More diseases tracked by using Google Trends

Effects of the Ambient Fine Particulate Matter on Public Awareness of Lung Cancer Risk in China: Evidence from the Internet-Based Big Data Platform

Background

In October 2013, the International Agency for Research on Cancer classified the particulate matter from outdoor air pollution as a group 1 carcinogen and declared that particulate matter can cause lung cancer. Fine particular matter (PM_2.5) pollution is becoming a serious public health concern in urban areas of China. It is essential to emphasize the importance of the public’s awareness and knowledge of modifiable risk factors of lung cancer for prevention.

Objective

The objective of our study was to explore the public’s awareness of the association of PM_2.5 with lung cancer risk in China by analyzing the relationship between the daily PM_2.5 concentration and searches for the term “lung cancer” on an Internet big data platform, Baidu.

Methods

We collected daily PM_2.5 concentration data and daily Baidu Index data in 31 Chinese capital cities from January 1, 2014 to December 31, 2016. We used Spearman correlation analysis to explore correlations between the daily Baidu Index for lung cancer searches and the daily average PM_2.5 concentration. Granger causality test was used to analyze the causal relationship between the 2 time-series variables.

Results

In 23 of the 31 cities, the pairwise correlation coefficients (Spearman rho) between the daily Baidu Index for lung cancer searches and the daily average PM_2.5 concentration were positive and statistically significant (P<.05). However, the correlation between the daily Baidu Index for lung cancer searches and the daily average PM_2.5 concentration was poor (all r²_s<.1). Results of Granger causality testing illustrated that there was no unidirectional causality from the daily PM_2.5 concentration to the daily Baidu Index for lung cancer searches, which was statistically significant at the 5% level for each city.

Conclusions

The daily average PM_2.5 concentration had a weak positive impact on the daily search interest for lung cancer on the Baidu search engine. Well-designed awareness campaigns are needed to enhance the general public’s awareness of the association of PM_2.5 with lung cancer risk, to lead the public to seek more information about PM_2.5 and its hazards, and to cope with their environment and its risks appropriately.

Low validity of Google Trends for behavioral forecasting of national suicide rates

Recent research suggests that search volumes of the most popular search engine worldwide, Google, provided via Google Trends, could be associated with national suicide rates in the USA, UK, and some Asian countries. However, search volumes have mostly been studied in an ad hoc fashion, without controls for spurious associations. This study evaluated the validity and utility of Google Trends search volumes for behavioral forecasting of suicide rates in the USA, Germany, Austria, and Switzerland. Suicide-related search terms were systematically collected and respective Google Trends search volumes evaluated for availability. Time spans covered 2004 to 2010 (USA, Switzerland) and 2004 to 2012 (Germany, Austria). Temporal associations of search volumes and suicide rates were investigated with time-series analyses that rigorously controlled for spurious associations. The number and reliability of analyzable search volume data increased with country size. Search volumes showed various temporal associations with suicide rates. However, associations differed both across and within countries and mostly followed no discernable patterns. The total number of significant associations roughly matched the number of expected Type I errors. These results suggest that the validity of Google Trends search volumes for behavioral forecasting of national suicide rates is low. The utility and validity of search volumes for the forecasting of suicide rates depend on two key assumptions ("the population that conducts searches consists mostly of individuals with suicidal ideation", "suicide-related search behavior is strongly linked with suicidal behavior"). We discuss strands of evidence that these two assumptions are likely not met. Implications for future research with Google Trends in the context of suicide research are also discussed.

Using Web-Based Search Data to Study the Public's Reactions to Societal Events: The Case of the Sandy Hook Shooting

Background

Internet search is the most common activity on the World Wide Web and generates a vast amount of user-reported data regarding their information-seeking preferences and behavior. Although this data has been successfully used to examine outbreaks, health care utilization, and outcomes related to quality of care, its value in informing public health policy remains unclear.

Objective

The aim of this study was to evaluate the role of Internet search query data in health policy development. To do so, we studied the public’s reaction to a major societal event in the context of the 2012 Sandy Hook School shooting incident.

Methods

Query data from the Yahoo! search engine regarding firearm-related searches was analyzed to examine changes in user-selected search terms and subsequent websites visited for a period of 14 days before and after the shooting incident.

Results

A total of 5,653,588 firearm-related search queries were analyzed. In the after period, queries increased for search terms related to “guns” (+50.06%), “shooting incident” (+333.71%), “ammunition” (+155.14%), and “gun-related laws” (+535.47%). The highest increase (+1054.37%) in Web traffic was seen by news websites following “shooting incident” queries whereas searches for “guns” (+61.02%) and “ammunition” (+173.15%) resulted in notable increases in visits to retail websites. Firearm-related queries generally returned to baseline levels after approximately 10 days.

Conclusions

Search engine queries present a viable infodemiology metric on public reactions and subsequent behaviors to major societal events and could be used by policymakers to inform policy development.

[What can we expect from clinical trials in psychiatry?]

Clinical trials in psychiatry allow to build the regulatory dossiers for market authorization but also to document the mechanism of action of new drugs, to build pharmacodynamics models, evaluate the treatment effects, propose prognosis, efficacy or tolerability biomarkers and altogether to assess the impact of drugs for patient, caregiver and society. However, clinical trials have shown some limitations. Number of recent dossiers failed to convince the regulators. The clinical and biological heterogeneity of psychiatric disorders, the pharmacokinetic and pharmacodynamics properties of the compounds, the lack of translatable biomarkers possibly explain these difficulties. Several breakthrough options are now available: quantitative system pharmacology analysis of drug effects variability, pharmacometry and pharmacoepidemiology, Big Data analysis, brain modelling. In addition to more classical approaches, these opportunities lead to a paradigm change for clinical trials in psychiatry.

Measuring Global Disease with Wikipedia: Success, Failure, and a Research Agenda

Effective disease monitoring provides a foundation for effective public health systems. This has historically been accomplished with patient contact and bureaucratic aggregation, which tends to be slow and expensive. Recent internet-based approaches promise to be real-time and cheap, with few parameters. However, the question of when and how these approaches work remains open. We addressed this question using Wikipedia access logs and category links. Our experiments, replicable and extensible using our open source code and data, test the effect of semantic article filtering, amount of training data, forecast horizon, and model staleness by comparing across 6 diseases and 4 countries using thousands of individual models. We found that our minimal-configuration, language-agnostic article selection process based on semantic relatedness is effective for improving predictions, and that our approach is relatively insensitive to the amount and age of training data. We also found, in contrast to prior work, very little forecasting value, and we argue that this is consistent with theoretical considerations about the nature of forecasting. These mixed results lead us to propose that the currently observational field of internet-based disease surveillance must pivot to include theoretical models of information flow as well as controlled experiments based on simulations of disease.

Internet search query analysis can be used to demonstrate the rapidly increasing public awareness of palliative care in the USA

BACKGROUND

A lack of public awareness of palliative care (PC) has been identified as one of the main barriers to appropriate PC access. Internet search query analysis is a novel methodology, which has been effectively used in surveillance of infectious diseases, and can be used to monitor public awareness of health-related topics.

OBJECTIVES

We aimed to demonstrate the utility of internet search query analysis to evaluate changes in public awareness of PC in the USA between 2005 and 2015.

METHODS

Google Trends provides a referenced score for the popularity of a search term, for defined regions over defined time periods. The popularity of the search term 'palliative care' was measured monthly between 1/1/2005 and 31/12/2015 in the USA and in the UK.

RESULTS

Results were analysed using independent t-tests and joinpoint analysis. The mean monthly popularity of the search term increased between 2008-2009 (p<0.001), 2011-2012 (p<0.001), 2013-2014 (p=0.004) and 2014-2015 (p=0.002) in the USA. Joinpoint analysis was used to evaluate the monthly percentage change (MPC) in the popularity of the search term. In the USA, the MPC increase was 0.6%/month (p<0.05); in the UK the MPC of 0.05% was non-significant.

DISCUSSION

Although internet search query surveillance is a novel methodology, it is freely accessible and has significant potential to monitor health-seeking behaviour among the public. PC is rapidly growing in the USA, and the rapidly increasing public awareness of PC as demonstrated in this study, in comparison with the UK, where PC is relatively well established is encouraging in increasingly ensuring appropriate PC access for all.

Disease surveillance based on Internet-based linear models: an Australian case study of previously unmodeled infection diseases

Effective disease surveillance is critical to the functioning of health systems. Traditional approaches are, however, limited in their ability to deliver timely information. Internet-based surveillance systems are a promising approach that may circumvent many of the limitations of traditional health surveillance systems and provide more intelligence on cases of infection, including cases from those that do not use the healthcare system. Infectious disease surveillance systems built on Internet search metrics have been shown to produce accurate estimates of disease weeks before traditional systems and are an economically attractive approach to surveillance; they are, however, also prone to error under certain circumstances. This study sought to explore previously unmodeled diseases by investigating the link between Google Trends search metrics and Australian weekly notification data. We propose using four alternative disease modelling strategies based on linear models that studied the length of the training period used for model construction, determined the most appropriate lag for search metrics, used wavelet transformation for denoising data and enabled the identification of key search queries for each disease. Out of the twenty-four diseases assessed with Australian data, our nowcasting results highlighted promise for two diseases of international concern, Ross River virus and pneumococcal disease.

Examining the themes of STD-related Internet searches to increase specificity of disease forecasting using Internet search terms

US surveillance of sexually transmitted diseases (STDs) is often delayed and incomplete which creates missed opportunities to identify and respond to trends in disease. Internet search engine data has the potential to be an efficient, economical and representative enhancement to the established surveillance system. Google Trends allows the download of de-identified search engine data, which has been used to demonstrate the positive and statistically significant association between STD-related search terms and STD rates. In this study, search engine user content was identified by surveying specific exposure groups of individuals (STD clinic patients and university students) aged 18–35. Participants were asked to list the terms they use to search for STD-related information. Google Correlate was used to validate search term content. On average STD clinic participant queries were longer compared to student queries. STD clinic participants were more likely to report using search terms that were related to symptomatology such as describing symptoms of STDs, while students were more likely to report searching for general information. These differences in search terms by subpopulation have implications for STD surveillance in populations at most risk for disease acquisition.

Trends in internet search activity, media coverage, and patient-centered health information after the FDA safety communications on surgical mesh for pelvic organ prolapse

INTRODUCTION

In July 2011, the US Food and Drug Administration (FDA) issued a safety communication regarding serious complications associated with surgical mesh for pelvic organ prolapse, prompting increased media and public attention. This study sought to analyze internet search activity and news article volume after this FDA warning and to evaluate the quality of websites providing patient-centered information.

METHODS

Google Trends™ was utilized to evaluate search engine trends for the term "pelvic organ prolapse" and associated terms between 1 January 2004 and 31 December 2014. Google News™ was utilized to quantify the number of news articles annually under the term "pelvic organ prolapse." The search results for the term "pelvic organ prolapse" were assessed for quality using the Health On the Net Foundation (HON) certification.

RESULTS

There was a significant increase in search activity from 37.42 in 2010 to 57.75 in 2011, at the time of the FDA communication (p = 0.021). No other annual interval had a statistically significant increase in search activity. The single highest monthly search activity, given the value of 100, was August 2011, immediately following the July 2011 notification, with the next highest value being 98 in July 2011. Linear regression analysis of news articles per year since the FDA communication revealed r² = 0.88, with a coefficient of 186. Quality assessment demonstrated that 42 % of websites were HON-certified, with .gov sites providing the highest quality information.

CONCLUSIONS

Although the 2011 FDA safety communication on surgical mesh was associated with increased public and media attention, the quality of relevant health information on the internet remains of poor quality. Future quality assurance measures may be critical in enabling patients to play active roles in their own healthcare.

Use of Internet Search Queries to Enhance Surveillance of Foodborne Illness

“Food poisoning” queries were correlated with the number of foodborne illness–related hospital stays.

As a supplement to or extension of methods used to determine trends in foodborne illness over time, we propose the use of Internet search metrics. We compared Internet query data for foodborne illness syndrome–related search terms from the most popular 5 Korean search engines using Health Insurance Review and Assessment Service inpatient stay data for 26 International Classification of Diseases, Tenth Revision, codes for foodborne illness in South Korea during 2010–2012. We used time-series analysis with Seasonal Autoregressive Integrated Moving Average (SARIMA) models. Internet search queries for “food poisoning” correlated most strongly with foodborne illness data (r = 0.70, p<0.001); furthermore, “food poisoning” queries correlated most strongly with the total number of inpatient stays related to foodborne illness during the next month (β = 0.069, SE 0.017, p<0.001). This approach, using the SARIMA model, could be used to effectively measure trends over time to enhance surveillance of foodborne illness in South Korea.

Internet-based media coverage on dengue in Sri Lanka between 2007 and 2015

Background

Internet-based media coverage to explore the extent of awareness of a disease and perceived severity of an outbreak at a national level can be used for early outbreak detection. Dengue has emerged as a major public health problem in Sri Lanka since 2009.

Objective

To compare Internet references to dengue in Sri Lana with references to other diseases (malaria and influenza) in Sri Lanka and to compare Internet references to dengue in Sri Lanka with notified cases of dengue in Sri Lanka.

Design

We examined Internet-based news media articles on dengue queried from HealthMap for Sri Lanka, for the period January 2007 to November 2015. For comparative purposes, we compared hits on dengue with hits on influenza and malaria.

Results

There were 565 hits on dengue between 2007 and 2015, with a rapid rise in 2009 and followed by a rising trend ever since. These hits were highly correlated with the national epidemiological trend of dengue. The volume of digital media coverage of dengue was much higher than of influenza and malaria.

Conclusions

Dengue in Sri Lanka is receiving increasing media attention. Our findings underpin previous claims that digital media reports reflect national epidemiological trends, both in annual trends and inter-annual seasonal variation, thus acting as proxy biosurveillance to provide early warning and situation awareness of emerging infectious diseases.

"Googling" for Cancer: An Infodemiological Assessment of Online Search Interests in Australia, Canada, New Zealand, the United Kingdom, and the United States

Background

The infodemiological analysis of queries from search engines to shed light on the status of various noncommunicable diseases has gained increasing popularity in recent years.

Objective

The aim of the study was to determine the international perspective on the distribution of information seeking in Google regarding “cancer” in major English-speaking countries.

Methods

We used Google Trends service to assess people’s interest in searching about “Cancer” classified as “Disease,” from January 2004 to December 2015 in Australia, Canada, New Zealand, the United Kingdom, and the United States. Then, we evaluated top cities and their relative search volumes (SVs) and country-specific “Top searches” and “Rising searches.” We also evaluated the cross-country correlations of SVs for cancer, as well as rank correlations of SVs from 2010 to 2014 with the incidence of cancer in 2012 in the abovementioned countries.

Results

From 2004 to 2015, the United States (relative SV [from 100]: 63), Canada (62), and Australia (61) were the top countries searching for cancer in Google, followed by New Zealand (54) and the United Kingdom (48). There was a consistent seasonality pattern in searching for cancer in the United States, Canada, Australia, and New Zealand. Baltimore (United States), St John’s (Canada), Sydney (Australia), Otaika (New Zealand), and Saint Albans (United Kingdom) had the highest search interest in their corresponding countries. “Breast cancer” was the cancer entity that consistently appeared high in the list of top searches in all 5 countries. The “Rising searches” were “pancreatic cancer” in Canada and “ovarian cancer” in New Zealand. Cross-correlation of SVs was strong between the United States, Canada, and Australia (>.70, P<.01).

Conclusions

Cancer maintained its popularity as a search term for people in the United States, Canada, and Australia, comparably higher than New Zealand and the United Kingdom. The increased interest in searching for keywords related to cancer shows the possible effectiveness of awareness campaigns in increasing societal demand for health information on the Web, to be met in community-wide communication or awareness interventions.

Tracking search engine queries for suicide in the United Kingdom, 2004-2013

OBJECTIVES

First, to determine if a cyclical trend is observed for search activity of suicide and three common suicide risk factors in the United Kingdom: depression, unemployment, and marital strain. Second, to test the validity of suicide search data as a potential marker of suicide risk by evaluating whether web searches for suicide associate with suicide rates among those of different ages and genders in the United Kingdom.

STUDY DESIGN

Cross-sectional.

METHODS

Search engine data was obtained from Google Trends, a publicly available repository of information of trends and patterns of user searches on Google. The following phrases were entered into Google Trends to analyse relative search volume for suicide, depression, job loss, and divorce, respectively: 'suicide'; 'depression + depressed + hopeless'; 'unemployed + lost job'; 'divorce'. Spearman's rank correlation coefficient was employed to test bivariate associations between suicide search activity and official suicide rates from the Office of National Statistics (ONS).

RESULTS

Cyclical trends were observed in search activity for suicide and depression-related search activity, with peaks in autumn and winter months, and a trough in summer months. A positive, non-significant association was found between suicide-related search activity and suicide rates in the general working-age population (15-64 years) (ρ = 0.164; P = 0.652). This association is stronger in younger age groups, particularly for those 25-34 years of age (ρ = 0.848; P = 0.002).

CONCLUSIONS

We give credence to a link between search activity for suicide and suicide rates in the United Kingdom from 2004 to 2013 for high risk sub-populations (i.e. male youth and young professionals). There remains a need for further research on how Google Trends can be used in other areas of disease surveillance and for work to provide greater geographical precision, as well as research on ways of mitigating the risk of internet use leading to suicide ideation in youth.

Comparative analysis of web search trends between experts and public for medicinal herbs in Korea

ETHNOPHARMACOLOGICAL RELEVANCE

The information and knowledge about ethno-medicinal herbs are getting stronger interest in Global and Korea after the agreement of the Nagoya Protocol. However, it is known that there is a serious asymmetry of ethno-medicinal information between experts and public, thus this study aimed to analyze the similarities and differences in interest between experts and public for medicinal herbs in Korea through big data analysis.

MATERIALS AND METHODS

The medicinal herbs selected in this study were the top 10 herbs in terms of the amounts purchased by TKM centers. And two representative web search engines were selected to collect the web search logs, i.e. big data, of experts and public for medicinal herbs in Korea. Comparative analysis was accomplished through descriptive statistical analysis, Pearson's correlation coefficient, and time-series graph analysis.

RESULTS

The web search traffic logs were collected for the past three years (2012-2014) from OASIS and NAVER, which are the representative web search engines of experts and public respectively in Korea. First, regarding OASIS, the most searched medicinal herb was Angelicae Gigantis Radix while the least searched was Alismatis Rhizoma; for NAVER, the most searched medicinal herb was Paeoniae Radix, unlike OASIS, and the least searched was Alismatis Rhizoma, as with OASIS. The coefficient between rank of herbs and OASIS was -0.401, and that between rank of herbs and NAVER was -0.387, and the correlational coefficient for web search trends of OASIS and NAVER during the past three years was 0.438. Also the correlation of interest between experts and public for each herb on monthly web trends basis was similar with regard to Glycyrrhizae Radix et Rhizoma and Angelicae Gigantis Radix, but different with regard to the other 8 medicinal herbs. Finally, significant outcomes or suggestions were figured out through time-series graph analysis.

CONCLUSION

This study presents meaningful results concerning the similarities and differences in interest between experts and public for popular medicinal herbs in Korea. Contrary to the common assumption that there exists big different interest between experts and public, this study revealed that there is apparent similar interest between experts and public on popular medicinal herbs in Korea.

Web search activity data accurately predict population chronic disease risk in the USA

BACKGROUND

The WHO framework for non-communicable disease (NCD) describes risks and outcomes comprising the majority of the global burden of disease. These factors are complex and interact at biological, behavioural, environmental and policy levels presenting challenges for population monitoring and intervention evaluation. This paper explores the utility of machine learning methods applied to population-level web search activity behaviour as a proxy for chronic disease risk factors.

METHODS

Web activity output for each element of the WHO's Causes of NCD framework was used as a basis for identifying relevant web search activity from 2004 to 2013 for the USA. Multiple linear regression models with regularisation were used to generate predictive algorithms, mapping web search activity to Centers for Disease Control and Prevention (CDC) measured risk factor/disease prevalence. Predictions for subsequent target years not included in the model derivation were tested against CDC data from population surveys using Pearson correlation and Spearman's r.

RESULTS

For 2011 and 2012, predicted prevalence was very strongly correlated with measured risk data ranging from fruits and vegetables consumed (r=0.81; 95% CI 0.68 to 0.89) to alcohol consumption (r=0.96; 95% CI 0.93 to 0.98). Mean difference between predicted and measured differences by State ranged from 0.03 to 2.16. Spearman's r for state-wise predicted versus measured prevalence varied from 0.82 to 0.93.

CONCLUSIONS

The high predictive validity of web search activity for NCD risk has potential to provide real-time information on population risk during policy implementation and other population-level NCD prevention efforts.

Using internet search queries for infectious disease surveillance: screening diseases for suitability

BACKGROUND

Internet-based surveillance systems provide a novel approach to monitoring infectious diseases. Surveillance systems built on internet data are economically, logistically and epidemiologically appealing and have shown significant promise. The potential for these systems has increased with increased internet availability and shifts in health-related information seeking behaviour. This approach to monitoring infectious diseases has, however, only been applied to single or small groups of select diseases. This study aims to systematically investigate the potential for developing surveillance and early warning systems using internet search data, for a wide range of infectious diseases.

METHODS

Official notifications for 64 infectious diseases in Australia were downloaded and correlated with frequencies for 164 internet search terms for the period 2009-13 using Spearman's rank correlations. Time series cross correlations were performed to assess the potential for search terms to be used in construction of early warning systems.

RESULTS

Notifications for 17 infectious diseases (26.6%) were found to be significantly correlated with a selected search term. The use of internet metrics as a means of surveillance has not previously been described for 12 (70.6%) of these diseases. The majority of diseases identified were vaccine-preventable, vector-borne or sexually transmissible; cross correlations, however, indicated that vector-borne and vaccine preventable diseases are best suited for development of early warning systems.

CONCLUSIONS

The findings of this study suggest that internet-based surveillance systems have broader applicability to monitoring infectious diseases than has previously been recognised. Furthermore, internet-based surveillance systems have a potential role in forecasting emerging infectious disease events, especially for vaccine-preventable and vector-borne diseases.

Improving Google Flu Trends estimates for the United States through transformation

Google Flu Trends (GFT) uses Internet search queries in an effort to provide early warning of increases in influenza-like illness (ILI). In the United States, GFT estimates the percentage of physician visits related to ILI (%ILINet) reported by the Centers for Disease Control and Prevention (CDC). However, during the 2012-13 influenza season, GFT overestimated %ILINet by an appreciable amount and estimated the peak in incidence three weeks late. Using data from 2010-14, we investigated the relationship between GFT estimates (%GFT) and %ILINet. Based on the relationship between the relative change in %GFT and the relative change in %ILINet, we transformed %GFT estimates to better correspond with %ILINet values. In 2010-13, our transformed %GFT estimates were within ± 10% of %ILINet values for 17 of the 29 weeks that %ILINet was above the seasonal baseline value determined by the CDC; in contrast, the original %GFT estimates were within ± 10% of %ILINet values for only two of these 29 weeks. Relative to the %ILINet peak in 2012-13, the peak in our transformed %GFT estimates was 2% lower and one week later, whereas the peak in the original %GFT estimates was 74% higher and three weeks later. The same transformation improved %GFT estimates using the recalibrated 2013 GFT model in early 2013-14. Our transformed %GFT estimates can be calculated approximately one week before %ILINet values are reported by the CDC and the transformation equation was stable over the time period investigated (2010-13). We anticipate our results will facilitate future use of GFT.

Risk assessment of vector-borne diseases for public health governance

OBJECTIVES

In the context of public health, risk governance (or risk analysis) is a framework for the assessment and subsequent management and/or control of the danger posed by an identified disease threat. Generic frameworks in which to carry out risk assessment have been developed by various agencies. These include monitoring, data collection, statistical analysis and dissemination. Due to the inherent complexity of disease systems, however, the generic approach must be modified for individual, disease-specific risk assessment frameworks.

STUDY DESIGN

The analysis was based on the review of the current risk assessments of vector-borne diseases adopted by the main Public Health organisations (OIE, WHO, ECDC, FAO, CDC etc…).

METHODS

Literature, legislation and statistical assessment of the risk analysis frameworks.

RESULTS

This review outlines the need for the development of a general public health risk assessment method for vector-borne diseases, in order to guarantee that sufficient information is gathered to apply robust models of risk assessment. Stochastic (especially spatial) methods, often in Bayesian frameworks are now gaining prominence in standard risk assessment procedures because of their ability to assess accurately model uncertainties.

CONCLUSIONS

Risk assessment needs to be addressed quantitatively wherever possible, and submitted with its quality assessment in order to enable successful public health measures to be adopted. In terms of current practice, often a series of different models and analyses are applied to the same problem, with results and outcomes that are difficult to compare because of the unknown model and data uncertainties. Therefore, the risk assessment areas in need of further research are identified in this article.

Global disease monitoring and forecasting with Wikipedia

Infectious disease is a leading threat to public health, economic stability, and other key social structures. Efforts to mitigate these impacts depend on accurate and timely monitoring to measure the risk and progress of disease. Traditional, biologically-focused monitoring techniques are accurate but costly and slow; in response, new techniques based on social internet data, such as social media and search queries, are emerging. These efforts are promising, but important challenges in the areas of scientific peer review, breadth of diseases and countries, and forecasting hamper their operational usefulness. We examine a freely available, open data source for this use: access logs from the online encyclopedia Wikipedia. Using linear models, language as a proxy for location, and a systematic yet simple article selection procedure, we tested 14 location-disease combinations and demonstrate that these data feasibly support an approach that overcomes these challenges. Specifically, our proof-of-concept yields models with r2 up to 0.92, forecasting value up to the 28 days tested, and several pairs of models similar enough to suggest that transferring models from one location to another without re-training is feasible. Based on these preliminary results, we close with a research agenda designed to overcome these challenges and produce a disease monitoring and forecasting system that is significantly more effective, robust, and globally comprehensive than the current state of the art.

The use of google trends in health care research: a systematic review

BACKGROUND

Google Trends is a novel, freely accessible tool that allows users to interact with Internet search data, which may provide deep insights into population behavior and health-related phenomena. However, there is limited knowledge about its potential uses and limitations. We therefore systematically reviewed health care literature using Google Trends to classify articles by topic and study aim; evaluate the methodology and validation of the tool; and address limitations for its use in research.

METHODS AND FINDINGS

PRISMA guidelines were followed. Two independent reviewers systematically identified studies utilizing Google Trends for health care research from MEDLINE and PubMed. Seventy studies met our inclusion criteria. Google Trends publications increased seven-fold from 2009 to 2013. Studies were classified into four topic domains: infectious disease (27% of articles), mental health and substance use (24%), other non-communicable diseases (16%), and general population behavior (33%). By use, 27% of articles utilized Google Trends for casual inference, 39% for description, and 34% for surveillance. Among surveillance studies, 92% were validated against a reference standard data source, and 80% of studies using correlation had a correlation statistic ≥0.70. Overall, 67% of articles provided a rationale for their search input. However, only 7% of articles were reproducible based on complete documentation of search strategy. We present a checklist to facilitate appropriate methodological documentation for future studies. A limitation of the study is the challenge of classifying heterogeneous studies utilizing a novel data source.

CONCLUSION

Google Trends is being used to study health phenomena in a variety of topic domains in myriad ways. However, poor documentation of methods precludes the reproducibility of the findings. Such documentation would enable other researchers to determine the consistency of results provided by Google Trends for a well-specified query over time. Furthermore, greater transparency can improve its reliability as a research tool.

Online reports of foodborne illness capture foods implicated in official foodborne outbreak reports

OBJECTIVE

Traditional surveillance systems capture only a fraction of the estimated 48 million yearly cases of foodborne illness in the United States. We assessed whether foodservice reviews on Yelp.com (a business review site) can be used to support foodborne illness surveillance efforts.

METHODS

We obtained reviews from 2005 to 2012 of 5824 foodservice businesses closest to 29 colleges. After extracting recent reviews describing episodes of foodborne illness, we compared implicated foods to foods in outbreak reports from the U.S. Centers for Disease Control and Prevention (CDC).

RESULTS

Broadly, the distribution of implicated foods across five categories was as follows: aquatic (16% Yelp, 12% CDC), dairy-eggs (23% Yelp, 23% CDC), fruits-nuts (7% Yelp, 7% CDC), meat-poultry (32% Yelp, 33% CDC), and vegetables (22% Yelp, 25% CDC). The distribution of foods across 19 more specific food categories was also similar, with Spearman correlations ranging from 0.60 to 0.85 for 2006-2011. The most implicated food categories in both Yelp and CDC were beef, dairy, grains-beans, poultry and vine-stalk.

CONCLUSIONS

Based on observations in this study and the increased usage of social media, we posit that online illness reports could complement traditional surveillance systems by providing near real-time information on foodborne illnesses, implicated foods and locations.

The representativeness of a European multi-center network for influenza-like-illness participatory surveillance

Background

The Internet is becoming more commonly used as a tool for disease surveillance. Similarly to other surveillance systems and to studies using online data collection, Internet-based surveillance will have biases in participation, affecting the generalizability of the results. Here we quantify the participation biases of Influenzanet, an ongoing European-wide network of Internet-based participatory surveillance systems for influenza-like-illness.

Methods

In 2011/2012 Influenzanet launched a standardized common framework for data collection applied to seven European countries. Influenzanet participants were compared to the general population of the participating countries to assess the representativeness of the sample in terms of a set of demographic, geographic, socio-economic and health indicators.

Results

More than 30,000 European residents registered to the system in the 2011/2012 season, and a subset of 25,481 participants were selected for this study. All age classes (10 years brackets) were represented in the cohort, including under 10 and over 70 years old. The Influenzanet population was not representative of the general population in terms of age distribution, underrepresenting the youngest and oldest age classes. The gender imbalance differed between countries. A counterbalance between gender-specific information-seeking behavior (more prominent in women) and Internet usage (with higher rates in male populations) may be at the origin of this difference. Once adjusted by demographic indicators, a similar propensity to commute was observed for each country, and the same top three transportation modes were used for six countries out of seven. Smokers were underrepresented in the majority of countries, as were individuals with diabetes; the representativeness of asthma prevalence and vaccination coverage for 65+ individuals in two successive seasons (2010/2011 and 2011/2012) varied between countries.

Conclusions

Existing demographic and national datasets allowed the quantification of the participation biases of a large cohort for influenza-like-illness surveillance in the general population. Significant differences were found between Influenzanet participants and the general population. The quantified biases need to be taken into account in the analysis of Influenzanet epidemiological studies and provide indications on populations groups that should be targeted in recruitment efforts.

Electronic supplementary material

The online version of this article (doi:10.1186/1471-2458-14-984) contains supplementary material, which is available to authorized users.

Global search demand for varicose vein information on the internet

BACKGROUND

Changes in internet search trends can provide healthcare professionals detailed information on prevalence of disease and symptoms. Chronic venous disease, more commonly known as varicose veins, is a common symptomatic disease among the adult population. This study aims to measure the change in global search demand for varicose vein information using Google over the past 8 years.

METHODS

The Google Trends instrument was used to measure the change in demand for the use of the local name for varicose veins in several countries across the world between January 2006 and December 2012. The measurements were normalised onto a scale relative to the largest volume of search requests received during a designated time and geographical location. Comparison of national levels of private healthcare and healthcare spending per capita to search demand was also undertaken using Organisation for Economic Co-operation and development economic measurements.

RESULTS

Global interest has increased significantly, with linear regression demonstrating a 3.72% year-on-year increase in demand over the 8-year time period (r(2 )= 0.385, p < 0.001). Annual demand significantly increased in the northern hemisphere (p < 0.001 Friedman) yet decreased in the southern hemisphere (p < 0.001 Friedman). Significant seasonality was observed, with warmer months experiencing greater search demand compared to cooler winter months (<0.001 Kruskal-Wallis). National levels of private healthcare did not appear to correlate in search demand (r(2 )= 0.120 p = 0.306). Healthcare spending per capita did not relate to search demand (r(2 )= 0.450 p = 0.077).

CONCLUSION

There is increasing demand for information about varicose veins on the internet, especially during the warmer months of the year. Online search demand does not appear to be related to healthcare spending.

Detecting the norovirus season in Sweden using search engine data--meeting the needs of hospital infection control teams

Norovirus outbreaks severely disrupt healthcare systems. We evaluated whether Websök, an internet-based surveillance system using search engine data, improved norovirus surveillance and response in Sweden. We compared Websök users' characteristics with the general population, cross-correlated weekly Websök searches with laboratory notifications between 2006 and 2013, compared the time Websök and laboratory data crossed the epidemic threshold and surveyed infection control teams about their perception and use of Websök. Users of Websök were not representative of the general population. Websök correlated with laboratory data (b = 0.88-0.89) and gave an earlier signal to the onset of the norovirus season compared with laboratory-based surveillance. 17/21 (81%) infection control teams answered the survey, of which 11 (65%) believed Websök could help with infection control plans. Websök is a low-resource, easily replicable system that detects the norovirus season as reliably as laboratory data, but earlier. Using Websök in routine surveillance can help infection control teams prepare for the yearly norovirus season.

Public health and pipe breaks in water distribution systems: analysis with internet search volume as a proxy

Drinking water distribution infrastructure has been identified as a factor in waterborne disease outbreaks and improved understanding of the public health risks associated with distribution system failures has been identified as a priority area for research. Pipe breaks may pose a risk, as their occurrence and repair can result in low or negative pressure, potentially allowing contamination of drinking water from adjacent soils. However, measuring this phenomenon is challenging because the most likely health impact is mild gastrointestinal (GI) illness, which is unlikely to result in a doctor or hospital visit. Here we present a novel method that uses data mining techniques and internet search volume to assess the relationship between pipe breaks and symptoms of GI illness in two U.S. cities. Weekly search volume for the terms diarrhea and vomiting was used as the response variable with the number of pipe breaks in each city as a covariate as well as additional covariates to control for seasonal patterns, search volume persistence, and other sources of GI illness. The fit and predictive accuracy of multiple regression and data mining techniques were compared, with the best performance obtained using random forest and bagged regression tree models. Pipe breaks were found to be an important and positively correlated predictor of internet search volume in multiple models in both cities, supporting previous investigations that indicated an increased risk of GI illness from distribution system disturbances.

The frequency and distribution of canine leishmaniosis diagnosed by veterinary practitioners in Europe

This study aimed to evaluate the frequency and spatial distribution of canine leishmaniosis (CanL) in France, Greece, Italy, Portugal and Spain. An online questionnaire investigated the location and frequency of CanL cases diagnosed by veterinary practitioners. Further data from the practice management systems of veterinary clinics in France were provided by a financial benchmarking company in relation to all treatment and test invoice data from participating practices. The geographical and temporal web interest in leishmaniosis was explored using Google Trends. Veterinary practitioners from France, Greece, Italy, Portugal and Spain completed 1231 questionnaires. The percentage of practice-attending dogs with a veterinary diagnosis of CanL ranged from 0.71% in France to 7.80% in Greece. However, due to regional differences in response rates, particularly in France, the mean regional estimates may better reflect the disease burden. Benchmarking data relating to approximately 180,000 dogs estimated that 0.05% of dogs attending veterinary clinics were treated for CanL or euthanased with suspected CanL in France. The regional frequency of Google web queries for leishmaniosis generally reflected the spatial patterns of disease identified from the other data sources. In conclusion, CanL was a relatively common diagnosis in veterinary clinics in many regions of the countries studied. Knowledge of CanL in endemic areas can direct the use of preventative measures and help estimate the likelihood of infection in dogs visiting or inhabiting these countries.

Does Saint Nicholas provoke seizures? Hints from Google Trends

PURPOSE

Stress is the most often reported seizure-precipitant in epilepsy. As most evidence for the relation between stress and epilepsy is derived from human self-reports, observational studies including a larger part of the population could provide additional proof. A stressor often reported to increase seizure frequency in children with epilepsy in the Netherlands is the national celebration of Saint Nicholas' eve (December 5) and the weeks before; this is the main period of festivities for children in this country. To study the relation between stress and epilepsy, we analyzed epilepsy information-seeking behavior on the Internet, an indirect measure of seizure frequency, around this national children's celebration.

METHODS

Google Trends was used to extract relative search percentages for 'epilepsy' on Google in the Netherlands, the United States, and the United Kingdom between 2004 and 2013. Relative search percentages during the Saint Nicholas period were compared with baseline.

RESULTS

Epilepsy searches increased by 14% in the Saint Nicholas period compared with baseline (p<0.001). This effect was not found for searches performed in the same period in the United States or the United Kingdom, countries where this holiday is not celebrated.

CONCLUSIONS

The increase in epilepsy information-seeking behavior in the Saint Nicholas period is possibly caused by an increased occurrence of epileptic seizures. This underscores the potential of health information-seeking behavior on the Internet to answer clinically relevant research questions and provides circumstantial evidence for a relation between stress and the occurrence of epileptic seizures.

Internet-based surveillance systems for monitoring emerging infectious diseases

Emerging infectious diseases present a complex challenge to public health officials and governments; these challenges have been compounded by rapidly shifting patterns of human behaviour and globalisation. The increase in emerging infectious diseases has led to calls for new technologies and approaches for detection, tracking, reporting, and response. Internet-based surveillance systems offer a novel and developing means of monitoring conditions of public health concern, including emerging infectious diseases. We review studies that have exploited internet use and search trends to monitor two such diseases: influenza and dengue. Internet-based surveillance systems have good congruence with traditional surveillance approaches. Additionally, internet-based approaches are logistically and economically appealing. However, they do not have the capacity to replace traditional surveillance systems; they should not be viewed as an alternative, but rather an extension. Future research should focus on using data generated through internet-based surveillance and response systems to bolster the capacity of traditional surveillance systems for emerging infectious diseases.

Gauging interest of the general public in laser-assisted in situ keratomileusis eye surgery

PURPOSE

To assess interest among members of the general public in laser-assisted in situ keratomileusis (LASIK) surgery and how levels of interest in this procedure have changed over time in the United States and other countries.

METHODS

Using the Google Trends Web site, we determined the weekly frequency of queries involving the term "LASIK" from January 1, 2007, through January 1, 2011, in the United States, United Kingdom, Canada, and India. We fit separate regression models for each of the countries to assess whether residents of these countries differed in their querying rates on specific dates and over time. Similar analyses were performed to compare 4 US states. Additional regression models compared general public interest in LASIK surgery before and after the release of a 2008 Food and Drug Administration report describing complaints associated with this procedure.

RESULTS

During 2007 to 2011, the Google query rate for "LASIK" was highest among persons residing in India, followed by the United Kingdom, Canada, and the United States. During this time period, the query rate declined by 40% in the United States, 24% in India, and 22% in the United Kingdom, and it increased by 8% in Canada. In all 4 of the US states examined, the query rate declined-by 52% in Florida, 56% in New York, 54% in Texas, and 42% in California. Interest in LASIK declined further among US citizens after the Food and Drug Administration report release.

CONCLUSIONS

Interest among the general public in LASIK surgery has been waning in recent years.

Chinese social media reaction to the MERS-CoV and avian influenza A(H7N9) outbreaks

Background

As internet and social media use have skyrocketed, epidemiologists have begun to use online data such as Google query data and Twitter trends to track the activity levels of influenza and other infectious diseases. In China, Weibo is an extremely popular microblogging site that is equivalent to Twitter. Capitalizing on the wealth of public opinion data contained in posts on Weibo, this study used Weibo as a measure of the Chinese people’s reactions to two different outbreaks: the 2012 Middle East Respiratory Syndrome Coronavirus (MERS-CoV) outbreak, and the 2013 outbreak of human infection of avian influenza A(H7N9) in China.

Methods

Keyword searches were performed in Weibo data collected by The University of Hong Kong’s Weiboscope project. Baseline values were determined for each keyword and reaction values per million posts in the days after outbreak information was released to the public.

Results

The results show that the Chinese people reacted significantly to both outbreaks online, where their social media reaction was two orders of magnitude stronger to the H7N9 influenza outbreak that happened in China than the MERS-CoV outbreak that was far away from China.

Conclusions

These results demonstrate that social media could be a useful measure of public awareness and reaction to disease outbreak information released by health authorities.

Correlation between national influenza surveillance data and google trends in South Korea

BACKGROUND

In South Korea, there is currently no syndromic surveillance system using internet search data, including Google Flu Trends. The purpose of this study was to investigate the correlation between national influenza surveillance data and Google Trends in South Korea.

METHODS

Our study was based on a publicly available search engine database, Google Trends, using 12 influenza-related queries, from September 9, 2007 to September 8, 2012. National surveillance data were obtained from the Korea Centers for Disease Control and Prevention (KCDC) influenza-like illness (ILI) and virologic surveillance system. Pearson's correlation coefficients were calculated to compare the national surveillance and the Google Trends data for the overall period and for 5 influenza seasons.

RESULTS

The correlation coefficient between the KCDC ILI and virologic surveillance data was 0.72 (p<0.05). The highest correlation was between the Google Trends query of H1N1 and the ILI data, with a correlation coefficient of 0.53 (p<0.05), for the overall study period. When compared with the KCDC virologic data, the Google Trends query of bird flu had the highest correlation with a correlation coefficient of 0.93 (p<0.05) in the 2010-11 season. The following queries showed a statistically significant correlation coefficient compared with ILI data for three consecutive seasons: Tamiflu (r = 0.59, 0.86, 0.90, p<0.05), new flu (r = 0.64, 0.43, 0.70, p<0.05) and flu (r = 0.68, 0.43, 0.77, p<0.05).

CONCLUSIONS

In our study, we found that the Google Trends for certain queries using the survey on influenza correlated with national surveillance data in South Korea. The results of this study showed that Google Trends in the Korean language can be used as complementary data for influenza surveillance but was insufficient for the use of predictive models, such as Google Flu Trends.

Web-search trends shed light on the nature of lunacy: relationship between moon phases and epilepsy information-seeking behavior

In old and modern times and across cultures, recurrent seizures have been attributed to the lunar phase. It is unclear whether this relationship should be classified as a myth or whether a true connection exists between moon phases and seizures. We analyzed the worldwide aggregated search queries related to epilepsy health-seeking behavior between 2005 and 2012. Epilepsy-related Internet searches increased in periods with a high moon illumination. The overall association was weak (r=0.11, 95% confidence interval: 0.07 to 0.14) but seems to be higher than most control search queries not related to epilepsy. Increased sleep deprivation during periods of full moon might explain this positive association and warrants further study into epilepsy-related health-seeking behavior on the Internet, the lunar phase, and its contribution to nocturnal luminance.

Using search queries for malaria surveillance, Thailand

Background

Internet search query trends have been shown to correlate with incidence trends for select infectious diseases and countries. Herein, the first use of Google search queries for malaria surveillance is investigated. The research focuses on Thailand where real-time malaria surveillance is crucial as malaria is re-emerging and developing resistance to pharmaceuticals in the region.

Methods

Official Thai malaria case data was acquired from the World Health Organization (WHO) from 2005 to 2009. Using Google correlate, an openly available online tool, and by surveying Thai physicians, search queries potentially related to malaria prevalence were identified. Four linear regression models were built from different sub-sets of malaria-related queries to be used in future predictions. The models’ accuracies were evaluated by their ability to predict the malaria outbreak in 2009, their correlation with the entire available malaria case data, and by Akaike information criterion (AIC).

Results

Each model captured the bulk of the variability in officially reported malaria incidence. Correlation in the validation set ranged from 0.75 to 0.92 and AIC values ranged from 808 to 586 for the models. While models using malaria-related and general health terms were successful, one model using only microscopy-related terms obtained equally high correlations to malaria case data trends. The model built strictly of queries provided by Thai physicians was the only one that consistently captured the well-documented second seasonal malaria peak in Thailand.

Conclusions

Models built from Google search queries were able to adequately estimate malaria activity trends in Thailand, from 2005–2010, according to official malaria case counts reported by WHO. While presenting their own limitations, these search queries may be valid real-time indicators of malaria incidence in the population, as correlations were on par with those of related studies for other infectious diseases. Additionally, this methodology provides a cost-effective description of malaria prevalence that can act as a complement to traditional public health surveillance. This and future studies will continue to identify ways to leverage web-based data to improve public health.

A method to assess seasonality of urinary tract infections based on medication sales and google trends

BACKGROUND

Despite the fact that urinary tract infection (UTI) is a very frequent disease, little is known about its seasonality in the community.

METHODS AND FINDINGS

To estimate seasonality of UTI using multiple time series constructed with available proxies of UTI. Eight time series based on two databases were used: sales of urinary antibacterial medications reported by a panel of pharmacy stores in France between 2000 and 2012, and search trends on the Google search engine for UTI-related terms between 2004 and 2012 in France, Germany, Italy, the USA, China, Australia and Brazil. Differences between summers and winters were statistically assessed with the Mann-Whitney test. We evaluated seasonality by applying the Harmonics Product Spectrum on Fast Fourier Transform. Seven time series out of eight displayed a significant increase in medication sales or web searches in the summer compared to the winter, ranging from 8% to 20%. The eight time series displayed a periodicity of one year. Annual increases were seen in the summer for UTI drug sales in France and Google searches in France, the USA, Germany, Italy, and China. Increases occurred in the austral summer for Google searches in Brazil and Australia.

CONCLUSIONS

An annual seasonality of UTIs was evidenced in seven different countries, with peaks during the summer.

Bariatric surgery interest around the world: what Google Trends can teach us

BACKGROUND

Bariatric surgery may prove an effective weight loss option for those struggling with severe obesity, but it is difficult to determine levels of interest in such procedures at the population level through traditional approaches. Analysis of Google Trend information may give providers and healthcare systems useful information regarding Internet users' interest in bariatric procedures. The objective of this study was to gather Google Trend information on worldwide Internet searches for "bariatric surgery", "gastric bypass", "gastric sleeve", "gastric plication", and "lap band" from 2004-2012 and to explore temporal relationships with relevant media events, economic variations, and policy modifications.

METHODS

Data were collected using Google Trends. Trend analyses were performed using Microsoft Excel Version 14.3.5 and Minitab V.16.0.

RESULTS

Trend analyses showed that total search volume for the term "bariatric surgery" has declined roughly 25% since January 2004, although interest increased approximately 5% from 2011 to 2012. Interest in lap band procedures declined 30% over the past 5 years, while "gastric sleeve" has increased 15%. Spikes in search numbers show an association with events such as changing policy and insurance guidelines and media coverage for bariatric procedures.

CONCLUSION

This report illustrates that variations in Internet search volume for terms related to bariatric surgery are multifactorial in origin. Although it is impossible to ascertain if reported Internet search volume is based on interest in potentially undergoing bariatric surgery or simply general interest, this analysis reveals that search volume appears to mirror real world events. Therefore, Google Trends could be a way to supplement understanding about interest in bariatric procedures.

Reassessing Google Flu Trends data for detection of seasonal and pandemic influenza: a comparative epidemiological study at three geographic scales

The goal of influenza-like illness (ILI) surveillance is to determine the timing, location and magnitude of outbreaks by monitoring the frequency and progression of clinical case incidence. Advances in computational and information technology have allowed for automated collection of higher volumes of electronic data and more timely analyses than previously possible. Novel surveillance systems, including those based on internet search query data like Google Flu Trends (GFT), are being used as surrogates for clinically-based reporting of influenza-like-illness (ILI). We investigated the reliability of GFT during the last decade (2003 to 2013), and compared weekly public health surveillance with search query data to characterize the timing and intensity of seasonal and pandemic influenza at the national (United States), regional (Mid-Atlantic) and local (New York City) levels. We identified substantial flaws in the original and updated GFT models at all three geographic scales, including completely missing the first wave of the 2009 influenza A/H1N1 pandemic, and greatly overestimating the intensity of the A/H3N2 epidemic during the 2012/2013 season. These results were obtained for both the original (2008) and the updated (2009) GFT algorithms. The performance of both models was problematic, perhaps because of changes in internet search behavior and differences in the seasonality, geographical heterogeneity and age-distribution of the epidemics between the periods of GFT model-fitting and prospective use. We conclude that GFT data may not provide reliable surveillance for seasonal or pandemic influenza and should be interpreted with caution until the algorithm can be improved and evaluated. Current internet search query data are no substitute for timely local clinical and laboratory surveillance, or national surveillance based on local data collection. New generation surveillance systems such as GFT should incorporate the use of near-real time electronic health data and computational methods for continued model-fitting and ongoing evaluation and improvement.

Evaluating the feasibility and participants' representativeness of an online nationwide surveillance system for influenza in France

The increasing Internet coverage and the widespread use of digital devices offer the possibility to develop new digital surveillance systems potentially capable to provide important aid to epidemiological and public health monitoring and research. In France, a new nationwide surveillance system for influenza-like illness, GrippeNet.fr, was introduced since the 2011/2012 season based on an online participatory mechanism and open to the general population. We evaluate the recruitment and participation of users to the first pilot season with respect to similar efforts in Europe to assess the feasibility of establishing a participative network of surveillance in France. We further investigate the representativeness of the GrippeNet.fr population along a set of indicators on geographical, demographic, socio-economic and health aspects. Participation was widespread in the country and with rates comparable to other European countries with partnered projects running since a longer time. It was not representative of the general population in terms of age and gender, however all age classes were represented, including the older classes (65+ years old), generally less familiar with the digital world, but considered at high risk for influenza complications. Once adjusted on demographic indicators, the GrippeNet.fr population is found to be more frequently employed, with a higher education level and vaccination rate with respect to the general population. A similar propensity to commute for work to different regions was observed, and no significant difference was found for asthma and diabetes. Results show the feasibility of the system, provide indications to inform adjusted epidemic analyses, and highlight the presence of specific population groups that need to be addressed by targeted communication strategies to achieve a higher representativeness in the following seasons.

Scoping review on search queries and social media for disease surveillance: a chronology of innovation

Background

The threat of a global pandemic posed by outbreaks of influenza H5N1 (1997) and Severe Acute Respiratory Syndrome (SARS, 2002), both diseases of zoonotic origin, provoked interest in improving early warning systems and reinforced the need for combining data from different sources. It led to the use of search query data from search engines such as Google and Yahoo! as an indicator of when and where influenza was occurring. This methodology has subsequently been extended to other diseases and has led to experimentation with new types of social media for disease surveillance.

Objective

The objective of this scoping review was to formally assess the current state of knowledge regarding the use of search queries and social media for disease surveillance in order to inform future work on early detection and more effective mitigation of the effects of foodborne illness.

Methods

Structured scoping review methods were used to identify, characterize, and evaluate all published primary research, expert review, and commentary articles regarding the use of social media in surveillance of infectious diseases from 2002-2011.

Results

Thirty-two primary research articles and 19 reviews and case studies were identified as relevant. Most relevant citations were peer-reviewed journal articles (29/32, 91%) published in 2010-11 (28/32, 88%) and reported use of a Google program for surveillance of influenza. Only four primary research articles investigated social media in the context of foodborne disease or gastroenteritis. Most authors (21/32 articles, 66%) reported that social media-based surveillance had comparable performance when compared to an existing surveillance program. The most commonly reported strengths of social media surveillance programs included their effectiveness (21/32, 66%) and rapid detection of disease (21/32, 66%). The most commonly reported weaknesses were the potential for false positive (16/32, 50%) and false negative (11/32, 34%) results. Most authors (24/32, 75%) recommended that social media programs should primarily be used to support existing surveillance programs.

Conclusions

The use of search queries and social media for disease surveillance are relatively recent phenomena (first reported in 2006). Both the tools themselves and the methodologies for exploiting them are evolving over time. While their accuracy, speed, and cost compare favorably with existing surveillance systems, the primary challenge is to refine the data signal by reducing surrounding noise. Further developments in digital disease surveillance have the potential to improve sensitivity and specificity, passively through advances in machine learning and actively through engagement of users. Adoption, even as supporting systems for existing surveillance, will entail a high level of familiarity with the tools and collaboration across jurisdictions.

Open data in public health surveillance systems: a case study using the French Sentinelles network

INTRODUCTION

Public Health Surveillance (PHS) produces an increasing number of health indicators. Exposing these data is at the core of interoperability; however no standard has yet been adopted for such information on the internet.

METHOD

Here, we compared two approaches to expose data from the French Sentinelles network, an information system focusing on communicable diseases surveillance in the general population. We implemented SDMX-HD (Statistical Data and Metadata Exchange-Health Domain), a standard supported by government agencies to exchange statistical data and OpenData (OData), a general purpose protocol proposed by Microsoft Corp. The same data were described using SDMX-HD (available at http://sdmx.sentiweb.fr) and using OData (http://odata.sentiweb.fr).

DISCUSSION

These two use cases proved the feasibility of opening public health data on the internet, and highlighted difficulties: SDMX, a full-featured solution, encouraged harmonization and reusability, sustainability, but required complex developments and tools; OData was much simpler to implement but required a "from scratch" description and did not encourage reusability. From an end-user perspective, integration in every-day tools is not achieved yet. These two approaches are a first step to interoperability in PHS.

Incidence of online health information search: a useful proxy for public health risk perception

Background

Internet users use search engines to look for information online, including health information. Researchers in medical informatics have found a high correlation of the occurrence of certain search queries and the incidence of certain diseases. Consumers’ search for information about diseases is related to current health status with regard to a disease and to the social environments that shape the public’s attitudes and behaviors.

Objective

This study aimed to investigate the extent to which public health risk perception as demonstrated by online information searches related to a health risk can be explained by the incidence of the health risk and social components of a specific population’s environment. Using an ecological perspective, we suggest that a population’s general concern for a health risk is formed by the incidence of the risk and social (eg, media attention) factors related with the risk.

Methods

We constructed a dataset that included state-level data from 32 states on the incidence of the flu; a number of social factors, such as media attention to the flu; private resources, such as education and health insurance coverage; public resources, such as hospital beds and primary physicians; and utilization of these resources, including inpatient days and outpatient visits. We then explored whether online information searches about the flu (seasonal and pandemic flu) can be predicted using these variables. We used factor analysis to construct indexes for sets of social factors (private resources, public resources). We then applied panel data multiple regression analysis to exploit both time-series and cross-sectional variation in the data over a 7-year period.

Results

Overall, the results provide evidence that the main effects of independent variables—the incidence of the flu (P<.001); social factors, including media attention (P<.001); private resources, including life quality (P<.001) and health lifestyles (P=.009); and public resources, such as hospital care utilization (P=.008) and public health funds (P=.02)—have significant effects on Web searches for queries related to the flu. After controlling for the number of reported disease cases and Internet access rate by state, we estimate the contribution of social factors to the public health risk perception levels by state (R²=23.37%). The interaction effects between flu incidence and social factors for our search terms did not add to the explanatory power of our regression models (R²<1%).

Conclusions

Our study suggests a practical way to measure the public’s health risk perception for certain diseases using online information search volume by state. The social environment influences public risk perception regardless of disease incidence. Thus, monitoring the social variables can be very helpful in being ready to respond to the public’s behavior in dealing with public health threats.

Tweaking and tweeting: exploring Twitter for nonmedical use of a psychostimulant drug (Adderall) among college students

Background

Adderall is the most commonly abused prescription stimulant among college students. Social media provides a real-time avenue for monitoring public health, specifically for this population.

Objective

This study explores discussion of Adderall on Twitter to identify variations in volume around college exam periods, differences across sets of colleges and universities, and commonly mentioned side effects and co-ingested substances.

Methods

Public-facing Twitter status messages containing the term “Adderall” were monitored from November 2011 to May 2012. Tweets were examined for mention of side effects and other commonly abused substances. Tweets from likely students containing GPS data were identified with clusters of nearby colleges and universities for regional comparison.

Results

213,633 tweets from 132,099 unique user accounts mentioned “Adderall.” The number of Adderall tweets peaked during traditional college and university final exam periods. Rates of Adderall tweeters were highest among college and university clusters in the northeast and south regions of the United States. 27,473 (12.9%) mentioned an alternative motive (eg, study aid) in the same tweet. The most common substances mentioned with Adderall were alcohol (4.8%) and stimulants (4.7%), and the most common side effects were sleep deprivation (5.0%) and loss of appetite (2.6%).

Conclusions

Twitter posts confirm the use of Adderall as a study aid among college students. Adderall discussions through social media such as Twitter may contribute to normative behavior regarding its abuse.

Using Google Trends for influenza surveillance in South China

BACKGROUND

Google Flu Trends was developed to estimate influenza activity in many countries; however there is currently no Google Flu Trends or other Internet search data used for influenza surveillance in China.

METHODS AND FINDINGS

Influenza surveillance data from 2008 through 2011 were obtained from provincial CDC influenza-like illness and virological surveillance systems of Guangdong, a province in south China. Internet search data were downloaded from the website of Google Trends. Pearson's correlation coefficients with 95% confidence intervals (95% CI) were calculated to compare surveillance data and internet search trends. The correlation between CDC ILI surveillance and CDC virus surveillance was 0.56 (95% CI: 0.43, 0.66). The strongest correlation was between the Google Trends term of Fever and ILI surveillance with a correlation coefficient of 0.73 (95% CI: 0.66, 0.79). When compared with influenza virological surveillance, the Google Trends term of Influenza A had the strongest correlation with a correlation coefficient of 0.64 (95% CI: 0.43, 0.79) in the 2009 H1N1 influenza pandemic period.

CONCLUSIONS

This study shows that Google Trends in Chinese can be used as a complementary source of data for influenza surveillance in south China. More research in the future should develop new models using search trends in Chinese language to estimate local disease activity and detect early signals of outbreaks.

Novel surveillance of psychological distress during the great recession

BACKGROUND

Economic stressors have been retrospectively associated with net population increases in nonspecific psychological distress (PD). However, no sentinels exist to evaluate contemporaneous associations. Aggregate Internet search query surveillance was used to monitor population changes in PD around the United States' Great Recession.

METHODS

Monthly PD query trends were compared with unemployment, underemployment, homes in delinquency and foreclosure, median home value or sale prices, and S&P 500 trends for 2004-2010. Time series analyses, where economic indicators predicted PD one to seven months into the future, were performed in 2011.

RESULT

PD queries surpassed 1,000,000 per month, of which 300,000 may be attributable to the Great Recession. A one percentage point increase in mortgage delinquencies and foreclosures was associated with a 16% (95%CI, 9-24) increase in PD queries one-month, and 11% (95%CI, 3-18) four months later, in reference to a pre-Great Recession mean. Unemployment and underemployment had similar associations half and one-quarter the intensity. "Anxiety disorder", "what is depression", "signs of depression", "depression symptoms", and "symptoms of depression" were the queries exhibiting the strongest associations with mortgage delinquencies and foreclosures, unemployment or underemployment. Housing prices and S&P 500 trends were not associated with PD queries.

LIMITATIONS

A non-traditional measure of PD was used. It is unclear if actual clinically significant depression or anxiety increased during the Great Recession. Alternative explanations for strong associations between the Great Recession and PD queries, such as media, were explored and rejected.

CONCLUSIONS

Because the economy is constantly changing, this work not only provides a snapshot of recent associations between the economy and PD queries but also a framework and toolkit for real-time surveillance going forward. Health resources, clinician screening patterns, and policy debate may be informed by changes in PD query trends.

Are cannabis prevalence estimates comparable across countries and regions? A cross-cultural validation using search engine query data

BACKGROUND

Prevalence estimation of cannabis use is usually based on self-report data. Although there is evidence on the reliability of this data source, its cross-cultural validity is still a major concern. External objective criteria are needed for this purpose. In this study, cannabis-related search engine query data are used as an external criterion.

METHODS

Data on cannabis use were taken from the 2007 European School Survey Project on Alcohol and Other Drugs (ESPAD). Provincial data came from three Italian nation-wide studies using the same methodology (2006-2008; ESPAD-Italia). Information on cannabis-related search engine query data was based on Google search volume indices (GSI). (1) Reliability analysis was conducted for GSI. (2) Latent measurement models of "true" cannabis prevalence were tested using perceived availability, web-based cannabis searches and self-reported prevalence as indicators. (3) Structure models were set up to test the influences of response tendencies and geographical position (latitude, longitude). In order to test the stability of the models, analyses were conducted on country level (Europe, US) and on provincial level in Italy.

RESULTS

Cannabis-related GSI were found to be highly reliable and constant over time. The overall measurement model was highly significant in both data sets. On country level, no significant effects of response bias indicators and geographical position on perceived availability, web-based cannabis searches and self-reported prevalence were found. On provincial level, latitude had a significant positive effect on availability indicating that perceived availability of cannabis in northern Italy was higher than expected from the other indicators.

CONCLUSION

Although GSI showed weaker associations with cannabis use than perceived availability, the findings underline the external validity and usefulness of search engine query data as external criteria. The findings suggest an acceptable relative comparability of national (provincial) prevalence estimates of cannabis use that are based on a common survey methodology. Search engine query data are a too weak indicator to base prevalence estimations on this source only, but in combination with other sources (waste water analysis, sales of cigarette paper) they may provide satisfactory estimates.