CoV-2 epidemic with incomplete data (preprint)

Designing public health responses to outbreaks requires close monitoring of population-level health indicators in real-time. Thus an accurate estimation of the epidemic curve is critical. We propose an approach to reconstruct epidemic curves in near real time. We apply this approach to characterize the early SARS-CoV-2 outbreak in two Spanish regions between the months of March and April 2020. We address two data collection problems that affected the reliability of the available real-time epidemiological data, namely, the frequent missing information documenting when a patient first experienced symptoms, and the frequent retrospective revision of historical information (including right censoring). This is done by using a novel back-calculating procedure based on imputing patients dates of symptom onset from reported cases, according to a dynamically-estimated backward reporting delay conditional distribution, and adjusting for right censoring using an existing package, NobBS, to estimate in real time (nowcast) cases by date of symptom onset. This process allows us to obtain an approximation of the time-varying reproduction number (Rt) in real-time. At each step, we evaluate how different assumptions affect the recovered epidemiological events and compare the proposed approach to the alternative procedure of merely using curves of case counts, by report day, to characterize the time-evolution of the outbreak. Finally, we assess how these real-time estimates compare with subsequently documented epidemiological information that is considered more reliable and complete that became available weeks to months later in time. Our approach may help improve accuracy, quantify uncertainty, and evaluate frequently unstated assumptions when recovering the epidemic curves from limited data obtained from public health surveillance systems in other locations.

Socioeconomic status determines COVID-19 incidence and related mortality in Santiago, Chile (preprint)

The current coronavirus disease 2019 (COVID-19) pandemic has impacted dense urban populations particularly hard. Here, we provide an in-depth characterization of disease incidence and mortality patterns, and their dependence on demographic and socioeconomic strata in Santiago, a highly segregated city and the capital of Chile. We find that among all age groups, there is a strong association between socioeconomic status and both mortality –measured either by direct COVID-19 attributed deaths or excess deaths– and public health capacity. Specifically, we show that behavioral factors like human mobility, as well as health system factors such as testing volumes, testing delays, and test positivity rates are associated with disease outcomes. These robust patterns suggest multiple possibly interacting pathways that can explain the observed disease burden and mortality differentials: (i) in lower socioeconomic status municipalities, human mobility was not reduced as much as in more affluent municipalities; (ii) testing volumes in these locations were insufficient early in the pandemic and public health interventions were applied too late to be effective; (iii) test positivity and testing delays were much higher in less affluent municipalities, indicating an impaired capacity of the health-care system to contain the spread of the epidemic; and (iv) infection fatality rates appear much higher in the lower end of the socioeconomic spectrum. Together, these findings highlight the exacerbated consequences of health-care inequalities in a large city of the developing world, and provide practical methodological approaches useful for characterizing COVID-19 burden and mortality in other segregated urban centers.

Stress as a meaningful variable in models of covid-19 spreading (preprint)

Covid-19 is a respiratory disease caused by the severe acute respiratory syndrome coronavirus – 2 (SARS-CoV-2), which was first identified in Wuhan China in December 2019. Because of Covid-19 worldwide spreading in 2020, urgent hygiene and lockdown measures have been implemented with fundamental consequences on the lives of people in all sectors of society. Besides visible socio-economic and documented health effects, the impact of Covid-19 on people’s mental health including stress-related effects on disease spreading remains yet to be addressed. Here, we argue about the relevance of incorporating stress factors into models of covid-19 spreading and the level of detail in which such models should take stress into consideration.

High-resolution Spatio-temporal Model for County-level COVID-19 Activity in the U.S (preprint)

We present an interpretable high-resolution spatio-temporal model to estimate COVID-19 deaths together with confirmed cases one-week ahead of the current time, at the county-level and weekly aggregated, in the United States. A notable feature of our spatio-temporal model is that it considers the (a) temporal auto- and pairwise correlation of the two local time series (confirmed cases and death of the COVID-19), (b) dynamics between locations (propagation between counties), and (c) covariates such as local within-community mobility and social demographic factors. The within-community mobility and demographic factors, such as total population and the proportion of the elderly, are included as important predictors since they are hypothesized to be important in determining the dynamics of COVID-19. To reduce the model's high-dimensionality, we impose sparsity structures as constraints and emphasize the impact of the top ten metropolitan areas in the nation, which we refer (and treat within our models) as hubs in spreading the disease. Our retrospective out-of-sample county-level predictions were able to forecast the subsequently observed COVID-19 activity accurately. The proposed multi-variate predictive models were designed to be highly interpretable, with clear identification and quantification of the most important factors that determine the dynamics of COVID-19. Ongoing work involves incorporating more covariates, such as education and income, to improve prediction accuracy and model interpretability.

An Early Warning Approach to Monitor COVID-19 Activity with Multiple Digital Traces in Near Real-Time (preprint)

Non-pharmaceutical interventions (NPIs) have been crucial in curbing COVID-19 in the United States (US). Consequently, relaxing NPIs through a phased re-opening of the US amid still-high levels of COVID-19 susceptibility could lead to new epidemic waves. This calls for a COVID-19 early warning system. Here we evaluate multiple digital data streams as early warning indicators of increasing or decreasing state-level US COVID-19 activity between January and June 2020. We estimate the timing of sharp changes in each data stream using a simple Bayesian model that calculates in near real-time the probability of exponential growth or decay. Analysis of COVID-19-related activity on social network microblogs, Internet searches, point-of-care medical software, and a metapopulation mechanistic model, as well as fever anomalies captured by smart thermometer networks, shows exponential growth roughly 2-3 weeks prior to comparable growth in confirmed COVID-19 cases and 3-4 weeks prior to comparable growth in COVID-19 deaths across the US over the last 6 months. We further observe exponential decay in confirmed cases and deaths 5-6 weeks after implementation of NPIs, as measured by anonymized and aggregated human mobility data from mobile phones. Finally, we propose a combined indicator for exponential growth in multiple data streams that may aid in developing an early warning system for future COVID-19 outbreaks. These efforts represent an initial exploratory framework, and both continued study of the predictive power of digital indicators as well as further development of the statistical approach are needed.