Web Search Data Research Articles

Eczema-related Web Search Data in Sweden: Investigating Search Patterns and the Influence of Weather.

As atopic eczema is triggered by environmental factors, such as temperature, differences in disease burden between and within countries are possible. One method to study this phenomenon is to perform web-search analysis, since the internet is commonly used to retrieve health-related information. This study, investigating the Google search volume regarding eczema in Swedish counties between April 2017 and March 2021, revealed a continuous increase in number of searches and that the search volume was higher in Northern than Southern Sweden. Gotland had the most searches per 100,000 inhabitants. In general, there was a negative correlation between search volume and temperature (r=-0.315, p < 0.001) and hours of sunshine (r=-0.213, p < 0.001), whereas there was a positive association between search volume and wind (r=0.229, p < 0.001). Search engine analysis is a rapid and cost-effective method of examining search behaviour regarding disease among the general population within a country and, thus, can enable the identification of regions with specific interests and needs.

Coronavirus-Related Searches on the Internet Predict COVID-19 Vaccination Rates in the Real World: A Behavioral Immune System Perspective

According to the smoke detector and functional flexibility principles of human behavioral immune system (BIS), the exposure to COVID-19 cues could motivate vaccine uptake. Using the tool of Google Trends, we tested that coronavirus-related searches—which assessed natural exposure to COVID-19 cues—would positively predict actual vaccination rates. As expected, coronavirus-related searches positively and significantly predicted vaccination rates in the United States (Study 1a) and across the globe (Study 2a) after accounting for a range of covariates. The stationary time series analyses with covariates and autocorrelation structure of the dependent variable confirmed that more coronavirus-related searches compared with last week indicated increases in vaccination rates compared with last week in the United States (Study 1b) and across the globe (Study 2b). With real-time web search data, psychological scientists could test their research questions in real-life settings and at a large scale to expand the ecological validity and generalizability of the findings.

Leveraging web search data to explore public interest in psoriasis in Germany

AbstractBackgroundLittle is known about what people are interested in when they search the web for psoriasis. The monitoring of online search behaviour allows the assessment of interest in a disease on a more representative level by including people outside of medical settings.ObjectivesThis study aims to reveal relevant aspects regarding psoriasis for people consulting the web.MethodsGoogle Ads Keyword Planner was used to select and evaluate relevant German keywords for ‘psoriasis' in the study period from September 2016 to August 2020 on German territory. All keywords were qualitatively reviewed and descriptively analysed in 11 categories. One‐way analysis of variance, Welch test and Pearson's correlation coefficient were used to assess differences within Germany, annual search volumes and seasons.ResultsOverall, 11,170,740 German language Google searches on the topic of psoriasis were made. In the general category, the most commonly searched disease aspect was related to the possible contagiousness of psoriasis (n = 122,040). ‘Head’ (51.1%), followed by ‘nails’ (15.9%), ‘hands’ (8.8%), ‘feet’ (7.8%), and ‘genital/anal area’ (6.1%), was the most frequently searched for localization. Pustular forms of psoriasis were overrepresented in the search volume. Almost half of the search terms for therapy were represented by keywords related to alternative care and phytotherapy, while less than 1% related to biologics. Differences between the federal states for search volume on treatment options (p &lt; 0.001) and between annual search volumes (p &lt; 0.001) were found.ConclusionsQueries on pustular forms and difficult‐to‐treat areas gained high online interest, which suggests higher disease impact or underdiagnosis. The high interest in alternative therapy options suggests possible therapy dissatisfaction, lack of patient information or lack of access to guideline‐conform therapy options. The possible infectiousness of psoriasis seems to remain an issue in the general population and should be addressed accordingly.

COVID-19 Propagation Prediction Model Using Improved Grey Wolf Optimization Algorithms in Combination with XGBoost and Bagging-Integrated Learning

Toward solving the slow convergence and low prediction accuracy problems associated with XGBoost in COVID-19-based transmission prediction, a novel algorithm based on guided aggregation is presented to optimize the XGBoost prediction model. In this study, we collect the early COVID-19 propagation data using web crawling techniques and use the Lasso algorithm to select the important attributes to simplify the attribute set. Moreover, to improve the global exploration and local mining capability of the grey wolf optimization (GWO) algorithm, a backward learning strategy has been introduced, and a chaotic search operator has been designed to improve GWO. In the end, the hyperparameters of XGBoost are continuously optimized using COLGWO in an iterative process, and Bagging is employed as a method of integrating the prediction effect of the COLGWO-XGBoost model optimization. The experiments, firstly, compared the search means and standard deviations of four search algorithms for eight standard test functions, and then, they compared and analyzed the prediction effects of fourteen models based on the COVID-19 web search data collected in China. Results show that the improved grey wolf algorithm has excellent performance benefits and that the combined model with integrated learning has good prediction ability. It demonstrates that the use of network search data in the early spread of COVID-19 can complement the historical information, and the combined model can be further extended to be applied to other prevention and control early warning tasks of public emergencies.

Application of Convolution Neural Networks in Web Search Log Mining for Effective Web Document Clustering

The volume of web search data stored in search engine log is increasing and has become big search log data. The web search log has been the source of data for mining based on web document clustering techniques to improve the efficiency and effectiveness of information retrieval. In this paper Deep Learning Model Convolution Neural Network(CNN) is used in big web search log data mining to learn the semantic representation of a document. These semantic documents vectors are clustered using K-means to group relevant documents for effective web document clustering. Experiment was done on the data set of web search query and associated clicked URLs to measure the quality of clusters based on document semantic representation using Deep learning model CNN. The clusters analysis was performed based on WCSS(the sum of squared distances of documents samples to their closest cluster center) and decrease in the WCSS in comparison to TF.IDF keyword based clusters confirm the effectiveness of CNN in web search log mining for effective web document clustering.

Understanding China\u2019s urban system evolution from web search index data

The spatial inequilibrium phenomenon is apparent during China’s rapid urbanization in the past four decades. As the fertility rate decreases and the population ages, this phenomenon will become more critical. To accurately forecast the future economic development of China, it is necessary to quantify the attractiveness of individual cities. This study introduces web search data to quantify the attractiveness of cities with a fine spatial scale (prefecture-level city) and relatively long-term span (nine years). Results confirm that the estimated city attractiveness can unravel a city’s capability to attract labor force, and suggest that tourism and health care functions of a city have a positive effect to the city’s attractiveness. Additionally, China’s north-south gap in economic development has been widened in the past decade, and 11 cities with nationwide influence have prosperous development potential. This study provides a new lens for predicting China’s economic development, as well as its spatial patterns.

The impact of anti-COVID-19 nonpharmaceutical interventions on hand, foot, and mouth disease-A spatiotemporal perspective in Xi'an, northwestern China.

Growing evidence has shown that anti‐COVID‐19 nonpharmaceutical interventions (NPIs) can support prevention and control of various infectious diseases, including intestinal diseases. However, most studies focused on the short‐term mitigating impact and neglected the dynamic impact over time. This study is aimed to investigate the dynamic impact of anti‐COVID‐19 NPIs on hand, foot, and mouth disease (HFMD) over time in Xi'an City, northwestern China. Based on the surveillance data of HFMD, meteorological and web search data, Bayesian Structural Time Series model and interrupted time series analysis were performed to quantitatively measure the impact of NPIs in sequent phases with different intensities and to predict the counterfactual number of HFMD cases. From 2013 to 2021, a total number of 172,898 HFMD cases were reported in Xi'an. In 2020, there appeared a significant decrease in HFMD incidence (−94.52%, 95% CI: −97.54% to −81.95%) in the first half of the year and the peak period shifted from June to October by a small margin of 6.74% compared to the previous years of 2013 to 2019. In 2021, the seasonality of HFMD incidence gradually returned to the bimodal temporal variation pattern with a significant average decline of 61.09%. In particular, the impact of NPIs on HFMD was more evident among young children (0–3 years), and the HFMD incidence reported in industrial areas had an unexpected increase of 51.71% in 2020 autumn and winter. Results suggested that both direct and indirect NPIs should be implemented as effective public health measures to reduce infectious disease and improve surveillance strategies, and HFMD incidence in Xi'an experienced a significant rebound to the previous seasonality after a prominent decline influenced by the anti‐COVID‐19 NPIs.

Cross-platform- and subgroup-differences in the well-being effects of Twitter, Instagram, and Facebook in the United States

Spatial aggregates of survey and web search data make it possible to identify the heterogeneous well-being effects of social media platforms. This study reports evidence from different sources of longitudinal data that suggests that the well-being effects of social media differ across platforms and population groups. The well-being effects of frequent social media visits are consistently positive for Facebook but negative for Instagram. Group-level analyses suggest that the positive well-being effects are experienced mainly by white, high-income populations at both the individual and the county level, while the adverse effects of Instagram use are observed on younger and Black populations. The findings are corroborated when geocoded web search data from Google is used and when self-reports from surveys are used in place of region-level aggregates. Greater Instagram use in regions is also linked to higher depression diagnoses across most sociodemographic groups.

Short-Term Forecasting of Air Travellers Outflows from Bali Using Web Search Data

Air travelers have become one of the strategic indicators in the transportation sector. The official data-released by Statistics Indonesia (BPS) for thirty days-lag, makes the condition of this indicator can’t be known in real-time. By the utilization of web search data that has been briskly evolving in recent years, this study aims to explore the possibility of using web search data in performing short-term forecasting to know the general outlook of the indicator earlier. Based on this study, web search data and official statistics figures show a strong correlation and having similar movement patterns over time. The application of web search data as a predictor in time series modeling, especially on time series regression and autoregressive model (SARIMA and SARIMAX), turn out a predicted value that well-approach the actual value of the response variable. In addition, it is proven that the use of web search data can increase model accuracy. The analysis results using SARIMAX model shows that the number of air traveller’s outflows from Bali in September and October 2021 will generally be higher than the number in August 2021. The increasing number of air travelers is thought due to a decrease in Covid-19 cases which has triggered the public's confidence in travelling about to rise again.

A Study on the Prediction of House Price Index in First‐Tier Cities in China Based on Heterogeneous Integrated Learning Model

To address the difficulty of low prediction accuracy, insufficient model stability, and certain lag associated with a single machine learning model in the prediction of house price, this paper proposes a multimodel fusion house price prediction model based on stacking integrated learning. Firstly, web search data affecting house prices were collected by web crawler technology, and Spearman correlation analysis was performed on the attribute set to reduce its complexity and establish a prediction index system for four first‐tier cities in China. Secondly, with the goal of improving accuracy, diversity, and generalization ability, the types of base learners as well as metalearners are determined, and the parameters of the base learners are optimized using the grey wolf optimization algorithm to produce the GWO‐stacking model, and the experimental results from four datasets demonstrate that the model has high prediction accuracy. Finally, to solve the issue of unintelligible black boxes in machine learning models, we have used the state‐of‐the‐art interpretation method SHAP combined with the LightGBM algorithm to interpret the model, and the result can be used as a basis for real estate policy planning and adjustment and even guide the demand of home buyers, thus improving the efficiency and effectiveness of government policy making.

Can crowdsourced data help to optimize atopic dermatitis treatment? Comparing web search data and environmental data in Germany.

Atopic dermatitis (AD) is one of the most common chronic skin diseases worldwide, showing various manifestations and having a severe impact on quality of life. As previous studies demonstrated, internet search analysis can help identify public interest in diseases and possible influencing factors on search behavior. To identify AD-related topics of interest in Germany using internet search volume. Google Ads Keyword Planner was used to identify AD-related search terms including their search volume in Germany on a national level as well as in 16 selected cities from January 2016 to December 2019. Identified keywords were qualitatively analyzed, and temporal trends as well as the influence of seasonal and environmental factors on search volume were assessed. Overall, 1222 AD-related search terms with a search volume of 8 842 360 searches were identified. An increase from 2016 to 2019 and seasonal peaks from January to April of each year were observed. Nationwide, the search volume correlated with mean monthly temperature and sun duration. With increasing temperature and sun duration, a significant decrease in search queries was observed. The most populated cities showed the lowest number of searches per 100 000 inhabitants (Berlin, Hamburg, and Munich). In the eight categories formed (comorbidities, general, influential factors, localization, stage of life, symptoms/severity, therapy/information, and questions on AD), the highest proportion of search queries were assigned to the category "therapy/information" in most cities. In this category, a focus on the topics of "alternative medicine" and "home remedies" could be observed. The overall high and increasing search volume indicates a high interest in AD-related topics, especially regarding treatment and disease education. Information provided by internet search volume analyses can optimize AD therapy and patient-centered care by providing insight into patient needs and predicting potential climatic trigger factors.

Laypeople's activity for seeking telephone number of EMS before and during the COVID-19 outbreak: An analysis of web search data.

ObjectiveTo explore trends and patterns of laypeople's activity for seeking telephone number of emergency medical services (EMS) based on analysis of online search traffic, including changes of the search activity with onset of the coronavirus disease 2019 (COVID-19) outbreak, in five countries – the United States of America (USA), India, Brazil, the United Kingdom (UK) and Russia.MethodsGoogle Trends (GT) country-level data on weekly relative search volumes (RSV) for top queries to seek EMS number were examined for January 2018–October 2021, including a comparison of RSVs between pre-COVID-19 period (January 2018–October 2019) and COVID-19 period (January 2020–October 2021), and evaluation of temporal associations of RSVs with weekly numbers of new COVID-19 cases.ResultsThe countries demonstrated diverse patterns of the search activity with significantly different mean RSVs (the USA 1.76, India 10.20, Brazil 2.51, the UK 6.42, Russia 56.79; p < 0.001). For all countries excepting the USA mean RSVs of the COVID-19 period were significantly higher compared with the pre-COVID-19 ones (India +74%, Brazil +148%, the UK +22%, Russia +9%; p ≤ 0.034), and exhibited positive correlations with numbers of new COVID-19 cases, more pronounced for 2021 (India rS = 0.538, Brazil 0.307, the UK 0.434, Russia 0.639; p ≤ 0.045).ConclusionLaypeople's activity for seeking EMS telephone number greatly varies between countries. It clearly responds to the spread of COVID-19 and could be reflective of public need for obtaining emergency help. Further studies are required to establish the role of GT for conducting real-time surveillance of population demand for EMS.

Early warning of COVID-19 hotspots using human mobility and web search query data

COVID-19 has disrupted the global economy and well-being of people at an unprecedented scale and magnitude. To contain the disease, an effective early warning system that predicts the locations of outbreaks is of crucial importance. Studies have shown the effectiveness of using large-scale mobility data to monitor the impacts of non-pharmaceutical interventions (e.g., lockdowns) through population density analysis. However, predicting the locations of potential outbreak occurrence is difficult using mobility data alone. Meanwhile, web search queries have been shown to be good predictors of the disease spread. In this study, we utilize a unique dataset of human mobility trajectories (GPS traces) and web search queries with common user identifiers (> 450 K users), to predict COVID-19 hotspot locations beforehand. More specifically, web search query analysis is conducted to identify users with high risk of COVID-19 contraction, and social contact analysis was further performed on the mobility patterns of these users to quantify the risk of an outbreak. Our approach is empirically tested using data collected from users in Tokyo, Japan. We show that by integrating COVID-19 related web search query analytics with social contact networks, we are able to predict COVID-19 hotspot locations 1–2 weeks beforehand, compared to just using social contact indexes or web search data analysis. This study proposes a novel method that can be used in early warning systems for disease outbreak hotspots, which can assist government agencies to prepare effective strategies to prevent further disease spread.Human mobility data and web search query data linked with common IDs are used to predict COVID-19 outbreaks. High risk social contact index captures both the contact density and COVID-19 contraction risks of individuals. Real world data was collected from 200 K individual users in Tokyo during the COVID-19 pandemic. Experiments showed that the index can be used for microscopic outbreak early warning.

A babel of web-searches: Googling unemployment during the pandemic

Researchers are increasingly exploiting web-searches to study phenomena for which timely and high-frequency data are not readily available. We propose a data-driven procedure which, exploiting machine learning techniques, solves the issue of identifying the list of queries linked to the phenomenon of interest, even in a cross-country setting. Queries are then aggregated in an indicator which can be used for causal inference. We apply this procedure to construct a search-based unemployment index and study the effect of lock-downs during the first wave of the covid-19 pandemic. In a Difference-in-Differences analysis, we show that the indicator rose significantly and persistently in the aftermath of lock-downs. This is not the case when using unprocessed (raw) web search data, which might return a partial figure of the labour market dynamics following lock-downs.

Web search data as health data? : Geographic differences, temporal trends, and topics of interest from internet search engine analyses in Germany

HintergrundDie eingeschränkt zeitnahe, kostengünstige und regionale Verfügbarkeit von Daten im Gesundheitswesen gilt als einer der limitierenden Faktoren für zeitgerechte Analysen im Bereich der Versorgungsforschung und damit für die Gesundheitsversorgung der allgemeinen Bevölkerung. Deshalb geraten zunehmend Internetsuchmaschinenanalysen in den Fokus.FragestellungWelchen Beitrag können Daten zum Internetsuchvolumen zu verschiedenen Erkrankungen in der Gesundheitsversorgung der deutschen Bevölkerung leisten? Wo liegt das Potenzial, und wo gibt es Grenzen?Material und MethodenEs erfolgten die Aufbereitung der aktuellen Literatur sowie eine selektive Übersicht der in den letzten 4 Jahren veröffentlichten Suchmaschinenanalysen zu dermatologischen, allergologischen und infektiologischen Erkrankungen in Deutschland.ErgebnisseDurch die Analyse von Suchmaschinendaten konnten zeitliche Entwicklungen wie Saisonalität, Monate mit dem höchsten Suchinteresse und geografische Unterschiede auf nationaler, Bundesländer- und Städteebene abgebildet und Interessenschwerpunkte bezüglich eines die Krankheit betreffenden Themas (z. B. Hautkrebs am Auge oder analer Juckreiz) aufgezeigt werden. Darüber hinaus fanden manche Studien einen Zusammenhang zwischen dem Suchvolumen und externen Faktoren (z. B. Temperatur, ärztliche Versorgungsstruktur) sowie zu registrierten Fällen (z. B. Hautkrebs, Borreliose).SchlussfolgerungInternetsuchmaschinendaten liefern als nahezu in Echtzeit verfügbare Datenquelle unter Berücksichtigung der aufgezeigten Fallstricke ein räumlich-zeitliches Abbild hinsichtlich der Bedürfnisse der internetnutzenden Bevölkerung. Sie können besonders nützlich in Situationen sein, in denen traditionelle Gesundheitsdaten begrenzt oder nicht vorhanden sind.Zusatzmaterial onlineDie Online-Version dieses Beitrags (10.1007/s00105-021-04918-x) enthält eine zusätzliche Tabelle. Beitrag und Zusatzmaterial stehen Ihnen im elektronischen Volltextarchiv auf https://www.springermedizin.de/der-hautarzt zur Verfügung. Sie finden das Zusatzmaterial am Beitragsende unter „Supplementary Information“.

Spatial-temporal disparities in the impact of the coronavirus disease 2019 on domestic tourism demand: A study of Emeishan National Park in Mainland China

The ongoing coronavirus disease 2019 pandemic, like other disasters or crises, can immensely influence visitors’ demand to visit affected destinations. The current study helps us better understand how this health crisis could affect the demand change from a micro-level perspective of small-scale tourist destinations. Based on the web search data from the Baidu Index, the present study adopts the Emeishan National Park in China as the study area and employs multiple methods to assess the spatial-temporal disparities in the impact of the coronavirus disease 2019 on domestic tourism demand. The main findings reveal that the demand changes during the observation period resemble a U-shaped curve along with the outbreak, spread, and control of the crisis, and such impacts exhibit different characteristics in the pre-event, prodromal, emergency, intermediate, and long-term recovery stages. During and after the pandemic, the short-distance market is the most vulnerable, but it presents the strongest resilience, while the medium- and long-distance markets are relatively less affected. Significant stratified heterogeneity in the tourism demand of domestic source markets also emerges before and after the crisis. Finally, some implications of promoting domestic tourism recovery in the post-pandemic era are discussed, and recommendations are made.

Characterizing internet search patterns for neurologic and neurosurgical conditions following celebrity diagnosis.

Social media and internet platforms have become significant drivers of mass-information. Highly publicized events, such as John McCain's announcement of his glioblastoma diagnosis, often drive national public interest in medical topics. Improved understanding of the temporality of interest spikes as well as the nature of the information that garners attention from outside the medical community can help inform ways in which the medical community can boost awareness of (and interest in) the field of neurosurgery. We utilized the "explore topics" feature on Google Trends to obtain web, news, and YouTube search data from May 1, 2015, to May 1, 2019 for the terms "glioblastoma," "brain tumor," "stroke," and "multiple sclerosis" to identify periods of visibly increased search interest. Search results for "glioblastoma" showed significantly elevated average interest during the period of July 3-23, 2017, as compared to that generated since this specific time period (42.6 vs. 8.73, P<0.001). This increased search activity therefore directly correlated with John McCain's public announcement of his glioblastoma diagnosis, and a similar search interest spike was evident using the search term "brain tumor" (87.3 vs. 64.2, P<0.001). Search results for "multiple sclerosis" showed - as a result of the online buzz created by Selma Blair's battle with the disease - significantly elevated average interest from October 8, 2018, to October 28, 2018, and February 11, 2019, to March 3, 2019, when compared to the average interest of the remaining time (59 vs. 40.16, P<0.001 and 69 vs. 40.16, P<0.001). Finally, there were no corresponding elevations in YouTube search interest for any of the terms associated with increased interest on Google Trends. Following major events related to the neurological disease of public figures there is an expected rise in Google search interest relevant to these topics. Our findings suggest that there is an optimal window of approximately 2 weeks following each of these events for activist and clinical groups to publicize their desired message, and for the field of neurosurgery and neurological science to increase public awareness regarding specific diseases, with a regression to baseline interest by 4 months following the event.

A machine-learning Strategy for Generating Pollen Estimates and Forecasts in North America

BACKGROUND AND AIM: Pollen information is important for risk communication, diagnosis of allergic disease, medical management, and surveillance. Pollen observations are sparse, however, and existing models have relatively low skill in estimating pollen concentrations for certain taxa and locations. The aim of this work is to leverage recently developed supervised Random Forest machine learning models to estimate daily pollen concentrations for multiple allergenic pollen taxa in the continental US. METHODS: We built upon a previously developed statistical supervised Random Forest machine learning model with demonstrated skill in estimating and forecasting four pollen types (Quercus, Cupressaceae, Ambrosia, and Poaceae) to make retrospective forecasts of seven other allergenic pollen taxa for the contiguous US. Meteorological, vegetation, and web search information were input to the models at city and regional scales and geographic coverage was expanded using data augmentation techniques. The models were further developed to estimate pollen concentrations in locations where there are no observations. Daily pollen concentration estimates were made for each of the 11 taxa. RESULTS:Daily estimates of pollen concentrations for 11 allergenic taxa for the period 2000 to 2020 were developed. Model forecast skill for the seven newly modeled taxa was assessed and compared with skill for the four previously modeled pollen types and available diagnostics for other pollen forecasting models. CONCLUSIONS:Weather, vegetation, and web search data can be used to estimate airborne pollen concentrations for prevalent allergenic taxa in the contiguous US. Reanalysis data products may be of use in diagnostic and management activities for people with allergic disease and for epidemiological analyses. Forecast models can be used in risk communication to facilitate exposure avoidance. Model forecasting skill is limited by availability of pollen observations and could be augmented by development of additional observation sites. KEYWORDS: pollen, random forest machine learning, allergic rhinitis, allergic asthma, forecasting, climate

Group-level human values estimated with web search data and archival data explain the geographic variation in COVID-19 severity in the United States

Objectives Personal Focus values (PF = Openness to change values + Self-enhancement values) motivate self-centred behaviours, and Social Focus values (SF = Conservation values + Self-transcendence values) promote self-sacrificial counterparts. This research investigated how a state-level PF-SF value-continuum would explain the geographic variation in COVID-19 severity in the United States. Design This research estimated state-level values by Google search data (from 2004 to 2019) on value-related words (e.g. family for conservation values) (Study 1a) and archival indicators (e.g. gun ownership rate for security values) (Study 1b). COVID-19 severity was measured by shorter time delay of first documented cases, shorter overall doubling times, higher reproductive ratio and higher case fatality ratio. Hierarchical and multilevel analyses examined how state-level values would predict COVID-19 severity across U.S. states (Studies 1a and 1b) and 3,135 counties (Study 2). Results State-level analyses accounting for spatial autocorrelation and covariates (e.g. COVID-19 testing rate, airport traffic, personality, etc.) revealed that the PF-SF value-continuums measured with different methods positively and significantly predicted COVID-19 severity, and the effects of state-level values on county-level COVID-19 severity were significant when county- and state-level covariates were controlled. Conclusion Social focus values may mitigate the devastating effect of COVID-19 in the United States.

Perceived Utility and Characterization of Personal Google Search Histories to Detect Data Patterns Proximal to a Suicide Attempt in Individuals Who Previously Attempted Suicide: Pilot Cohort Study.

BackgroundDespite decades of research to better understand suicide risk and to develop detection and prevention methods, suicide is still one of the leading causes of death globally. While large-scale studies using real-world evidence from electronic health records can identify who is at risk, they have not been successful at pinpointing when someone is at risk. Personalized social media and online search history data, by contrast, could provide an ongoing real-world datastream revealing internal thoughts and personal states of mind.ObjectiveWe conducted this study to determine the feasibility and acceptability of using personalized online information-seeking behavior in the identification of risk for suicide attempts.MethodsThis was a cohort survey study to assess attitudes of participants with a prior suicide attempt about using web search data for suicide prevention purposes, dates of lifetime suicide attempts, and an optional one-time download of their past web searches on Google. The study was conducted at the University of Washington School of Medicine Psychiatry Research Offices. The main outcomes were participants’ opinions on internet search data for suicide prediction and intervention and any potential change in online information-seeking behavior proximal to a suicide attempt. Individualized nonparametric association analysis was used to assess the magnitude of difference in web search data features derived from time periods proximal (7, 15, 30, and 60 days) to the suicide attempts versus the typical (baseline) search behavior of participants.ResultsA total of 62 participants who had attempted suicide in the past agreed to participate in the study. Internet search activity varied from person to person (median 2-24 searches per day). Changes in online search behavior proximal to suicide attempts were evident up to 60 days before attempt. For a subset of attempts (7/30, 23%) search features showed associations from 2 months to a week before the attempt. The top 3 search constructs associated with attempts were online searching patterns (9/30 attempts, 30%), semantic relatedness of search queries to suicide methods (7/30 attempts, 23%), and anger (7/30 attempts, 23%). Participants (40/59, 68%) indicated that use of this personalized web search data for prevention purposes was acceptable with noninvasive potential interventions such as connection to a real person (eg, friend, family member, or counselor); however, concerns were raised about detection accuracy, privacy, and the potential for overly invasive intervention.ConclusionsChanges in online search behavior may be a useful and acceptable means of detecting suicide risk. Personalized analysis of online information-seeking behavior showed notable changes in search behavior and search terms that are tied to early warning signs of suicide and are evident 2 months to 7 days before a suicide attempt.