The effects of synoptic weather on influenza infection incidences: a retrospective study utilizing digital disease surveillance.

The environmental drivers and mechanisms of influenza dynamics remain unclear. The recent development of influenza surveillance--particularly the emergence of digital epidemiology--provides an opportunity to further understand this puzzle as an area within applied human biometeorology. This paper investigates the short-term weather effects on human influenza activity at a synoptic scale during cold seasons. Using 10 years (2005-2014) of municipal level influenza surveillance data (an adjustment of the Google Flu Trends estimation from the Centers for Disease Control's virologic surveillance data) and daily spatial synoptic classification weather types, we explore and compare the effects of weather exposure on the influenza infection incidences in 79 cities across the USA. We find that during the cold seasons the presence of the polar [i.e., dry polar (DP) and moist polar (MP)] weather types is significantly associated with increasing influenza likelihood in 62 and 68% of the studied cities, respectively, while the presence of tropical [i.e., dry tropical (DT) and moist tropical (MT)] weather types is associated with a significantly decreasing occurrence of influenza in 56 and 43% of the cities, respectively. The MP and the DP weather types exhibit similar close positive correlations with influenza infection incidences, indicating that both cold-dry and cold-moist air provide favorable conditions for the occurrence of influenza in the cold seasons. Additionally, when tropical weather types are present, the humid (MT) and the dry (DT) weather types have similar strong impacts to inhibit the occurrence of influenza. These findings suggest that temperature is a more dominating atmospheric factor than moisture that impacts the occurrences of influenza in cold seasons.

Evaluating Google Flu Trends in Latin America: Important Lessons for the Next Phase of Digital Disease Detection.

BACKGROUND: Latin America has a substantial burden of influenza and rising Internet access and could benefit from real-time influenza epidemic prediction web tools such as Google Flu Trends (GFT) to assist in risk communication and resource allocation during epidemics. However, there has never been a published assessment of GFT's accuracy in most Latin American countries or in any low- to middle-income country. Our aim was to evaluate GFT in Argentina, Bolivia, Brazil, Chile, Mexico, Paraguay, Peru, and Uruguay. METHODS: Weekly influenza-test positive proportions for the eight countries were obtained from FluNet for the period January 2011-December 2014. Concurrent weekly Google-predicted influenza activity in the same countries was abstracted from GFT. Pearson correlation coefficients between observed and Google-predicted influenza activity trends were determined for each country. Permutation tests were used to examine background seasonal correlation between FluNet and GFT by country. RESULTS: There were frequent GFT prediction errors, with correlation ranging from r = -0.53 to 0.91. GFT-predicted influenza activity best correlated with FluNet data in Mexico follow by Uruguay, Argentina, Chile, Brazil, Peru, Bolivia and Paraguay. Correlation was generally highest in the more temperate countries with more regular influenza seasonality and lowest in tropical regions. A substantial amount of autocorrelation was noted, suggestive that GFT is not fully specific for influenza virus activity. CONCLUSIONS: We note substantial inaccuracies with GFT-predicted influenza activity compared with FluNet throughout Latin America, particularly among tropical countries with irregular influenza seasonality. Our findings offer valuable lessons for future Internet-based biosurveillance tools.

Google Flu Trends Spatial Variability Validated Against Emergency Department Influenza-Related Visits.

BACKGROUND: Influenza is a deadly and costly public health problem. Variations in its seasonal patterns cause dangerous surges in emergency department (ED) patient volume. Google Flu Trends (GFT) can provide faster influenza surveillance information than traditional CDC methods, potentially leading to improved public health preparedness. GFT has been found to correlate well with reported influenza and to improve influenza prediction models. However, previous validation studies have focused on isolated clinical locations. OBJECTIVE: The purpose of the study was to measure GFT surveillance effectiveness by correlating GFT with influenza-related ED visits in 19 US cities across seven influenza seasons, and to explore which city characteristics lead to better or worse GFT effectiveness. METHODS: Using Healthcare Cost and Utilization Project data, we collected weekly counts of ED visits for all patients with diagnosis (International Statistical Classification of Diseases 9) codes for influenza-related visits from 2005-2011 in 19 different US cities. We measured the correlation between weekly volume of GFT searches and influenza-related ED visits (ie, GFT ED surveillance effectiveness) per city. We evaluated the relationship between 15 publically available city indicators (11 sociodemographic, two health care utilization, and two climate) and GFT surveillance effectiveness using univariate linear regression. RESULTS: Correlation between city-level GFT and influenza-related ED visits had a median of .84, ranging from .67 to .93 across 19 cities. Temporal variability was observed, with median correlation ranging from .78 in 2009 to .94 in 2005. City indicators significantly associated (P<.10) with improved GFT surveillance include higher proportion of female population, higher proportion with Medicare coverage, higher ED visits per capita, and lower socioeconomic status. CONCLUSIONS: GFT is strongly correlated with ED influenza-related visits at the city level, but unexplained variation over geographic location and time limits its utility as standalone surveillance. GFT is likely most useful as an early signal used in conjunction with other more comprehensive surveillance techniques. City indicators associated with improved GFT surveillance provide some insight into the variability of GFT effectiveness. For example, populations with lower socioeconomic status may have a greater tendency to initially turn to the Internet for health questions, thus leading to increased GFT effectiveness. GFT has the potential to provide valuable information to ED providers for patient care and to administrators for ED surge preparedness.

Inference for reaction networks using the linear noise approximation.

We consider inference for the reaction rates in discretely observed networks such as those found in models for systems biology, population ecology, and epidemics. Most such networks are neither slow enough nor small enough for inference via the true state-dependent Markov jump process to be feasible. Typically, inference is conducted by approximating the dynamics through an ordinary differential equation (ODE) or a stochastic differential equation (SDE). The former ignores the stochasticity in the true model and can lead to inaccurate inferences. The latter is more accurate but is harder to implement as the transition density of the SDE model is generally unknown. The linear noise approximation (LNA) arises from a first-order Taylor expansion of the approximating SDE about a deterministic solution and can be viewed as a compromise between the ODE and SDE models. It is a stochastic model, but discrete time transition probabilities for the LNA are available through the solution of a series of ordinary differential equations. We describe how a restarting LNA can be efficiently used to perform inference for a general class of reaction networks; evaluate the accuracy of such an approach; and show how and when this approach is either statistically or computationally more efficient than ODE or SDE methods. We apply the LNA to analyze Google Flu Trends data from the North and South Islands of New Zealand, and are able to obtain more accurate short-term forecasts of new flu cases than another recently proposed method, although at a greater computational cost.

Cumulative query method for influenza surveillance using search engine data.

BACKGROUND: Internet search queries have become an important data source in syndromic surveillance system. However, there is currently no syndromic surveillance system using Internet search query data in South Korea. OBJECTIVES: The objective of this study was to examine correlations between our cumulative query method and national influenza surveillance data. METHODS: Our study was based on the local search engine, Daum (approximately 25% market share), and influenza-like illness (ILI) data from the Korea Centers for Disease Control and Prevention. A quota sampling survey was conducted with 200 participants to obtain popular queries. We divided the study period into two sets: Set 1 (the 2009/10 epidemiological year for development set 1 and 2010/11 for validation set 1) and Set 2 (2010/11 for development Set 2 and 2011/12 for validation Set 2). Pearson's correlation coefficients were calculated between the Daum data and the ILI data for the development set. We selected the combined queries for which the correlation coefficients were .7 or higher and listed them in descending order. Then, we created a cumulative query method n representing the number of cumulative combined queries in descending order of the correlation coefficient. RESULTS: In validation set 1, 13 cumulative query methods were applied, and 8 had higher correlation coefficients (min=.916, max=.943) than that of the highest single combined query. Further, 11 of 13 cumulative query methods had an r value of ≥.7, but 4 of 13 combined queries had an r value of ≥.7. In validation set 2, 8 of 15 cumulative query methods showed higher correlation coefficients (min=.975, max=.987) than that of the highest single combined query. All 15 cumulative query methods had an r value of ≥.7, but 6 of 15 combined queries had an r value of ≥.7. CONCLUSIONS: Cumulative query method showed relatively higher correlation with national influenza surveillance data than combined queries in the development and validation set.

A case study of the New York City 2012-2013 influenza season with daily geocoded Twitter data from temporal and spatiotemporal perspectives.

BACKGROUND: Twitter has shown some usefulness in predicting influenza cases on a weekly basis in multiple countries and on different geographic scales. Recently, Broniatowski and colleagues suggested Twitter's relevance at the city-level for New York City. Here, we look to dive deeper into the case of New York City by analyzing daily Twitter data from temporal and spatiotemporal perspectives. Also, through manual coding of all tweets, we look to gain qualitative insights that can help direct future automated searches. OBJECTIVE: The intent of the study was first to validate the temporal predictive strength of daily Twitter data for influenza-like illness emergency department (ILI-ED) visits during the New York City 2012-2013 influenza season against other available and established datasets (Google search query, or GSQ), and second, to examine the spatial distribution and the spread of geocoded tweets as proxies for potential cases. METHODS: From the Twitter Streaming API, 2972 tweets were collected in the New York City region matching the keywords "flu", "influenza", "gripe", and "high fever". The tweets were categorized according to the scheme developed by Lamb et al. A new fourth category was added as an evaluator guess for the probability of the subject(s) being sick to account for strength of confidence in the validity of the statement. Temporal correlations were made for tweets against daily ILI-ED visits and daily GSQ volume. The best models were used for linear regression for forecasting ILI visits. A weighted, retrospective Poisson model with SaTScan software (n=1484), and vector map were used for spatiotemporal analysis. RESULTS: Infection-related tweets (R=.763) correlated better than GSQ time series (R=.683) for the same keywords and had a lower mean average percent error (8.4 vs 11.8) for ILI-ED visit prediction in January, the most volatile month of flu. SaTScan identified primary outbreak cluster of high-probability infection tweets with a 2.74 relative risk ratio compared to medium-probability infection tweets at P=.001 in Northern Brooklyn, in a radius that includes Barclay's Center and the Atlantic Avenue Terminal. CONCLUSIONS: While others have looked at weekly regional tweets, this study is the first to stress test Twitter for daily city-level data for New York City. Extraction of personal testimonies of infection-related tweets suggests Twitter's strength both qualitatively and quantitatively for ILI-ED prediction compared to alternative daily datasets mixed with awareness-based data such as GSQ. Additionally, granular Twitter data provide important spatiotemporal insights. A tweet vector-map may be useful for visualization of city-level spread when local gold standard data are otherwise unavailable.

Performance of eHealth data sources in local influenza surveillance: a 5-year open cohort study.

BACKGROUND: There is abundant global interest in using syndromic data from population-wide health information systems--referred to as eHealth resources--to improve infectious disease surveillance. Recently, the necessity for these systems to achieve two potentially conflicting requirements has been emphasized. First, they must be evidence-based; second, they must be adjusted for the diversity of populations, lifestyles, and environments. OBJECTIVE: The primary objective was to examine correlations between data from Google Flu Trends (GFT), computer-supported telenursing centers, health service websites, and influenza case rates during seasonal and pandemic influenza outbreaks. The secondary objective was to investigate associations between eHealth data, media coverage, and the interaction between circulating influenza strain(s) and the age-related population immunity. METHODS: An open cohort design was used for a five-year study in a Swedish county (population 427,000). Syndromic eHealth data were collected from GFT, telenursing call centers, and local health service website visits at page level. Data on mass media coverage of influenza was collected from the major regional newspaper. The performance of eHealth data in surveillance was measured by correlation effect size and time lag to clinically diagnosed influenza cases. RESULTS: Local media coverage data and influenza case rates showed correlations with large effect sizes only for the influenza A (A) pH1N1 outbreak in 2009 (r=.74, 95% CI .42-.90; P<.001) and the severe seasonal A H3N2 outbreak in 2011-2012 (r=.79, 95% CI .42-.93; P=.001), with media coverage preceding case rates with one week. Correlations between GFT and influenza case data showed large effect sizes for all outbreaks, the largest being the seasonal A H3N2 outbreak in 2008-2009 (r=.96, 95% CI .88-.99; P<.001). The preceding time lag decreased from two weeks during the first outbreaks to one week from the 2009 A pH1N1 pandemic. Telenursing data and influenza case data showed correlations with large effect sizes for all outbreaks after the seasonal B and A H1 outbreak in 2007-2008, with a time lag decreasing from two weeks for the seasonal A H3N2 outbreak in 2008-2009 (r=.95, 95% CI .82-.98; P<.001) to none for the A p H1N1 outbreak in 2009 (r=.84, 95% CI .62-.94; P<.001). Large effect sizes were also observed between website visits and influenza case data. CONCLUSIONS: Correlations between the eHealth data and influenza case rates in a Swedish county showed large effect sizes throughout a five-year period, while the time lag between signals in eHealth data and influenza rates changed. Further research is needed on analytic methods for adjusting eHealth surveillance systems to shifts in media coverage and to variations in age-group related immunity between virus strains. The results can be used to inform the development of alert-generating eHealth surveillance systems that can be subject for prospective evaluations in routine public health practice.

Psychological AI: Designing Algorithms Informed by Human Psychology.

Psychological artificial intelligence (AI) applies insights from psychology to design computer algorithms. Its core domain is decision-making under uncertainty, that is, ill-defined situations that can change in unexpected ways rather than well-defined, stable problems, such as chess and Go. Psychological theories about heuristic processes under uncertainty can provide possible insights. I provide two illustrations. The first shows how recency-the human tendency to rely on the most recent information and ignore base rates-can be built into a simple algorithm that predicts the flu substantially better than did Google Flu Trends's big-data algorithms. The second uses a result from memory research-the paradoxical effect that making numbers less precise increases recall-in the design of algorithms that predict recidivism. These case studies provide an existence proof that psychological AI can help design efficient and transparent algorithms.

Improved Real-Time Influenza Surveillance: Using Internet Search Data in Eight Latin American Countries.

BACKGROUND: Novel influenza surveillance systems that leverage Internet-based real-time data sources including Internet search frequencies, social-network information, and crowd-sourced flu surveillance tools have shown improved accuracy over the past few years in data-rich countries like the United States. These systems not only track flu activity accurately, but they also report flu estimates a week or more ahead of the publication of reports produced by healthcare-based systems, such as those implemented and managed by the Centers for Disease Control and Prevention. Previous work has shown that the predictive capabilities of novel flu surveillance systems, like Google Flu Trends (GFT), in developing countries in Latin America have not yet delivered acceptable flu estimates. OBJECTIVE: The aim of this study was to show that recent methodological improvements on the use of Internet search engine information to track diseases can lead to improved retrospective flu estimates in multiple countries in Latin America. METHODS: A machine learning-based methodology that uses flu-related Internet search activity and historical information to monitor flu activity, named ARGO (AutoRegression with Google search), was extended to generate flu predictions for 8 Latin American countries (Argentina, Bolivia, Brazil, Chile, Mexico, Paraguay, Peru, and Uruguay) for the time period: January 2012 to December of 2016. These retrospective (out-of-sample) Influenza activity predictions were compared with historically observed flu suspected cases in each country, as reported by Flunet, an influenza surveillance database maintained by the World Health Organization. For a baseline comparison, retrospective (out-of-sample) flu estimates were produced for the same time period using autoregressive models that only leverage historical flu activity information. RESULTS: Our results show that ARGO-like models' predictive power outperform autoregressive models in 6 out of 8 countries in the 2012-2016 time period. Moreover, ARGO significantly improves on historical flu estimates produced by the now discontinued GFT for the time period of 2012-2015, where GFT information is publicly available. CONCLUSIONS: We demonstrate here that a self-correcting machine learning method, leveraging Internet-based disease-related search activity and historical flu trends, has the potential to produce reliable and timely flu estimates in multiple Latin American countries. This methodology may prove helpful to local public health officials who design and implement interventions aimed at mitigating the effects of influenza outbreaks. Our methodology generally outperforms both the now-discontinued tool GFT, and autoregressive methodologies that exploit only historical flu activity to produce future disease estimates.

Association of influenza-like illness and climatic parameters with hemoptysis.

BACKGROUND: Although some authors have suggested that there is some seasonal periodicity of hemoptysis, or relation to respiratory tract infections, the association of influenza or climatic parameters with hemoptysis has been poorly investigated. Our aim was to describe the relationship between influenza and climatic parameters with severe hemoptysis that required bronchial artery embolization (BAE). METHODS: All consecutive subjects with at least one episode of hemoptysis that required BAE during a 5-y period were included. We applied a general multivariable additive seemingly causal model corresponding to a lagged variable autoregressive model with the exogenous variables as monthly mean temperature, lagged monthly mean temperature (-1), and monthly mean influenza activity, and the number of embolizations as the endogenous variable. RESULTS: We found a significant association between severe hemoptysis requiring BAE and low monthly mean temperature and influenza activity. Other climatic factors, such as atmospheric pressure, rainfall, relative humidity, or wind speed, failed to show significant association with the occurrence of life-threatening hemoptysis. CONCLUSIONS: There is a strong long running relationship between severe hemoptysis and low monthly mean temperature. A weaker association of hemoptysis with influenza activity was also found.