Reappraising the utility of Google Flu Trends

Global infectious disease early warning models: An updated review and lessons from the COVID-19 pandemic

An early warning model for infectious diseases is a crucial tool for timely monitoring, prevention, and control of disease outbreaks. The integration of diverse multi-source data using big data and artificial intelligence techniques has emerged as a key approach in advancing these early warning models. This paper presents a comprehensive review of widely utilized early warning models for infectious diseases around the globe. Unlike previous review studies, this review encompasses newly developed approaches such as the combined model and Hawkes model after the COVID-19 pandemic, providing a thorough evaluation of their current application status and development prospects for the first time. These models not only rely on conventional surveillance data but also incorporate information from various sources. We aim to provide valuable insights for enhancing global infectious disease surveillance and early warning systems, as well as informing future research in this field, by summarizing the underlying modeling concepts, algorithms, and application scenarios of each model.

Investigating Google Trends to forecast acute febrile illness outbreaks in North India reported through the Integrated Disease Surveillance Program

BACKGROUND

Acute Febrile Illness (AFI) like Malaria, Dengue, Chikungunya, and Enteric fever still remain the most common cause of seeking healthcare in low-middle-income countries and need to be constantly monitored for any impending outbreak. Digital epidemiology promises to assist traditional health surveillance. The health data (including AFI) collated by Google using specialised platforms like Google Trends (GT) is known to correlate with actual disease trends. The present study thus aims to assess the potential of GT to support routine surveillance system and forecast AFI outbreaks reported through the Indian Integrated Disease Surveillance Programme (IDSP).

METHODS

We utilised Haryana's IDSP portal to retrieve the weekly data of the most commonly reported infectious diseases causing AFI between 2011 and 2020. Internet search trends were downloaded using GT. Descriptive statistics estimated the burden of the AFI and Bland-Altman's plot depicted statistical agreement between the two. We adopted the Box-Jenkins approach to attain the final SARIMA model and explain the time-dependent weekly incidence of AFI.

RESULTS

The time series plot of the reported AFI displayed trends. Martin- Bland plots depicted acceptable agreement between two datasets for all Chikungunya and Dengue. Among the models evaluated, the Malaria model [SARIMA(1,1,1)(1,1,1)] demonstrated the best performance with a balanced fit and reasonable accuracy, while the Enteric Fever model [SARIMA(0,1,0)(1,1,1)] exhibited low prediction error but weak seasonal significance. In contrast, the Dengue [SARIMA(1,1,0)(1,1,0)] and Chikungunya [ARIMA(1,0,0)(0,0,0)] models had high forecast errors, limiting their predictive reliability. Overall, GT supplemented the prediction performance of the SARIMA models with adjusted R2 of 46%, 50%, 50%, and 52% compared to the original 43%, 49%, 20%, and 48%.

CONCLUSIONS

Our study observed modest improvements in GT-based SARIMA forecasting models compared to routine IDSP mechanisms for predicting AFI outbreaks in Haryana, highlighting the potential for further enhancement. As more granular GT data becomes available, its integration with traditional surveillance systems could significantly enhance forecasting accuracy for AFI and other infectious disease outbreaks. At no additional cost to the health system, GT can serve as a valuable, real-time digital epidemiology tool, strengthening public health preparedness and enabling timely interventions for the early containment of emerging diseases.

Drawing on the Development Experiences of Infectious Disease Surveillance Systems Around the World

High-quality infectious disease surveillance systems are foundational to infectious disease prevention and control. Current major infectious disease surveillance systems globally can be categorized as either indicator-based, which are more specific, or event-based, which are more timely. Modern surveillance systems commonly utilize multi-source data, strengthened information sharing, advanced technology, and improved early warning accuracy and sensitivity. International experience may provide valuable insights for China. China's existing infectious disease surveillance systems require urgent enhancements to monitor emerging infectious diseases and improve the integration and learning capabilities of early warning models. Methods such as establishing multi-stage surveillance systems, promoting cross-sectoral and cross-provincial data sharing, applying advanced technologies like artificial intelligence, and cultivating professional talent should be adopted to enhance the development of intelligent and multipoint-triggered infectious disease surveillance systems in China.

Spatio-temporal evaluation of social media as a tool for livestock disease surveillance

Recent outbreaks of Avian Influenza across Europe have highlighted the potential for syndromic surveillance systems that consider other modes of data, namely social media. This study investigates the feasibility of using social media, primarily Twitter, to monitor illness outbreaks such as avian flu. Using temporal, geographical, and correlation analyses, we investigated the association between avian influenza tweets and officially verified cases in the United Kingdom in 2021 and 2022. Pearson correlation coefficient, bivariate Moran's I analysis and time series analysis, were among the methodologies used. The findings show a weak, statistically insignificant relationship between the number of tweets and confirmed cases in a temporal context, implying that relying simply on social media data for surveillance may be insufficient. The spatial analysis provided insights into the overlaps between confirmed cases and tweet locations, shedding light on regionally targeted interventions during outbreaks. Although social media can be useful for understanding public sentiment and concerns during outbreaks, it must be combined with traditional surveillance methods and official data sources for a more accurate and comprehensive approach. Improved data mining techniques and real-time analysis can improve outbreak detection and response even further. This study underscores the need of having a strong surveillance system in place to properly monitor and manage disease outbreaks and protect public health.

The evolution of the COVID-19 pandemic through the lens of google searches

Real-time data is essential for policymakers to adapt to a rapidly evolving situation like the COVID-19 pandemic. Using data from 221 countries and territories, we demonstrate the capacity of Google search data to anticipate reported COVID-19 cases and understand how containment policies are associated with changes in socioeconomic indicators. First, search interest in COVID-specific symptoms such as "loss of smell" strongly correlated with cases initially, but the association diminished as COVID-19 evolved; general terms such as "COVID symptoms" remained strongly associated with cases. Moreover, trends in search interest preceded trends in reported cases, particularly in the first year of the pandemic. Second, countries with more restrictive containment policies experienced greater search interest in unemployment and mental health terms after policies were implemented, indicating socio-economic externalities. Higher-income countries experienced a larger increase in searches related to unemployment and a larger reduction in relationship and family planning keywords relative to lower-income countries. The results demonstrate that real-time search interest can be a valuable tool to inform policies across multiple stages of the pandemic.

Population Interest in Information on Obesity, Nutrition, and Occupational Health and Its Relationship with the Prevalence of Obesity: An Infodemiological Study

OBJECTIVE

To identify and analyze population interest in obesity, nutrition, and occupational health and safety and its relationship with the worldwide prevalence of obesity through information search trends.

METHOD

In this ecological study, data were obtained through online access to Google Trends using the topics "obesity", "nutrition", and "occupational health and safety". Obesity data were obtained from the World Health Organization (WHO) website for crude adult prevalence and estimates by region. The variables studied were relative search volume (RSV), temporal evolution, milestone, trend, and seasonality. The temporal evolution of the search trends was examined by regression analysis (R2). To assess the relationship between quantitative variables, the Spearman correlation coefficient (Rho) was used. Seasonality was verified using the augmented Dickey-Fuller (ADF) test.

RESULTS

The RSV trends were as follows: obesity (R2 = 0.04, p = 0.004); nutrition (R2 = 0.42, p < 0.001); and occupational health and safety (R2 = 0.45, p < 0.001). The analysis of seasonality showed the absence of a temporal pattern (p < 0.05 for all terms). The associations between world obesity prevalence (WOP) and the different RSVs were as follows: WOP versus RSV obesity, Rho = -0.79, p = 0.003; WOP versus RSV nutrition, Rho = 0.57, p = 0.044; and WOP versus RSV occupational health and safety, Rho = -0.93, p = 0.001.

CONCLUSIONS

Population interest in obesity continues to be a trend in countries with the highest prevalence, although there are clear signs popularity loss in favor of searches focused on possible solutions and treatments, with a notable increase in searches related to nutrition and diet. Despite the fact that most people spend a large part of their time in the workplace and that interventions including various strategies have been shown to be useful in combating overweight and obesity, there has been a decrease in the population's interest in information related to obesity in the workplace. This information can be used as a guide for public health approaches to obesity and its relationship to nutrition and a healthy diet, approaches that are of equal utility and applicability in occupational health.

Web-based surveillance of respiratory infection outbreaks: retrospective analysis of Italian COVID-19 epidemic waves using Google Trends

Introduction

Large-scale diagnostic testing has been proven insufficient to promptly monitor the spread of the Coronavirus disease 2019. Electronic resources may provide better insight into the early detection of epidemics. We aimed to retrospectively explore whether the Google search volume has been useful in detecting Severe Acute Respiratory Syndrome Coronavirus outbreaks early compared to the swab-based surveillance system.

Methods

The Google Trends website was used by applying the research to three Italian regions (Lombardy, Marche, and Sicily), covering 16 million Italian citizens. An autoregressive-moving-average model was fitted, and residual charts were plotted to detect outliers in weekly searches of five keywords. Signals that occurred during periods labelled as free from epidemics were used to measure Positive Predictive Values and False Negative Rates in anticipating the epidemic wave occurrence.

Results

Signals from "fever," "cough," and "sore throat" showed better performance than those from "loss of smell" and "loss of taste." More than 80% of true epidemic waves were detected early by the occurrence of at least an outlier signal in Lombardy, although this implies a 20% false alarm signals. Performance was poorer for Sicily and Marche.

Conclusion

Monitoring the volume of Google searches can be a valuable tool for early detection of respiratory infectious disease outbreaks, particularly in areas with high access to home internet. The inclusion of web-based syndromic keywords is promising as it could facilitate the containment of COVID-19 and perhaps other unknown infectious diseases in the future.

County-level estimates of suicide mortality in the USA: a modelling study

BACKGROUND

Suicide is one of the leading causes of death in the USA and population risk prediction models can inform decisions on the type, location, and timing of public health interventions. We aimed to develop a prediction model to estimate county-level suicide risk in the USA using population characteristics.

METHODS

We obtained data on all deaths by suicide reported to the National Vital Statistics System between Jan 1, 2005, and Dec 31, 2019, and age, sex, race, and county of residence of the decedents were extracted to calculate baseline risk. We also obtained county-level annual measures of socioeconomic predictors of suicide risk (unemployment, weekly wage, poverty prevalence, median household income, and population density) and state-level prevalence of major depressive disorder and firearm ownership from US public sources. We applied conditional autoregressive models, which account for spatiotemporal autocorrelation in response and predictors, to estimate county-level suicide risk.

FINDINGS

Estimates derived from conditional autoregressive models were more accurate than from models not adjusted for spatiotemporal autocorrelation. Inclusion of suicide risk and protective covariates further reduced errors. Suicide risk was estimated to increase with each SD increase in firearm ownership (2·8% [95% credible interval (CrI) 1·8 to 3·9]), prevalence of major depressive episode (1·0% [0·4 to 1·5]), and unemployment rate (2·8% [1·9 to 3·8]). Conversely, risk was estimated to decrease by 4·3% (-5·1 to -3·2) for each SD increase in median household income and by 4·3% (-5·8 to -2·5) for each SD increase in population density. An increase in the heterogeneity in county-specific suicide risk was also observed during the study period.

INTERPRETATION

Area-level characteristics and the conditional autoregressive models can estimate population-level suicide risk. Availability of near real-time situational data are necessary for the translation of these models into a surveillance setting. Monitoring changes in population-level risk of suicide could help public health agencies select and deploy targeted interventions quickly.

FUNDING

US National Institute of Mental Health.

Impact of COVID-19 on search in an organisation

COVID-19 has created unprecedented organisational challenges, yet no study has examined the impact on information search. A case study in a knowledge-intensive organisation was undertaken on 2.5 million search queries during the pandemic. A surge of unique users and COVID-19 search queries in March 2020 may equate to 'peak uncertainty and activity', demonstrating the importance of corporate search engines in times of crisis. Search volumes dropped 24% after lockdowns; an 'L-shaped' recovery may be a surrogate for business activity. COVID-19 search queries transitioned from awareness, to impact, strategy, response and ways of working that may influence future search design. Low click through rates imply some information needs were not met and searches on mental health increased. In extreme situations (i.e. a pandemic), companies may need to move faster, monitoring and exploiting their enterprise search logs in real time as these reflect uncertainty and anxiety that may exist in the enterprise.

How do El Niño Southern Oscillation (ENSO) and local meteorological factors affect the incidence of seasonal influenza in New York state

Background

Research is lacking in examining how multiple climate factors affect the incidence of seasonal influenza. We investigated the associations between El Niño Southern Oscillation (ENSO), meteorological factors, and influenza incidence in New York State, United States.

Method

We collected emergency department visit data for influenza from the New York State Department of Health. ENSO index was obtained from the National Oceanic and Atmospheric Administration. Meteorological factors, Google Flu Search Index (GFI), and Influenza-like illness (ILI) data in New York State were also collected. Wavelet analysis was used to quantitatively estimate the coherence and phase difference of ENSO, temperature, precipitation, relative humidity, and absolute humidity with emergency department visits of influenza in New York State. Generalized additive models (GAM) were employed to examine the exposure-response relationships between ENSO, weather, and influenza. GFI and ILI data were used to simulate synchronous influenza visits.

Results

The influenza epidemic in New York State had multiple periodic and was primarily on the 1-year scale. The incidence of influenza closely followed the low ENSO index by an average of two months, and the lag period of ENSO on influenza was shorter during 2015-2018. Low temperature in the previous 2 weeks and low absolute humidity in the prior week were positively associated with influenza incidence in New York State. We found an l-shaped association between ENSO index and influenza, a parabolic relationship between temperature in the previous two weeks and influenza, and a linear negative association between absolute humidity in the previous week and influenza. The simulation models including GFI and ILI had higher accuracy for influenza visit estimation.

Conclusions

Low ENSO index, low temperature, and low absolute humidity may drive the influenza epidemics in New York State. The findings can help us deepen the understanding of the climate-influenza association, and help to develop an influenza forecasting model.

Predicting health crises from early warning signs in patient medical records

The COVID-19 global pandemic has caused unprecedented worldwide changes in healthcare delivery. While containment and mitigation approaches have been intensified, the progressive increase in the number of cases has overwhelmed health systems globally, highlighting the need for anticipation and prediction to be the basis of an efficient response system. This study demonstrates the role of population health metrics as early warning signs of future health crises. We retrospectively collected data from the emergency department of a large academic hospital in the northeastern United States from 01/01/2019 to 08/07/2021. A total of 377,694 patient records and 303 features were included for analysis. Departing from a multivariate artificial intelligence (AI) model initially developed to predict the risk of high-flow oxygen therapy or mechanical ventilation requirement during the COVID-19 pandemic, a total of 19 original variables and eight engineered features showing to be most predictive of the outcome were selected for further analysis. The temporal trends of the selected variables before and during the pandemic were characterized to determine their potential roles as early warning signs of future health crises. Temporal analysis of the individual variables included in the high-flow oxygen model showed that at a population level, the respiratory rate, temperature, low oxygen saturation, number of diagnoses during the first encounter, heart rate, BMI, age, sex, and neutrophil percentage demonstrated observable and traceable changes eight weeks before the first COVID-19 public health emergency declaration. Additionally, the engineered rule-based features built from the original variables also exhibited a pre-pandemic surge that preceded the first pandemic wave in spring 2020. Our findings suggest that the changes in routine population health metrics may serve as early warnings of future crises. This justifies the development of patient health surveillance systems, that can continuously monitor population health features, and alarm of new approaching public health crises before they become devastating.

Using search trends to analyze web-based users' behavior profiles connected with COVID-19 in mainland China: infodemiology study based on hot words and Baidu Index

Background

Mainland China, the world's most populous region, experienced a large-scale coronavirus disease 2019 (COVID-19) outbreak in 2020 and 2021, respectively. Existing infodemiology studies have primarily concentrated on the prospective surveillance of confirmed cases or symptoms which met the criterion for investigators; nevertheless, the actual impact regarding COVID-19 on the public and subsequent attitudes of different groups towards the COVID-19 epidemic were neglected.

Methods

This study aimed to examine the public web-based search trends and behavior patterns related to COVID-19 outbreaks in mainland China by using hot words and Baidu Index (BI). The initial hot words (the high-frequency words on the Internet) and the epidemic data (2019/12/01-2021/11/30) were mined from infodemiology platforms. The final hot words table was established by two-rounds of hot words screening and double-level hot words classification. Temporal distribution and demographic portraits of COVID-19 were queried by search trends service supplied from BI to perform the correlation analysis. Further, we used the parameter estimation to quantitatively forecast the geographical distribution of COVID-19 in the future.

Results

The final English-Chinese bilingual table was established including six domains and 32 subordinate hot words. According to the temporal distribution of domains and subordinate hot words in 2020 and 2021, the peaks of searching subordinate hot words and COVID-19 outbreak periods had significant temporal correlation and the subordinate hot words in COVID-19 Related and Territory domains were reliable for COVID-19 surveillance. Gender distribution results showed that Territory domain (the male proportion: 67.69%; standard deviation (SD): 5.88%) and Symptoms/Symptom and Public Health (the female proportion: 57.95%, 56.61%; SD: 0, 9.06%) domains were searched more by male and female groups respectively. The results of age distribution of hot words showed that people aged 20-50 (middle-aged people) had a higher online search intensity, and the group of 20-29, 30-39 years old focused more on Media and Symptoms/Symptom (proportion: 45.43%, 51.66%; SD: 15.37%, 16.59%) domains respectively. Finally, based on frequency rankings of searching hot words and confirmed cases in Mainland China, the epidemic situation of provinces and Chinese administrative divisions were divided into 5 levels of early-warning regions. Central, East and South China regions would be impacted again by the COVID-19 in the future.

How reported outbreak data can shape individual behavior in a social world

Agencies reporting on disease outbreaks face many choices about what to report and the scale of its dissemination. Reporting impacts an epidemic by influencing individual decisions directly, and the social network in which they are made. We simulated a dynamic multiplex network model-with coupled infection and communication layers-to examine behavioral impacts from the nature and scale of epidemiological information reporting. We explored how adherence to protective behaviors (social distancing) can be facilitated through epidemiological reporting, social construction of perceived risk, and local monitoring of direct connections, but eroded via social reassurance. We varied reported information (total active cases, daily new cases, hospitalizations, hospital capacity exceeded, or deaths) at one of two scales (population level or community level). Total active and new case reporting at the population level were the most effective approaches, relative to the other reporting approaches. Case reporting, which synergizes with test-trace-and-isolate and vaccination policies, should remain a priority throughout an epidemic.

Introduction to supervised machine learning in clinical epidemiology

Machine learning refers to a series of processes in which a computer finds rules from a vast amount of data. With recent advances in computer technology and the availability of a wide variety of health data, machine learning has rapidly developed and been applied in medical research. Currently, there are three types of machine learning: supervised, unsupervised, and reinforcement learning. In medical research, supervised learning is commonly used for diagnoses and prognoses, while unsupervised learning is used for phenotyping a disease, and reinforcement learning for maximizing favorable results, such as optimization of total patients' waiting time in the emergency department. The present article focuses on the concept and application of supervised learning in medicine, the most commonly used machine learning approach in medicine, and provides a brief explanation of four algorithms widely used for prediction (random forests, gradient-boosted decision tree, support vector machine, and neural network). Among these algorithms, the neural network has further developed into deep learning algorithms to solve more complex tasks. Along with simple classification problems, deep learning is commonly used to process medical imaging, such as retinal fundus photographs for diabetic retinopathy diagnosis. Although machine learning can bring new insights into medicine by processing a vast amount of data that are often beyond human capacity, algorithms can also fail when domain knowledge is neglected. The combination of algorithms and human cognitive ability is a key to the successful application of machine learning in medicine.

A Novel Missing Data Imputation Approach for Time Series Air Quality Data Based on Logistic Regression

Missing values in air quality datasets bring trouble to exploration and decision making about the environment. Few imputation methods aim at time series air quality data so that they fail to handle the timeliness of the data. Moreover, most imputation methods prefer low-missing-rate datasets to relatively high-missing-rate datasets. This paper proposes a novel missing data imputation method, called FTLRI, for time series air quality data based on the traditional logistic regression and a presented “first Five & last Three” model, which can explain relationships between disparate attributes and extract data that are extremely relevant, both in terms of time and attributes, to the missing data, respectively. To investigate the performance of FTLRI, it is benchmarked with five classical baselines and a new dynamic imputation method using a neural network with average hourly concentration data of pollutants from three disparate stations in Lanzhou in 2019 under different missing rates. The results show that FTLRI has a significant advantage over the compared imputation approaches, both in the particular short-term and long-term time series air quality data. Furthermore, FTLRI has good performance on datasets with a relatively high missing rate, since it only selects the data extremely related to the missing values instead of relying on all the other data like other methods.

Spatial Variation in Humidity and the Onset of Seasonal Influenza Across the Contiguous United States

In recent years, environmental factors, particularly humidity, have been used to inform influenza prediction models. This study aims to quantify the relationship between humidity and influenza incidence at the state-level in the contiguous United States. Piecewise segmented regressions were performed on specific humidity data from NASA's Atmospheric Infrared Sounder (AIRS) and incident influenza estimates from Google Flu Trends to identify threshold values of humidity that signal the onset of an influenza outbreak. Our results suggest that influenza incidence increases after reaching a humidity threshold that is state-specific. A linear regression showed that the state-specific thresholds were associated with annual average humidity conditions (R 2 = 0.9). Threshold values statistically significantly varied by region (F-statistic = 8.274, p < 0.001) and of their 36 pairwise combinations, 13 pairs had at least marginally statistically significant differences in their means. All of the significant comparisons included either the South or Southeast region, which had higher humidity threshold values. Results from this study improve our understanding of the significance of humidity in the transmission of influenza and reinforce the need for local and regional conditions to be considered in this relationship. Ultimately this could help researchers to produce more accurate forecasts of seasonal influenza onset and provide health officials with better information prior to outbreaks.

Web Crawling of Social Media and Related Web Platforms to Analyze Backyard Poultry Owners Responses to the 2018-2020 Newcastle Disease (ND) Outbreak in Southern California

As social media becomes an ever-increasing staple of everyday life and a growing percentage of people turn to community driven platforms as a primary source of information, the data created from these posts can provide a new source of information from which to better understand an event in near real-time. The 2018-2020 outbreak of Newcastle disease (ND) in Southern California is the third outbreak of ND in Southern California within a 50-year time span. These outbreaks are thought to be primarily driven by non-commercial poultry (i.e. backyard and game fowl) in the region. Here we employed a commercial "web crawling" tool between June of 2018 and July of 2020 which encompassed the majority of the outbreak in order to collect all available online mentions of "virulent Newcastle Disease" (vND), the terminology commonly used by the California Department of Food and Agriculture (CDFA), United States Department of Agriculture (USDA), and the general public, in relation to the outbreak. A total of 2,498 posts in English and Spanish were returned using a Boolean logic-based string search. While the number of posts was relatively small, their impact as measured by the number of visitors to the website and the number of people viewing the post (where provided) was much larger. Posts with negative sentiment were found to have a larger audience relative to posts with a positive sentiment. In addition, posts with negative sentiment peaked in May of 2019 which preceded the formation of the anti-depopulation group Save Our Birds (SOB). As the usage and impact of social media grows, the ability to utilize tools to analyze social media may improve both response and outreach-based strategies for various disease outbreaks including vND in Southern California which has a large non-commercial poultry population. This article is protected by copyright. All rights reserved.

United States Influenza Search Patterns Since the Emergence of COVID-19: Infodemiology Study

Background

The emergence and media coverage of COVID-19 may have affected influenza search patterns, possibly affecting influenza surveillance results using Google Trends.

Objective

We aimed to investigate if the emergence of COVID-19 was associated with modifications in influenza search patterns in the United States.

Methods

We retrieved US Google Trends data (relative number of searches for specified terms) for the topics influenza, Coronavirus disease 2019, and symptoms shared between influenza and COVID-19. We calculated the correlations between influenza and COVID-19 search data for a 1-year period after the first COVID-19 diagnosis in the United States (January 21, 2020 to January 20, 2021). We constructed a seasonal autoregressive integrated moving average model and compared predicted search volumes, using the 4 previous years, with Google Trends relative search volume data. We built a similar model for shared symptoms data. We also assessed correlations for the past 5 years between Google Trends influenza data, US Centers for Diseases Control and Prevention influenza-like illness data, and influenza media coverage data.

Results

We observed a nonsignificant weak correlation (ρ= –0.171; P=0.23) between COVID-19 and influenza Google Trends data. Influenza search volumes for 2020-2021 distinctly deviated from values predicted by seasonal autoregressive integrated moving average models—for 6 weeks within the first 13 weeks after the first COVID-19 infection was confirmed in the United States, the observed volume of searches was higher than the upper bound of 95% confidence intervals for predicted values. Similar results were observed for shared symptoms with influenza and COVID-19 data. The correlation between Google Trends influenza data and CDC influenza-like-illness data decreased after the emergence of COVID-19 (2020-2021: ρ=0.643; 2019-2020: ρ=0.902), while the correlation between Google Trends influenza data and influenza media coverage volume remained stable (2020-2021: ρ=0.746; 2019-2020: ρ=0.707).

Conclusions

Relevant differences were observed between predicted and observed influenza Google Trends data the year after the onset of the COVID-19 pandemic in the United States. Such differences are possibly due to media coverage, suggesting limitations to the use of Google Trends as a flu surveillance tool.

Improving Google Flu Trends for COVID-19 estimates using Weibo posts

While incomplete non-medical data has been integrated into prediction models for epidemics, the accuracy and the generalizability of the data are difficult to guarantee. To comprehensively evaluate the ability and applicability of using social media data to predict the development of COVID-19, a new confirmed case prediction algorithm improving the Google Flu Trends algorithm is established, called Weibo COVID-19 Trends (WCT), based on the post dataset generated by all users in Wuhan on Sina Weibo. A genetic algorithm is designed to select the keyword set for filtering COVID-19 related posts. WCT can constantly outperform the highest average test score in the training set between daily new confirmed case counts and the prediction results. It remains to produce the best prediction results among other algorithms when the number of forecast days increases from one to eight days with the highest correlation score from 0.98 (P ​< 0.01) to 0.86 (P ​< 0.01) during all analysis period. Additionally, WCT effectively improves the Google Flu Trends algorithm's shortcoming of overestimating the epidemic peak value. This study offers a highly adaptive approach for feature engineering of third-party data in epidemic prediction, providing useful insights for the prediction of newly emerging infectious diseases at an early stage.

A novel data-driven methodology for influenza outbreak detection and prediction

Influenza is an infectious disease that leads to an estimated 5 million cases of severe illness and 650,000 respiratory deaths worldwide each year. The early detection and prediction of influenza outbreaks are crucial for efficient resource planning to save patient’s lives and healthcare costs. We propose a new data-driven methodology for influenza outbreak detection and prediction at very local levels. A doctor’s diagnostic dataset of influenza-like illness from more than 3000 clinics in Malaysia is used in this study because these diagnostic data are reliable and can be captured promptly. A new region index (RI) of the influenza outbreak is proposed based on the diagnostic dataset. By analysing the anomalies in the weekly RI value, potential outbreaks are identified using statistical methods. An ensemble learning method is developed to predict potential influenza outbreaks. Cross-validation is conducted to optimize the hyperparameters of the ensemble model. A testing data set is used to provide an unbiased evaluation of the model. The proposed methodology is shown to be sensitive and accurate at influenza outbreak prediction, with average of 75% recall, 74% precision, and 83% accuracy scores across five regions in Malaysia. The results are also validated by Google Flu Trends data, news reports, and surveillance data released by World Health Organization.

Multiscale influenza forecasting

Influenza forecasting in the United States (US) is complex and challenging due to spatial and temporal variability, nested geographic scales of interest, and heterogeneous surveillance participation. Here we present Dante, a multiscale influenza forecasting model that learns rather than prescribes spatial, temporal, and surveillance data structure and generates coherent forecasts across state, regional, and national scales. We retrospectively compare Dante's short-term and seasonal forecasts for previous flu seasons to the Dynamic Bayesian Model (DBM), a leading competitor. Dante outperformed DBM for nearly all spatial units, flu seasons, geographic scales, and forecasting targets. Dante's sharper and more accurate forecasts also suggest greater public health utility. Dante placed 1st in the Centers for Disease Control and Prevention's prospective 2018/19 FluSight challenge in both the national and regional competition and the state competition. The methodology underpinning Dante can be used in other seasonal disease forecasting contexts having nested geographic scales of interest.

The impact of the pandemic declaration on public awareness and behavior: Focusing on COVID-19 google searches

The unprecedented outbreaks of epidemics such as the coronavirus has caused major socio-economic changes. To analyze public risk awareness and behavior in response to the outbreak of epidemic diseases, this study focuses on RSV (Relative Search Volume) provided by Google Trends. This study uses the social big data provided by Google RSV to investigate how the WHO's pandemic declaration affected public awareness and behavior. 37 OECD countries were analyzed and clustered according to the degree of reaction to the declaration, and the United States, France and Germany were selected for comparative study. The results of this study statistically confirmed that the pandemic declaration increased public awareness and had the effect of increasing searches for information on COVID-19 by more than 20%. In addition, this rapid rise in RSV also reflected interest in the COVID-19 test and had the effect of inducing individuals to be tested, which helped identify new cases. The significance of this study is that it provided the theoretical foundation for using RSV and its implications to understand and strategically utilize public awareness and behavior in situations where the WHO and governments must launch policies in response to the outbreak of new infectious diseases such as COVID-19.

Associations between Google Search Trends for Symptoms and COVID-19 Confirmed and Death Cases in the United States

We utilize functional data analysis techniques to investigate patterns of COVID-19 positivity and mortality in the US and their associations with Google search trends for COVID-19-related symptoms. Specifically, we represent state-level time series data for COVID-19 and Google search trends for symptoms as smoothed functional curves. Given these functional data, we explore the modes of variation in the data using functional principal component analysis (FPCA). We also apply functional clustering analysis to identify patterns of COVID-19 confirmed case and death trajectories across the US. Moreover, we quantify the associations between Google COVID-19 search trends for symptoms and COVID-19 confirmed case and death trajectories using dynamic correlation. Finally, we examine the dynamics of correlations for the top nine Google search trends of symptoms commonly associated with COVID-19 confirmed case and death trajectories. Our results reveal and characterize distinct patterns for COVID-19 spread and mortality across the US. The dynamics of these correlations suggest the feasibility of using Google queries to forecast COVID-19 cases and mortality for up to three weeks in advance. Our results and analysis framework set the stage for the development of predictive models for forecasting COVID-19 confirmed cases and deaths using historical data and Google search trends for nine symptoms associated with both outcomes.

Temporal Trends and Interest in Coronary Artery Calcium Scoring Over Time: An Infodemiology Study

Objective

To evaluate interest in coronary artery calcium (CAC) among the general public during the past 17 years and to compare trends with real-world data on number of CAC procedures performed.

Methods

We used Google Trends, a publicly available database, to access search query data in a systematic and quantitative fashion to search for CAC-related key terms. Search terms included calcium test, heart score, calcium score, coronary calcium, and calcium test score. We accessed Google Trends in January 2021 and analyzed data from 2004 to 2020.

Results

From 2004 to December 31, 2020, CAC-related search interest (in relative search volume) increased continually worldwide (+201.9%) and in the United States (+354.8%). Three main events strongly influenced search interest in CAC: reports of a CAC scan of the president of the United States led to a transient 10-fold increase in early January 2018. American College of Cardiology/American Heart Association guideline release led to a sustained increase, and lockdown after the global pandemic due to COVID-19 led to a transient decrease. Real-world data on performed CAC scans showed an increase between 2006 and 2017 (+200.0%); during the same time period, relative search volume for CAC-related search terms increased in a similar pattern (+70.6%-1511.1%). For the search term coronary calcium scan near me, a potential representative of active online search for CAC scanning, we found a +28.8% increase in 2020 compared with 2017.

Conclusion

Google Trends, a valuable tool for assessing public interest in health-related topics, suggests increased overall interest in CAC during the last 17 years that mirrors real-world usage data. Increased interest is seemingly linked to reports of CAC testing in world leaders and endorsement in major guidelines.

Decreased public pursuit of cancer-related information during the COVID-19 pandemic in the United States

Background

In response to the prioritization of healthcare resources towards the COVID-19 pandemic, routine cancer screening and diagnostic have been disrupted, potentially explaining the apparent COVID-era decline in cancer cases and mortality. In this study, we identified temporal trends in public interest in cancer-related health information using the nowcasting tool Google Trends.

Methods

We used Google Trends to query search terms related to cancer types for short-term (September 2019–September 2020) and long-term (September 2016–September 2020) trends in the US. We compared average relative search volumes (RSV) for specified time ranges to detect recent and seasonal variation.

Results

General search interest declined for all cancer types beginning in March 2020, with changes in search interest for “Breast cancer,” “Colorectal cancer,” and “Melanoma” of − 30.6%, − 28.2%, and − 26.7%, respectively, and compared with the mean RSV of the two previous months. In the same time range, search interest for “Telemedicine” has increased by + 907.1% and has reached a 4-year peak with a sustained increased level of search interest. Absolute cancer mortality has declined and is presently at a 4-year low; however, search interest in cancer has been recuperating since July 2020.

Conclusion

We observed a marked decline in searches for cancer-related health information that mirrors the reduction in new cancer diagnoses and cancer mortality during the COVID-19 pandemic. Health professions need to be prepared for the coming demand for cancer-related healthcare, foreshadowed by recovering interest in cancer-related information on Google Trends.

Supplementary Information

The online version contains supplementary material available at 10.1007/s10552-021-01409-1.

Internet search patterns reveal clinical course of COVID-19 disease progression and pandemic spread across 32 countries

Effective public health response to novel pandemics relies on accurate and timely surveillance of pandemic spread, as well as characterization of the clinical course of the disease in affected individuals. We sought to determine whether Internet search patterns can be useful for tracking COVID-19 spread, and whether these data could also be useful in understanding the clinical progression of the disease in 32 countries across six continents. Temporal correlation analyses were conducted to characterize the relationships between a range of COVID-19 symptom-specific search terms and reported COVID-19 cases and deaths for each country from January 1 through April 20, 2020. Increases in COVID-19 symptom-related searches preceded increases in reported COVID-19 cases and deaths by an average of 18.53 days (95% CI 15.98-21.08) and 22.16 days (20.33-23.99), respectively. Cross-country ensemble averaging was used to derive average temporal profiles for each search term, which were combined to create a search-data-based view of the clinical course of disease progression. Internet search patterns revealed a clear temporal pattern of disease progression for COVID-19: Initial symptoms of fever, dry cough, sore throat and chills were followed by shortness of breath an average of 5.22 days (3.30-7.14) after initial symptom onset, matching the clinical course reported in the medical literature. This study shows that Internet search data can be useful for characterizing the detailed clinical course of a disease. These data are available in real-time at population scale, providing important benefits as a complementary resource for tracking pandemics, especially before widespread laboratory testing is available.

Using Baidu search values to monitor and predict the confirmed cases of COVID-19 in China: - evidence from Baidu index

BACKGROUND

New coronavirus disease 2019 (COVID-19) has posed a severe threat to human life and caused a global pandemic. The current research aimed to explore whether the search-engine query patterns could serve as a potential tool for monitoring the outbreak of COVID-19.

METHODS

We collected the number of COVID-19 confirmed cases between January 11, 2020, and April 22, 2020, from the Center for Systems Science and Engineering (CSSE) at Johns Hopkins University (JHU). The search index values of the most common symptoms of COVID-19 (e.g., fever, cough, fatigue) were retrieved from the Baidu Index. Spearman's correlation analysis was used to analyze the association between the Baidu index values for each COVID-19-related symptom and the number of confirmed cases. Regional distributions among 34 provinces/ regions in China were also analyzed.

RESULTS

Daily growth of confirmed cases and Baidu index values for each COVID-19-related symptom presented robust positive correlations during the outbreak (fever: rs=0.705, p=9.623× 10- 6; cough: rs=0.592, p=4.485× 10- 4; fatigue: rs=0.629, p=1.494× 10- 4; sputum production: rs=0.648, p=8.206× 10- 5; shortness of breath: rs=0.656, p=6.182× 10-5). The average search-to-confirmed interval (STCI) was 19.8 days in China. The daily Baidu Index value's optimal time lags were the 4 days for cough, 2 days for fatigue, 3 days for sputum production, 1 day for shortness of breath, and 0 days for fever.

CONCLUSION

The searches of COVID-19-related symptoms on the Baidu search engine were significantly correlated to the number of confirmed cases. Since the Baidu search engine could reflect the public's attention to the pandemic and the regional epidemics of viruses, relevant departments need to pay more attention to areas with high searches of COVID-19-related symptoms and take precautionary measures to prevent these potentially infected persons from further spreading.

Big Data and Actuarial Science

This article investigates the impact of big data on the actuarial sector. The growing fields of applications of data analytics and data mining raise the ability for insurance companies to conduct more accurate policy pricing by incorporating a broader variety of data due to increased data availability. The analyzed areas of this paper span from automobile insurance policy pricing, mortality and healthcare modeling to estimation of harvest-, climate- and cyber risk as well as assessment of catastrophe risk such as storms, hurricanes, tornadoes, geomagnetic events, earthquakes, floods, and fires. We evaluate the current use of big data in these contexts and how the utilization of data analytics and data mining contribute to the prediction capabilities and accuracy of policy premium pricing of insurance companies. We find a high penetration of insurance policy pricing in almost all actuarial fields except in the modeling and pricing of cyber security risk due to lack of data in this area and prevailing data asymmetries, for which we identify the application of artificial intelligence, in particular machine learning techniques, as a possible solution to improve policy pricing accuracy and results.

Accurate spatiotemporal mapping of drug overdose deaths by machine learning of drug-related web-searches

Persons who inject drugs (PWID) are at increased risk for overdose death (ODD), infections with HIV, hepatitis B (HBV) and hepatitis C virus (HCV), and noninfectious health conditions. Spatiotemporal identification of PWID communities is essential for developing efficient and cost-effective public health interventions for reducing morbidity and mortality associated with injection-drug use (IDU). Reported ODDs are a strong indicator of the extent of IDU in different geographic regions. However, ODD quantification can take time, with delays in ODD reporting occurring due to a range of factors including death investigation and drug testing. This delayed ODD reporting may affect efficient early interventions for infectious diseases. We present a novel model, Dynamic Overdose Vulnerability Estimator (DOVE), for assessment and spatiotemporal mapping of ODDs in different U.S. jurisdictions. Using Google^® Web-search volumes (i.e., the fraction of all searches that include certain words), we identified a strong association between the reported ODD rates and drug-related search terms for 2004–2017. A machine learning model (Extremely Random Forest) was developed to produce yearly ODD estimates at state and county levels, as well as monthly estimates at state level. Regarding the total number of ODDs per year, DOVE’s error was only 3.52% (Median Absolute Error, MAE) in the United States for 2005–2017. DOVE estimated 66,463 ODDs out of the reported 70,237 (94.48%) during 2017. For that year, the MAE of the individual ODD rates was 4.43%, 7.34%, and 12.75% among yearly estimates for states, yearly estimates for counties, and monthly estimates for states, respectively. These results indicate suitability of the DOVE ODD estimates for dynamic IDU assessment in most states, which may alert for possible increased morbidity and mortality associated with IDU. ODD estimates produced by DOVE offer an opportunity for a spatiotemporal ODD mapping. Timely identification of potential mortality trends among PWID might assist in developing efficient ODD prevention and HBV, HCV, and HIV infection elimination programs by targeting public health interventions to the most vulnerable PWID communities.

Leveraging Internet Search Data to Improve the Prediction and Prevention of Noncommunicable Diseases: Retrospective Observational Study

BACKGROUND

As human society enters an era of vast and easily accessible social media, a growing number of people are exploiting the internet to search and exchange medical information. Because internet search data could reflect population interest in particular health topics, they provide a new way of understanding health concerns regarding noncommunicable diseases (NCDs) and the role they play in their prevention.

OBJECTIVE

We aimed to explore the association of internet search data for NCDs with published disease incidence and mortality rates in the United States and to grasp the health concerns toward NCDs.

METHODS

We tracked NCDs by examining the correlations among the incidence rates, mortality rates, and internet searches in the United States from 2004 to 2017, and we established forecast models based on the relationship between the disease rates and internet searches.

RESULTS

Incidence and mortality rates of 29 diseases in the United States were statistically significantly correlated with the relative search volumes (RSVs) of their search terms (P<.05). From the perspective of the goodness of fit of the multiple regression prediction models, the results were closest to 1 for diabetes mellitus, stroke, atrial fibrillation and flutter, Hodgkin lymphoma, and testicular cancer; the coefficients of determination of their linear regression models for predicting incidence were 80%, 88%, 96%, 80%, and 78%, respectively. Meanwhile, the coefficient of determination of their linear regression models for predicting mortality was 82%, 62%, 94%, 78%, and 62%, respectively.

CONCLUSIONS

An advanced understanding of search behaviors could augment traditional epidemiologic surveillance and could be used as a reference to aid in disease prediction and prevention.

Track Iran's national COVID-19 response committee's major concerns using two-stage unsupervised topic modeling

Background

Since the World Health Organization (WHO) declared the COVID-19 as a Public Health Emergency of International Concern (PHEIC) on January 31, 2020, governments have been enfaced with crisis for timely responses. The efficacy of these responses directly depends on the social behaviors of the target society. People react to these actions with respect to the information they received from different channels, such as news and social networks. Thus, analyzing news demonstrates a brief view of the information users received during the outbreak.

Methods

The raw data used in this study is collected from official news channels of news wires and agencies in Telegram messenger, which exceeds 2,400,000 posts. The posts that are quoted by NCRC’s members are collected, cleaned, and divided into sentences. The topic modeling and tracking are utilized in a two-stage framework, which is customized for this problem to separate miscellaneous sentences from those presenting concerns. The first stage is fed with embedding vectors of sentences where they are grouped by the Mapper algorithm. Sentences belonging to singleton nodes are labeled as miscellaneous sentences. The remained sentences are vectorized, adopting Tf-IDF weighting schema in the second stage and topically modeled by the LDA method. Finally, relevant topics are aligned to the list of policies and actions, named topic themes, that are set up by the NCRC.

Results

Our results show that major concerns presented in about half of the sentences are (1) PCR lab. test, diagnosis, and screening, (2) Closure of the education system, and (3) awareness actions about washing hands and facial mask usage. Among the eight themes, intra-provincial travel and traffic restrictions, as well as briefing the national and provincial status, are under-presented. The timeline of concerns annotated by the preventive actions illustrates the changes in concerns addressed by NCRC. This timeline shows that although the announcements and public responses are not lagged behind the events, but cannot be considered as timely. Furthermore, the fluctuating series of concerns reveal that the NCRC has not a long-time response map, and members react to the closest announced policy/act.

Conclusion

The results of our study can be used as a quantitative indicator for evaluating the availability of an on-time public response of Iran’s NCRC during the first three months of the outbreak. Moreover, it can be used in comparative studies to investigate the differences between awareness acts in various countries. Results of our customized-design framework showed that about one-third of the discussions of the NCRC’s members cover miscellaneous topics that must be removed from the data.

How the ecology and evolution of the COVID-19 pandemic changed learning

The coronavirus disease 2019 (COVID-19) pandemic introduced an abrupt change in human behavior globally. Here, we discuss unique insights the pandemic has provided into the eco-evolutionary role of pathogens in ecosystems and present data that indicates the pandemic may have fundamentally changed our learning choices. COVID-19 has indirectly affected many organisms and processes by changing the behavior of humans to avoid being infected. The pandemic also changed our learning behavior by affecting the relative importance of information and forcing teaching and learning into a framework that accommodates human behavioral measures to avoid disease transmission. Not only are these indirect effects on the environment occurring through a unique mechanistic pathway in ecology, the pandemic along with its effects on us provides a profound example of the role risk can play in the transmission of information between the at risk. Ultimately, these changes in our learning behavior led to this special issue "Taking learning online in Ecology and Evolution." The special issue was a call to the community to take learning in new directions, including online and distributed experiences. The topics examined include a significant component of DIY ecology and evolution that is experiential but done individually, opportunities to use online tools and apps to be more inclusive, student-focused strategies for teaching online, how to reinvent conferences, strategies to retain experiential learning safely, emerging forms of teaching such as citizen science, apps and podcasting, and ideas on how to accommodate ever changing constraints in the college classroom, to name a few. The collective consensus in our fields is that these times are challenging but we can continue to improve and innovate on existing developments, and more broadly and importantly, this situation may provide an opportunity to reset some of the existing practices that fail to promote an effective and inclusive learning environment.

Online Public Attention During the Early Days of the COVID-19 Pandemic: Infoveillance Study Based on Baidu Index

BACKGROUND

The COVID-19 pandemic has become a global public health event, attracting worldwide attention. As a tool to monitor public awareness, internet search engines have been widely used in public health emergencies.

OBJECTIVE

This study aims to use online search data (Baidu Index) to monitor the public's attention and verify internet search engines' function in public attention monitoring of public health emergencies.

METHODS

We collected the Baidu Index and the case monitoring data from January 20, 2020, to April 20, 2020. We combined the Baidu Index of keywords related to COVID-19 to describe the public attention's temporal trend and spatial distribution, and conducted the time lag cross-correlation analysis.

RESULTS

The Baidu Index temporal trend indicated that the changes of the Baidu Index had a clear correspondence with the development time node of the pandemic. The Baidu Index spatial distribution showed that in the regions of central and eastern China, with denser populations, larger internet user bases, and higher economic development levels, the public was more concerned about COVID-19. In addition, the Baidu Index was significantly correlated with six case indicators of new confirmed cases, new death cases, new cured discharge cases, cumulative confirmed cases, cumulative death cases, and cumulative cured discharge cases. Moreover, the Baidu Index was 0-4 days earlier than new confirmed and new death cases, and about 20 days earlier than new cured and discharged cases while 3-5 days later than the change of cumulative cases.

CONCLUSIONS

The national public's demand for epidemic information is urgent regardless of whether it is located in the hardest hit area. The public was more sensitive to the daily new case data that represents the progress of the epidemic, but the public's attention to the epidemic situation in other areas may lag behind. We could set the Baidu Index as the sentinel and the database in the online infoveillance system for infectious disease and public health emergencies. According to the monitoring data, the government needs to prevent and control the possible outbreak in advance and communicate the risks to the public so as to ensure the physical and psychological health of the public in the epidemic.

A Study on Fight Against COVID-19 from Latest Technological Intervention

Uncontrolled spread of pandemic COVID-19 in India and across the globe over several months, created an impact as never before any pandemic would have created. This certainly demands a technological intervention from all possibility to overcome the situation and lead a normal life as early as possible. AI/Machine learning responds to the situation, through inspecting different aspects of the pandemic. This paper analyses and studies those aspects, (I) Quarantine and statistical aspect: Quarantine potentially affected candidates (person who is in touch, travel history) through Data analytics/Machine learning. (II) Diagnosis and Treatment aspect: Early detection and fast treatment will save lives. Diagnosis using deep learning assists radiologist from saving their effort and time to a greater extent and arrives faster conclusion. (III) Prevention aspect: Monitoring and enforce social distancing through visual social distancing using deep learning and Computer vision.

A demonstration project of Global Alliance against Chronic Respiratory Diseases: Prediction of interactions between air pollution and allergen exposure-the Mobile Airways Sentinel NetworK-Impact of air POLLution on Asthma and Rhinitis approach

This review analyzes the state and recent progress in the field of information support for pollen allergy sufferers. For decades, information available for the patients and allergologists consisted of pollen counts, which are vital but insufficient. New technology paves the way to substantial increase in amount and diversity of the data. This paper reviews old and newly suggested methods to predict pollen and air pollutant concentrations in the air and proposes an allergy risk concept, which combines the pollen and pollution information and transforms it into a qualitative risk index. This new index is available in an app (Mobile Airways Sentinel NetworK-air) that was developed in the frame of the European Union grant Impact of Air POLLution on sleep, Asthma and Rhinitis (a project of European Institute of Innovation and Technology-Health). On-going transformation of the pollen allergy information support is based on new technological solutions for pollen and air quality monitoring and predictions. The new information-technology and artificial-intelligence-based solutions help to convert this information into easy-to-use services for both medical practitioners and allergy sufferers.

Investigating the utility of Google trends for Zika and Chikungunya surveillance in Venezuela

INTRODUCTION

Chikungunya and Zika Virus are vector-borne diseases responsible for a substantial disease burden in the Americas. Between 2013 and 2016, no cases of Chikungunya or Zika Virus were reported by the Venezuelan Ministry of Health. However, peaks of undiagnosed fever cases have been observed during the same period. In the context of scarce data, alternative surveillance methods are needed. Assuming that unusual peaks of acute fever cases correspond to the incidences of both diseases, this study aims to evaluate the use of Google Trends as an indicator of the epidemic behavior of Chikungunya and Zika.

METHODS

Time-series cross-correlations of acute fever cases reported by the Venezuelan Ministry of Health and data on Google search queries related to Chikungunya and Zika were calculated.

RESULTS

A temporal distinction has been made so that acute febrile cases occurring between 25th of June 2014 and 23rd of April 2015 were attributed to the Chikungunya virus, while cases occurring between 30th of April 2015 and 29th of April 2016 were ascribed to the Zika virus. The highest cross-correlations for each disease were shown at a lag of 0 (r = 0.784) for Chikungunya and at + 1 (r = 0.754) for Zika.

CONCLUSION

The strong positive correlation between Google search queries and official data on acute febrile cases suggests that this resource can be used as an indicator of endemic urban arboviruses activity. In the Venezuelan context, Internet search queries might help to overcome some of the gaps that exist in the national surveillance system.

Correlations of Online Search Engine Trends With Coronavirus Disease (COVID-19) Incidence: Infodemiology Study

BACKGROUND

The coronavirus disease (COVID-19) is the latest pandemic of the digital age. With the internet harvesting large amounts of data from the general population in real time, public databases such as Google Trends (GT) and the Baidu Index (BI) can be an expedient tool to assist public health efforts.

OBJECTIVE

The aim of this study is to apply digital epidemiology to the current COVID-19 pandemic to determine the utility of providing adjunctive epidemiologic information on outbreaks of this disease and evaluate this methodology in the case of future pandemics.

METHODS

An epidemiologic time series analysis of online search trends relating to the COVID-19 pandemic was performed from January 9, 2020, to April 6, 2020. BI was used to obtain online search data for China, while GT was used for worldwide data, the countries of Italy and Spain, and the US states of New York and Washington. These data were compared to real-world confirmed cases and deaths of COVID-19. Chronologic patterns were assessed in relation to disease patterns, significant events, and media reports.

RESULTS

Worldwide search terms for shortness of breath, anosmia, dysgeusia and ageusia, headache, chest pain, and sneezing had strong correlations (r>0.60, P<.001) to both new daily confirmed cases and deaths from COVID-19. GT COVID-19 (search term) and GT coronavirus (virus) searches predated real-world confirmed cases by 12 days (r=0.85, SD 0.10 and r=0.76, SD 0.09, respectively, P<.001). Searches for symptoms of diarrhea, fever, shortness of breath, cough, nasal obstruction, and rhinorrhea all had a negative lag greater than 1 week compared to new daily cases, while searches for anosmia and dysgeusia peaked worldwide and in China with positive lags of 5 days and 6 weeks, respectively, corresponding with widespread media coverage of these symptoms in COVID-19.

CONCLUSIONS

This study demonstrates the utility of digital epidemiology in providing helpful surveillance data of disease outbreaks like COVID-19. Although certain online search trends for this disease were influenced by media coverage, many search terms reflected clinical manifestations of the disease and showed strong correlations with real-world cases and deaths.

Global Preparedness Against COVID-19: We Must Leverage the Power of Digital Health

The coronavirus disease (COVID-19) pandemic has revealed many areas of public health preparedness that are lacking, especially in lower- and middle-income countries. Digital interventions provide many opportunities for strengthening health systems and could be vital resources in the current public health emergency. We provide several use cases for infection control, home-based diagnosis and screening, empowerment through information, public health surveillance and epidemiology, and leveraging crowd-sourced data. A thoughtful, concerted effort-leveraging existing experience and robust enterprise-grade technologies-can have a substantive impact on the immediate and distal consequences of COVID-19.