Global monitoring of the impact of the COVID-19 pandemic through online surveys sampled from the Facebook user base

Exploring the Big Data Paradox for various estimands using vaccination data from the global COVID-19 Trends and Impact Survey (CTIS)

Selection bias poses a substantial challenge to valid statistical inference in nonprobability samples. This study compared estimates of the first-dose COVID-19 vaccination rates among Indian adults in 2021 from a large nonprobability sample, the COVID-19 Trends and Impact Survey (CTIS), and a small probability survey, the Center for Voting Options and Trends in Election Research (CVoter), against national benchmark data from the COVID Vaccine Intelligence Network. Notably, CTIS exhibits a larger estimation error on average (0.37) compared to CVoter (0.14). Additionally, we explored the accuracy (regarding mean squared error) of CTIS in estimating successive differences (over time) and subgroup differences (for females versus males) in mean vaccine uptakes. Compared to the overall vaccination rates, targeting these alternative estimands comparing differences or relative differences in two means increased the effective sample size. These results suggest that the Big Data Paradox can manifest in countries beyond the United States and may not apply equally to every estimand of interest.

Vaccine effectiveness against emerging COVID-19 variants using digital health data

BACKGROUND

Participatory surveillance of self-reported symptoms and vaccination status can be used to supplement traditional public health surveillance and provide insights into vaccine effectiveness and changes in the symptoms produced by an infectious disease. The University of Maryland COVID Trends and Impact Survey provides an example of participatory surveillance that leveraged Facebook's active user base to provide self-reported symptom and vaccination data in near real-time.

METHODS

Here, we develop a methodology for identifying changes in vaccine effectiveness and COVID-19 symptomatology using the University of Maryland COVID Trends and Impact Survey data from three middle-income countries (Guatemala, Mexico, and South Africa). We implement conditional logistic regression to develop estimates of vaccine effectiveness conditioned on the prevalence of various definitions of self-reported COVID-like illness in lieu of confirmed diagnostic test results.

RESULTS

We highlight a reduction in vaccine effectiveness during Omicron-dominated waves of infections when compared to periods dominated by the Delta variant (median change across COVID-like illness definitions: -0.40, IQR[-0.45, -0.35]. Further, we identify a shift in COVID-19 symptomatology towards upper respiratory type symptoms (i.e., cough and sore throat) during Omicron periods of infections. Stratifying COVID-like illness by the National Institutes of Health's (NIH) description of mild and severe COVID-19 symptoms reveals a similar level of vaccine protection across different levels of COVID-19 severity during the Omicron period.

CONCLUSIONS

Participatory surveillance data alongside methodologies described in this study are particularly useful for resource-constrained settings where diagnostic testing results may be delayed or limited.

Changes to Public Health Surveillance Methods Due to the COVID-19 Pandemic: Scoping Review

BACKGROUND

Public health surveillance plays a vital role in informing public health decision-making. The onset of the COVID-19 pandemic in early 2020 caused a widespread shift in public health priorities. Global efforts focused on COVID-19 monitoring and contact tracing. Existing public health programs were interrupted due to physical distancing measures and reallocation of resources. The onset of the COVID-19 pandemic intersected with advancements in technologies that have the potential to support public health surveillance efforts.

OBJECTIVE

This scoping review aims to explore emergent public health surveillance methods during the early COVID-19 pandemic to characterize the impact of the pandemic on surveillance methods.

METHODS

A scoping search was conducted in multiple databases and by scanning key government and public health organization websites from March 2020 to January 2022. Published papers and gray literature that described the application of new or revised approaches to public health surveillance were included. Papers that discussed the implications of novel public health surveillance approaches from ethical, legal, security, and equity perspectives were also included. The surveillance subject, method, location, and setting were extracted from each paper to identify trends in surveillance practices. Two public health epidemiologists were invited to provide their perspectives as peer reviewers.

RESULTS

Of the 14,238 unique papers, a total of 241 papers describing novel surveillance methods and changes to surveillance methods are included. Eighty papers were review papers and 161 were single studies. Overall, the literature heavily featured papers detailing surveillance of COVID-19 transmission (n=187). Surveillance of other infectious diseases was also described, including other pathogens (n=12). Other public health topics included vaccines (n=9), mental health (n=11), substance use (n=4), healthy nutrition (n=1), maternal and child health (n=3), antimicrobial resistance (n=2), and misinformation (n=6). The literature was dominated by applications of digital surveillance, for example, by using big data through mobility tracking and infodemiology (n=163). Wastewater surveillance was also heavily represented (n=48). Other papers described adaptations to programs or methods that existed prior to the COVID-19 pandemic (n=9). The scoping search also found 109 papers that discuss the ethical, legal, security, and equity implications of emerging surveillance methods. The peer reviewer public health epidemiologists noted that additional changes likely exist, beyond what has been reported and available for evidence syntheses.

CONCLUSIONS

The COVID-19 pandemic accelerated advancements in surveillance and the adoption of new technologies, especially for digital and wastewater surveillance methods. Given the investments in these systems, further applications for public health surveillance are likely. The literature for surveillance methods was dominated by surveillance of infectious diseases, particularly COVID-19. A substantial amount of literature on the ethical, legal, security, and equity implications of these emerging surveillance methods also points to a need for cautious consideration of potential harm.

Performance and explainability of feature selection-boosted tree-based classifiers for COVID-19 detection

In this paper, we evaluate the performance and analyze the explainability of machine learning models boosted by feature selection in predicting COVID-19-positive cases from self-reported information. In essence, this work describes a methodology to identify COVID-19 infections that considers the large amount of information collected by the University of Maryland Global COVID-19 Trends and Impact Survey (UMD-CTIS). More precisely, this methodology performs a feature selection stage based on the recursive feature elimination (RFE) method to reduce the number of input variables without compromising detection accuracy. A tree-based supervised machine learning model is then optimized with the selected features to detect COVID-19-active cases. In contrast to previous approaches that use a limited set of selected symptoms, the proposed approach builds the detection engine considering a broad range of features including self-reported symptoms, local community information, vaccination acceptance, and isolation measures, among others. To implement the methodology, three different supervised classifiers were used: random forests (RF), light gradient boosting (LGB), and extreme gradient boosting (XGB). Based on data collected from the UMD-CTIS, we evaluated the detection performance of the methodology for four countries (Brazil, Canada, Japan, and South Africa) and two periods (2020 and 2021). The proposed approach was assessed in terms of various quality metrics: F1-score, sensitivity, specificity, precision, receiver operating characteristic (ROC), and area under the ROC curve (AUC). This work also shows the normalized daily incidence curves obtained by the proposed approach for the four countries. Finally, we perform an explainability analysis using Shapley values and feature importance to determine the relevance of each feature and the corresponding contribution for each country and each country/year.

Attractor landscapes: a unifying conceptual model for understanding behaviour change across scales of observation

Models and theories in behaviour change science are not in short supply, but they almost exclusively pertain to a particular facet of behaviour, such as automaticity or reasoned action, or to a single scale of observation such as individuals or communities. We present a highly generalisable conceptual model which is widely used in complex systems research from biology to physics, in an accessible form to behavioural scientists. The proposed model of attractor landscapes can be used to understand human behaviour change on different levels, from individuals to dyads, groups and societies. We use the model as a tool to present neglected ideas in contemporary behaviour change science, such as hysteresis and nonlinearity. The model of attractor landscapes can deepen understanding of well-known features of behaviour change (research), including short-livedness of intervention effects, problematicity of focusing on behavioural initiation while neglecting behavioural maintenance, continuum and stage models of behaviour change understood within a single accommodating framework, and the concept of resilience. We also demonstrate potential methods of analysis and outline avenues for future research.

Engaging a national-scale cohort of smart thermometer users in participatory surveillance

Participatory surveillance systems crowdsource individual reports to rapidly assess population health phenomena. The value of these systems increases when more people join and persistently contribute. We examine the level of and factors associated with engagement in participatory surveillance among a retrospective, national-scale cohort of individuals using smartphone-connected thermometers with a companion app that allows them to report demographic and symptom information. Between January 1, 2020 and October 29, 2022, 1,325,845 participants took 20,617,435 temperature readings, yielding 3,529,377 episodes of consecutive readings. There were 1,735,805 (49.2%) episodes with self-reported symptoms (including reports of no symptoms). Compared to before the pandemic, participants were more likely to report their symptoms during pandemic waves, especially after the winter wave began (September 13, 2020) (OR across pandemic periods range from 3.0 to 4.0). Further, symptoms were more likely to be reported during febrile episodes (OR = 2.6, 95% CI = 2.6-2.6), and for new participants, during their first episode (OR = 2.4, 95% CI = 2.4-2.5). Compared with participants aged 50-65 years old, participants over 65 years were less likely to report their symptoms (OR = 0.3, 95% CI = 0.3-0.3). Participants in a household with both adults and children (OR = 1.6 [1.6-1.7]) were more likely to report symptoms. We find that the use of smart thermometers with companion apps facilitates the collection of data on a large, national scale, and provides real time insight into transmissible disease phenomena. Nearly half of individuals using these devices are willing to report their symptoms after taking their temperature, although participation varies among individuals and over pandemic stages.

Consistent comparison of symptom-based methods for COVID-19 infection detection

BACKGROUND

During the global pandemic crisis, various detection methods of COVID-19-positive cases based on self-reported information were introduced to provide quick diagnosis tools for effectively planning and managing healthcare resources. These methods typically identify positive cases based on a particular combination of symptoms, and they have been evaluated using different datasets.

PURPOSE

This paper presents a comprehensive comparison of various COVID-19 detection methods based on self-reported information using the University of Maryland Global COVID-19 Trends and Impact Survey (UMD-CTIS), a large health surveillance platform, which was launched in partnership with Facebook.

METHODS

Detection methods were implemented to identify COVID-19-positive cases among UMD-CTIS participants reporting at least one symptom and a recent antigen test result (positive or negative) for six countries and two periods. Multiple detection methods were implemented for three different categories: rule-based approaches, logistic regression techniques, and tree-based machine-learning models. These methods were evaluated using different metrics including F1-score, sensitivity, specificity, and precision. An explainability analysis has also been conducted to compare methods.

RESULTS

Fifteen methods were evaluated for six countries and two periods. We identify the best method for each category: rule-based methods (F1-score: 51.48% - 71.11%), logistic regression techniques (F1-score: 39.91% - 71.13%), and tree-based machine learning models (F1-score: 45.07% - 73.72%). According to the explainability analysis, the relevance of the reported symptoms in COVID-19 detection varies between countries and years. However, there are two variables consistently relevant across approaches: stuffy or runny nose, and aches or muscle pain.

CONCLUSIONS

Regarding the categories of detection methods, evaluating detection methods using homogeneous data across countries and years provides a solid and consistent comparison. An explainability analysis of a tree-based machine-learning model can assist in identifying infected individuals specifically based on their relevant symptoms. This study is limited by the self-report nature of data, which cannot replace clinical diagnosis.

COVID-19 activity risk calculator as a gamified public health intervention tool

The Coronavirus disease 2019 (COVID-19) pandemic, caused by the virus severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), has impacted over 200 countries leading to hospitalizations and deaths of millions of people. Public health interventions, such as risk estimators, can reduce the spread of pandemics and epidemics through influencing behavior, which impacts risk of exposure and infection. Current publicly available COVID-19 risk estimation tools have had variable effectiveness during the pandemic due to their dependency on rapidly evolving factors such as community transmission levels and variants. There has also been confusion surrounding certain personal protective strategies such as risk reduction by mask-wearing and vaccination. In order to create a simple easy-to-use tool for estimating different individual risks associated with carrying out daily-life activity, we developed COVID-19 Activity Risk Calculator (CovARC). CovARC is a gamified public health intervention as users can "play with" how different risks associated with COVID-19 can change depending on several different factors when carrying out routine daily activities. Empowering the public to make informed, data-driven decisions about safely engaging in activities may help to reduce COVID-19 levels in the community. In this study, we demonstrate a streamlined, scalable and accurate COVID-19 risk calculation system. Our study also demonstrates the quantitative impact of vaccination and mask-wearing during periods of high case counts. Validation of this impact could inform and support policy decisions regarding case thresholds for mask mandates, and other public health interventions.

Trends and symptoms of SARS-CoV-2 infection: a longitudinal study on an Alpine population representative sample

OBJECTIVES

The continuous monitoring of SARS-CoV-2 infection waves and the emergence of novel pathogens pose a challenge for effective public health surveillance strategies based on diagnostics. Longitudinal population representative studies on incident events and symptoms of SARS-CoV-2 infection are scarce. We aimed at describing the evolution of the COVID-19 pandemic during 2020 and 2021 through regular monitoring of self-reported symptoms in an Alpine community sample.

DESIGN

To this purpose, we designed a longitudinal population representative study, the Cooperative Health Research in South Tyrol COVID-19 study.

PARTICIPANTS AND OUTCOME MEASURES

A sample of 845 participants was retrospectively investigated for active and past infections with swab and blood tests, by August 2020, allowing adjusted cumulative incidence estimation. Of them, 700 participants without previous infection or vaccination were followed up monthly until July 2021 for first-time infection and symptom self-reporting: COVID-19 anamnesis, social contacts, lifestyle and sociodemographic data were assessed remotely through digital questionnaires. Temporal symptom trajectories and infection rates were modelled through longitudinal clustering and dynamic correlation analysis. Negative binomial regression and random forest analysis assessed the relative importance of symptoms.

RESULTS

At baseline, the cumulative incidence of SARS-CoV-2 infection was 1.10% (95% CI 0.51%, 2.10%). Symptom trajectories mimicked both self-reported and confirmed cases of incident infections. Cluster analysis identified two groups of high-frequency and low-frequency symptoms. Symptoms like fever and loss of smell fell in the low-frequency cluster. Symptoms most discriminative of test positivity (loss of smell, fatigue and joint-muscle aches) confirmed prior evidence.

CONCLUSIONS

Regular symptom tracking from population representative samples is an effective screening tool auxiliary to laboratory diagnostics for novel pathogens at critical times, as manifested in this study of COVID-19 patterns. Integrated surveillance systems might benefit from more direct involvement of citizens' active symptom tracking.

Smart Thermometer-Based Participatory Surveillance to Discern the Role of Children in Household Viral Transmission During the COVID-19 Pandemic

Importance

Children's role in spreading virus during the COVID-19 pandemic is yet to be elucidated, and measuring household transmission traditionally requires contact tracing.

Objective

To discern children's role in household viral transmission during the pandemic when enveloped viruses were at historic lows and the predominance of viral illnesses were attributed to COVID-19.

Design, Setting, and Participants

This cohort study of a voluntary US cohort tracked data from participatory surveillance using commercially available thermometers with a companion smartphone app from October 2019 to October 2022. Eligible participants were individuals with temperature measurements in households with multiple members between October 2019 and October 2022 who opted into data sharing.

Main Outcomes and Measures

Proportion of household transmissions with a pediatric index case and changes in transmissions during school breaks were assessed using app and thermometer data.

Results

A total of 862 577 individuals from 320 073 households with multiple participants (462 000 female [53.6%] and 463 368 adults [53.7%]) were included. The number of febrile episodes forecast new COVID-19 cases. Within-household transmission was inferred in 54 506 (15.4%) febrile episodes and increased from the fourth pandemic period, March to July 2021 (3263 of 32 294 [10.1%]) to the Omicron BA.1/BA.2 wave (16 516 of 94 316 [17.5%]; P < .001). Among 38 787 transmissions in 166 170 households with adults and children, a median (IQR) 70.4% (61.4%-77.6%) had a pediatric index case; proportions fluctuated weekly from 36.9% to 84.6%. A pediatric index case was 0.6 to 0.8 times less frequent during typical school breaks. The winter break decrease was from 68.4% (95% CI, 57.1%-77.8%) to 41.7% (95% CI, 34.3%-49.5%) at the end of 2020 (P < .001). At the beginning of 2022, it dropped from 80.3% (95% CI, 75.1%-84.6%) to 54.5% (95% CI, 51.3%-57.7%) (P < .001). During summer breaks, rates dropped from 81.4% (95% CI, 74.0%-87.1%) to 62.5% (95% CI, 56.3%-68.3%) by August 2021 (P = .02) and from 83.8% (95% CI, 79.2%-87.5) to 62.8% (95% CI, 57.1%-68.1%) by July 2022 (P < .001). These patterns persisted over 2 school years.

Conclusions and Relevance

In this cohort study using participatory surveillance to measure within-household transmission at a national scale, we discerned an important role for children in the spread of viral infection within households during the COVID-19 pandemic, heightened when schools were in session, supporting a role for school attendance in COVID-19 spread.

Machine learning clinical decision support systems for surveillance: a case study on pertussis and RSV in children

We tested the performance of a machine learning (ML) algorithm based on signs and symptoms for the diagnosis of RSV infection or pertussis in the first year of age to support clinical decisions and provide timely data for public health surveillance. We used data from a retrospective case series of children in the first year of life investigated for acute respiratory infections in the emergency room from 2015 to 2020. We collected data from PCR laboratory tests for confirming pertussis or RSV infection, clinical symptoms, and routine blood testing results, which were used for the algorithm development. We used a LightGBM model to develop 2 sets of models for predicting pertussis and RSV infection: for each type of infection, we developed one model trained with the combination of clinical symptoms and results from routine blood test (white blood cell count, lymphocyte fraction and C-reactive protein), and one with symptoms only. All analyses were performed using Python 3.7.4 with Shapley values (Shap values) visualization package for predictor visualization. The performance of the models was assessed through confusion matrices. The models were developed on a dataset of 599 children. The recall for the pertussis model combining symptoms and routine laboratory tests was 0.72, and 0.74 with clinical symptoms only. For RSV infection, recall was 0.68 with clinical symptoms and laboratory tests and 0.71 with clinical symptoms only. The F1 score for the pertussis model was 0.72 in both models, and, for RSV infection, it was 0.69 and 0.75. ML models can support the diagnosis and surveillance of infectious diseases such as pertussis or RSV infection in children based on common symptoms and laboratory tests. ML-based clinical decision support systems may be developed in the future in large networks to create accurate tools for clinical support and public health surveillance.

Social Media Sentiment about COVID-19 Vaccination Predicts Vaccine Acceptance among Peruvian Social Media Users the Next Day

Drawing upon theories of risk and decision making, we present a theoretical framework for how the emotional attributes of social media content influence risk behaviors. We apply our framework to understanding how COVID-19 vaccination Twitter posts influence acceptance of the vaccine in Peru, the country with the highest relative number of COVID-19 excess deaths. By employing computational methods, topic modeling, and vector autoregressive time series analysis, we show that the prominence of expressed emotions about COVID-19 vaccination in social media content is associated with the daily percentage of Peruvian social media survey respondents who are vaccine-accepting over 231 days. Our findings show that net (positive) sentiment and trust emotions expressed in tweets about COVID-19 are positively associated with vaccine acceptance among survey respondents one day after the post occurs. This study demonstrates that the emotional attributes of social media content, besides veracity or informational attributes, may influence vaccine acceptance for better or worse based on its valence.

Spatiotemporal Trends in Self-Reported Mask-Wearing Behavior in the United States: Analysis of a Large Cross-sectional Survey

BACKGROUND

Face mask wearing has been identified as an effective strategy to prevent the transmission of SARS-CoV-2, yet mask mandates were never imposed nationally in the United States. This decision resulted in a patchwork of local policies and varying compliance, potentially generating heterogeneities in the local trajectories of COVID-19 in the United States. Although numerous studies have investigated the patterns and predictors of masking behavior nationally, most suffer from survey biases and none have been able to characterize mask wearing at fine spatial scales across the United States through different phases of the pandemic.

OBJECTIVE

Urgently needed is a debiased spatiotemporal characterization of mask-wearing behavior in the United States. This information is critical to further assess the effectiveness of masking, evaluate the drivers of transmission at different time points during the pandemic, and guide future public health decisions through, for example, forecasting disease surges.

METHODS

We analyzed spatiotemporal masking patterns in over 8 million behavioral survey responses from across the United States, starting in September 2020 through May 2021. We adjusted for sample size and representation using binomial regression models and survey raking, respectively, to produce county-level monthly estimates of masking behavior. We additionally debiased self-reported masking estimates using bias measures derived by comparing vaccination data from the same survey to official records at the county level. Lastly, we evaluated whether individuals' perceptions of their social environment can serve as a less biased form of behavioral surveillance than self-reported data.

RESULTS

We found that county-level masking behavior was spatially heterogeneous along an urban-rural gradient, with mask wearing peaking in winter 2021 and declining sharply through May 2021. Our results identified regions where targeted public health efforts could have been most effective and suggest that individuals' frequency of mask wearing may be influenced by national guidance and disease prevalence. We validated our bias correction approach by comparing debiased self-reported mask-wearing estimates with community-reported estimates, after addressing issues of a small sample size and representation. Self-reported behavior estimates were especially prone to social desirability and nonresponse biases, and our findings demonstrated that these biases can be reduced if individuals are asked to report on community rather than self behaviors.

CONCLUSIONS

Our work highlights the importance of characterizing public health behaviors at fine spatiotemporal scales to capture heterogeneities that may drive outbreak trajectories. Our findings also emphasize the need for a standardized approach to incorporating behavioral big data into public health response efforts. Even large surveys are prone to bias; thus, we advocate for a social sensing approach to behavioral surveillance to enable more accurate estimates of health behaviors. Finally, we invite the public health and behavioral research communities to use our publicly available estimates to consider how bias-corrected behavioral estimates may improve our understanding of protective behaviors during crises and their impact on disease dynamics.

Trajectories of psychological distress over multiple COVID-19 lockdowns in Australia

The impact of the global COVID-19 pandemic, including the indirect effect of policy responses, on psychological distress has been the subject of much research. However, there has been little consideration of how the prevalence of psychological distress changed with the duration and repetition of lockdowns, or the rate of resolution of psychological distress once lockdowns ended. This study describes the trajectories of psychological distress over multiple lockdowns during the first two years of the pandemic across five Australian states for the period May 2020 to December 2021 and examines whether psychological distress trajectories varied as a function of time spent in lockdown, or time since lockdown ended.

A total of N = 574,306 Australian adults completed Facebook surveys over 611 days (on average 940 participants per day). Trajectories of psychological distress (depression and anxiety) were regressed on lockdown duration and time since lockdown ended. Random effects reflecting the duration of each lockdown were included to account for varying effects on psychological distress associated with lockdown length.

The prevalence of psychological distress was higher during periods of lockdown, more so for longer lockdowns relative to shorter lockdowns. Psychological distress increased rapidly over the first ten weeks of lockdowns spanning at least twelve weeks, though less rapidly for short lockdowns of three weeks or less. Psychological distress levels tended to stabilise, or even decrease, after ten consecutive weeks of lockdown. After lockdown restrictions were lifted, psychological distress rapidly subsided but did not return to pre-lockdown levels within four weeks, although continued to decline afterwards.

In Australia short lockdowns of pre-announced durations were associated with slower rises in psychological distress. Lockdowns may have left some temporary residual population effect, but we cannot discern whether this reflects longer term trends in increasing psychological distress. However, the findings do re-emphasise the resilience of individuals to major life stressors.

Global state of education-related inequality in COVID-19 vaccine coverage, structural barriers, vaccine hesitancy, and vaccine refusal: findings from the Global COVID-19 Trends and Impact Survey

BACKGROUND

COVID-19 vaccine coverage and experiences of structural and attitudinal barriers to vaccination vary across populations. Education-related inequality in COVID-19 vaccine coverage and barriers within and between countries can provide insight into the hypothesised role of education as a correlate of vaccine access and acceptability. We aimed to characterise patterns of within-country education-related inequality in COVID-19 vaccine indicators across 90 countries.

METHODS

This study used data from the University of Maryland Social Data Science Center Global COVID-19 Trends and Impact Survey. Data from 90 countries (more than 14 million participants aged 18 years and older) were included in our analyses. We assessed education-related inequalities globally, across country-income groupings, and nationally for four indicators (self-reported receipt of COVID-19 vaccine, structural barriers to vaccination, vaccine hesitancy, and vaccine refusal) for the study period June 1-Dec 31, 2021. We calculated an absolute summary measure of inequality to assess the latest situation of inequality and time trends and explored the association between government vaccine availability policies and education-related inequality.

FINDINGS

Nearly all countries had higher self-reported receipt of a COVID-19 vaccine among the most educated respondents than the least educated respondents. Education-related inequality in structural barriers, vaccine hesitancy, and vaccine refusal varied across countries, and was most pronounced in high-income countries, overall. Low-income and lower-middle-income countries reported widespread experiences of structural barriers and high levels of vaccine hesitancy alongside low levels of education-related inequality. Globally, vaccine hesitancy in unvaccinated people was higher among those with lower education and vaccine refusal was higher among those with higher education, especially in high-income countries. Over the study period, education-related inequalities in self-reported receipt of a COVID-19 vaccine declined, globally and across all country income groupings. Government policies expanding vaccine availability were associated with lower education-related inequality in self-reported receipt of vaccine.

INTERPRETATION

This study serves as a baseline for continued inequality monitoring and could help to inform targeted actions for the equitable uptake of vaccines.

FUNDING

Gavi, the Vaccine Alliance.

Digital public health interventions at scale: The impact of social media advertising on beliefs and outcomes related to COVID vaccines

Public health organizations increasingly use social media advertising campaigns in pursuit of public health goals. In this paper, we evaluate the impact of about $40 million of social media advertisements that were run and experimentally tested on Facebook and Instagram, aimed at increasing COVID-19 vaccination rates in the first year of the vaccine roll-out. The 819 randomized experiments in our sample were run by 174 different public health organizations and collectively reached 2.1 billion individuals in 15 languages. We find that these campaigns are, on average, effective at influencing self-reported beliefs-shifting opinions close to 1% at baseline with a cost per influenced person of about $3.41. Combining this result with an estimate of the relationship between survey outcomes and vaccination rates derived from observational data yields an estimated cost per additional vaccination of about $5.68. There is further evidence that campaigns are especially effective at influencing users' knowledge of how to get vaccines. Our results represent, to the best of our knowledge, the largest set of online public health interventions analyzed to date.

Using survey data to estimate the impact of the omicron variant on vaccine efficacy against COVID-19 infection

Symptoms-based detection of SARS-CoV-2 infection is not a substitute for precise diagnostic tests but can provide insight into the likely level of infection in a given population. This study uses symptoms data collected in the Global COVID-19 Trends and Impact Surveys (UMD Global CTIS), and data on variants sequencing from GISAID. This work, conducted in January of 2022 during the emergence of the Omicron variant (subvariant BA.1), aims to improve the quality of infection detection from the available symptoms and to use the resulting estimates of infection levels to assess the changes in vaccine efficacy during a change of dominant variant; from the Delta dominant to the Omicron dominant period. Our approach produced a new symptoms-based classifier, Random Forest, that was compared to a ground-truth subset of cases with known diagnostic test status. This classifier was compared with other competing classifiers and shown to exhibit an increased performance with respect to the ground-truth data. Using the Random Forest classifier, and knowing the vaccination status of the subjects, we then proceeded to analyse the evolution of vaccine efficacy towards infection during different periods, geographies and dominant variants. In South Africa, where the first significant wave of Omicron occurred, a significant reduction of vaccine efficacy is observed from August-September 2021 to December 2021. For instance, the efficacy drops from 0.81 to 0.30 for those vaccinated with 2 doses (of Pfizer/BioNTech), and from 0.51 to 0.09 for those vaccinated with one dose (of Pfizer/BioNTech or Johnson & Johnson). We also extended the study to other countries in which Omicron has been detected, comparing the situation in October 2021 (before Omicron) with that of December 2021. While the reduction measured is smaller than in South Africa, we still found, for instance, an average drop in vaccine efficacy from 0.53 to 0.45 among those vaccinated with two doses. Moreover, we found a significant negative (Pearson) correlation of around - 0.6 between the measured prevalence of Omicron in several countries and the vaccine efficacy in those same countries. This prediction, in January of 2022, of the decreased vaccine efficacy towards Omicron is in line with the subsequent increase of Omicron infections in the first half of 2022.

Spatiotemporal trends in self-reported mask-wearing behavior in the United States: Analysis of a large cross-sectional survey

Background

Face mask-wearing has been identified as an effective strategy to prevent transmission of SARS-CoV-2, yet mask mandates were never imposed nationally in the United States. This decision resulted in a patchwork of local policies and varying compliance potentially generating heterogeneities in the local trajectories of COVID-19 in the U.S. While numerous studies have investigated patterns and predictors of masking behavior nationally, most suffer from survey biases and none have been able to characterize mask-wearing at fine spatial scales across the U.S. through different phases of the pandemic.

Objective

Urgently needed is a debiased spatiotemporal characterization of mask-wearing behavior in the U.S. This information is critical to further assess the effectiveness of masking, evaluate drivers of transmission at different time points during the pandemic, and guide future public health decisions through, for example, forecasting disease surges.

Methods

We analyze spatiotemporal masking patterns in over eight million behavioral survey responses from across the United States starting in September 2020 through May 2021. We adjust for sample size and representation using binomial regression models and survey raking, respectively, to produce county-level monthly estimates of masking behavior. We additionally debias self-reported masking estimates using bias measures derived by comparing vaccination data from the same survey to official records at the county-level. Lastly, we evaluate whether individuals' perceptions of their social environment can serve as a less biased form of behavioral surveillance than self-reported data.

Results

We find that county-level masking behavior is spatially heterogeneous along an urban-rural gradient, with mask-wearing peaking in winter 2021 and declining sharply through May 2021. Our results identify regions where targeted public health efforts could have been most effective and suggest that individuals' frequency of mask-wearing may be influenced by national guidance and disease prevalence. We validate our bias-correction approach by comparing debiased self-reported mask-wearing estimates with community-reported estimates, after addressing issues of small sample size and representation. Self-reported behavior estimates are especially prone to social desirability and non-response biases and our findings demonstrate that these biases can be reduced if individuals are asked to report on community rather than self behaviors.

Conclusions

Our work highlights the importance of characterizing public health behaviors at fine spatiotemporal scales to capture heterogeneities that may drive outbreak trajectories. Our findings also emphasize the need for a standardized approach to incorporating behavioral big data into public health response efforts. Even large surveys are prone to bias; thus, we advocate for a social sensing approach to behavioral surveillance to enable more accurate estimates of health behaviors. Finally, we invite the public health and behavioral research communities to use our publicly available estimates to consider how bias-corrected behavioral estimates may improve our understanding of protective behaviors during crises and their impact on disease dynamics.

Unaffordability of COVID-19 tests: assessing age-related inequalities in 83 countries

BACKGROUND

Diagnostic testing for SARS-CoV-2 is critical to manage the pandemic and its different waves. The requirement to pay out-of-pocket (OOP) for testing potentially represents both a financial barrier to access and, for those who manage to make the payment, a source of financial hardship, as they may be forced to reduce spending on other necessities. This study aims to assess age-related inequality in affordability of COVID-19 tests.

METHODS

Daily data from the Global COVID-19 Trends and Impact Survey among adult respondents across 83 countries from July 2020 to April 2021 was used to monitor age-related inequalities across three indicators: the experiences of, first, reducing spending on necessities because of paying OOP for testing, second, facing financial barriers to get tested (from January to April 2021), and third, having anxiety related to household finance in the future. Logistic regressions were used to assess the association of age with each of these.

RESULTS

Among the population ever tested, the adjusted odds of reducing spending on necessities due to the cost of the test decreased non-linearly with age from 2.3 [CI95%: 2.1-2.5] among ages 18-24 to 1.6 [CI95%: 1.5-1.8] among ages 45-54. Among the population never tested, odds of facing any type of barrier to testing were highest among the youngest age group 2.5 [CI95%:2.4-2.5] and decreased with age. Finally, among those reporting reducing spending on necessities, the odds of reporting anxiety about their future finances decreased non-linearly with age, with the two younger groups being 2.4-2.5 times more anxious than the oldest age group. Among those reporting financial barriers due to COVID-19 test cost, there was an inverse U-shape relationship.

CONCLUSIONS

COVID-19 testing was associated with a reduction in spending on necessities at varying levels by age. Younger people were more likely to face financial barrier to get tested. Both negative outcomes generated anxiety across all age-groups but more frequently among the younger ones. To reduce age-related inequalities in the affordability of COVID-19 test, these findings support calls for exempting everyone from paying OOP for testing and, removing other type of barriers than financial ones.

Emulation of epidemics via Bluetooth-based virtual safe virus spread: Experimental setup, software, and data

We describe an experimental setup and a currently running experiment for evaluating how physical interactions over time and between individuals affect the spread of epidemics. Our experiment involves the voluntary use of the Safe Blues Android app by participants at The University of Auckland (UoA) City Campus in New Zealand. The app spreads multiple virtual safe virus strands via Bluetooth depending on the physical proximity of the subjects. The evolution of the virtual epidemics is recorded as they spread through the population. The data is presented as a real-time (and historical) dashboard. A simulation model is applied to calibrate strand parameters. Participants' locations are not recorded, but participants are rewarded based on the duration of participation within a geofenced area, and aggregate participation numbers serve as part of the data. The 2021 experimental data is available as an open-source anonymized dataset, and once the experiment is complete, the remaining data will be made available. This paper outlines the experimental setup, software, subject-recruitment practices, ethical considerations, and dataset description. The paper also highlights current experimental results in view of the lockdown that started in New Zealand at 23:59 on August 17, 2021. The experiment was initially planned in the New Zealand environment, expected to be free of COVID and lockdowns after 2020. However, a COVID Delta strain lockdown shuffled the cards and the experiment is currently extended into 2022.

Education-related inequalities in beliefs and behaviors pertaining to COVID-19 non-pharmaceutical interventions

BACKGROUND

The coronavirus pandemic has exposed existing social inequalities in relation to disease preventive behaviors, risk of exposure, testing and healthcare access, and consequences as a result of illness and containment measures across different population groups. However, due to a lack of data, to date there has been limited evidence of the extent of such within-country inequalities globally.

METHODS

We examined education-related inequalities in four COVID-19 prevention and testing indicators within 90 countries, using data from the University of Maryland Social Data Science Center Global COVID-19 Trends and Impact Survey, in partnership with Facebook, over the period 1 June 2021 to 31 December 2021. The overall level of education-related inequalities, as well as how they differ across country income groups and how they have changed over time were analyzed using the Slope Index of Inequality (SII) and the Relative Index of Inequality (RII). We also assessed whether these education-related inequalities were associated with government policies and responses.

RESULTS

Education-related inequalities in beliefs, mask wearing, social distancing and testing varied across the study countries. Mask wearing and beliefs in the effectiveness of social distancing and mask wearing were overall more common among people with a higher level of education. Even after controlling for other sociodemographic and health-related factors, social distancing practice was higher among the most educated in low/lower middle income countries, but was higher overall among the least educated in high income countries. Overall there were low education-related inequalities in COVID-19 testing, though there was variation across countries.

CONCLUSIONS

The study highlights important within-country education-related differences in COVID-19 beliefs, preventive behaviors and testing, as well as differing trends across country income groups. This has implications for considering and targeting specific population groups when designing public health interventions and messaging during the COVID-19 pandemic and future health emergencies.

Real-Time Analysis of Predictors of COVID-19 Infection Spread in Countries in the European Union Through a New Tool

Objectives: Real-time data analysis during a pandemic is crucial. This paper aims to introduce a novel interactive tool called Covid-Predictor-Tracker using several sources of COVID-19 data, which allows examining developments over time and across countries. Exemplified here by investigating relative effects of vaccination to non-pharmaceutical interventions on COVID-19 spread.

Methods: We combine >100 indicators from the Global COVID-19 Trends and Impact Survey, Johns Hopkins University, Our World in Data, European Centre for Disease Prevention and Control, National Centers for Environmental Information, and Eurostat using random forests, hierarchical clustering, and rank correlation to predict COVID-19 cases.

Results: Between 2/2020 and 1/2022, we found among the non-pharmaceutical interventions “mask usage” to have strong effects after the percentage of people vaccinated at least once, followed by country-specific measures such as lock-downs. Countries with similar characteristics share ranks of infection predictors. Gender and age distribution, healthcare expenditures and cultural participation interact with restriction measures.

Conclusion: Including time-aware machine learning models in COVID-19 infection dashboards allows to disentangle and rank predictors of COVID-19 cases per country to support policy evaluation. Our open-source tool can be updated daily with continuous data streams, and expanded as the pandemic evolves.

Adaptive resources allocation CUSUM for binomial count data monitoring with application to COVID-19 hotspot detection

In this paper, we present an efficient statistical method (denoted as 'Adaptive Resources Allocation CUSUM') to robustly and efficiently detect the hotspot with limited sampling resources. Our main idea is to combine the multi-arm bandit (MAB) and change-point detection methods to balance the exploration and exploitation of resource allocation for hotspot detection. Further, a Bayesian weighted update is used to update the posterior distribution of the infection rate. Then, the upper confidence bound (UCB) is used for resource allocation and planning. Finally, CUSUM monitoring statistics to detect the change point as well as the change location. For performance evaluation, we compare the performance of the proposed method with several benchmark methods in the literature and showed the proposed algorithm is able to achieve a lower detection delay and higher detection precision. Finally, this method is applied to hotspot detection in a real case study of county-level daily positive COVID-19 cases in Washington State WA) and demonstrates the effectiveness with very limited distributed samples.

Relationship Between COVID-19 Threat Beliefs and Individual Differences in Demographics, Personality, and Related Beliefs

Individual differences in demographics, personality, and other related beliefs are associated with coronavirus disease 2019 (COVID-19) threat beliefs. However, the relative contributions of these different types of individual differences to COVID-19 threat beliefs are not known. In this study, a total of 1,700 participants in Croatia (68% female; age 18-86 years) completed a survey that included questions about COVID-19 risks, questions about related beliefs including vaccination beliefs, trust in the health system, trust in scientists, and trust in the political system, the HEXACO 60 personality inventory, as well as demographic questions about gender, age, chronic diseases, and region. We used hierarchical regression analyses to examine the proportion of variance explained by demographics, personality, and other related beliefs. All three types of individual differences explained a part of the variance of COVID-19 threat beliefs, with related beliefs explaining the largest part. Personality facets explained a slightly larger amount of variance than personality factors. These results have implications for communication about COVID-19.

Epidemiology of SARS-CoV-2 Infection in Italy Using Real-World Data: Methodology and Cohort Description of the Second Phase of Web-Based EPICOVID19 Study

Digital technologies have been extensively employed in response to the SARS-CoV-2 pandemic worldwide. This study describes the methodology of the two-phase internet-based EPICOVID19 survey, and the characteristics of the adult volunteer respondents who lived in Italy during the first (April-May 2020) and the second wave (January-February 2021) of the epidemic. Validated scales and ad hoc questionnaires were used to collect socio-demographic, medical and behavioural characteristics, as well as information on COVID-19. Among those who provided email addresses during phase I (105,355), 41,473 participated in phase II (mean age 50.7 years ± 13.5 SD, 60.6% females). After a median follow-up of ten months, 52.8% had undergone nasopharyngeal swab (NPS) testing and 13.2% had a positive result. More than 40% had undergone serological test (ST) and 11.9% were positive. Out of the 2073 participants with at least one positive ST, 72.8% had only negative results from NPS or never performed it. These results indicate that a large fraction of individuals remained undiagnosed, possibly contributing to the spread of the virus in the community. Participatory online surveys offer a unique opportunity to collect relevant data at individual level from large samples during confinement.