Influenza Forecasting Research Articles

Forecasting influenza epidemics in Hong Kong using Google search queries data: A new integrated approach

Forecasting influenza epidemics has important practical implications. However, the performance of traditional methods adopting in Hong Kong influenza forecasting is limited due to its particularity. This paper proposes an integrated approach for Hong Kong influenza epidemics forecasting. The novelties of our approach mainly include: firstly, we adopt a model for Google search queries data collection and selection in Hong Kong to substitute Google Correlate. Secondly, we adopt the stacked autoencoder (SAE) to reduce the dimensionality of Google search queries data. Thirdly, we adopt a signal decomposition method named variational mode decomposition (VMD) to decompose the influenza data into modes with different frequencies, which can extract the characteristic. Fourthly, we use artificial neural networks (ANN) to forecast these modes of influenza epidemics extracted by VMD respectively, then these forecasts of each mode are added to generate the final forecasting results. From the perspective of forecasting accuracy and hypothesis tests, the empirical results show that our proposed integrated approach SAE-VMD-ANN significantly outperforms some other benchmark models both in the whole period and influenza season. The performance of our proposed model during the COVID-19 pandemic is checked too.

Predicting seasonal influenza using supermarket retail records.

Increased availability of epidemiological data, novel digital data streams, and the rise of powerful machine learning approaches have generated a surge of research activity on real-time epidemic forecast systems. In this paper, we propose the use of a novel data source, namely retail market data to improve seasonal influenza forecasting. Specifically, we consider supermarket retail data as a proxy signal for influenza, through the identification of sentinel baskets, i.e., products bought together by a population of selected customers. We develop a nowcasting and forecasting framework that provides estimates for influenza incidence in Italy up to 4 weeks ahead. We make use of the Support Vector Regression (SVR) model to produce the predictions of seasonal flu incidence. Our predictions outperform both a baseline autoregressive model and a second baseline based on product purchases. The results show quantitatively the value of incorporating retail market data in forecasting models, acting as a proxy that can be used for the real-time analysis of epidemics.

Enhancing Influenza Epidemics Forecasting Accuracy in China with Both Official and Unofficial Online News Articles, 2019–2020

Real-time online data sources have contributed to timely and accurate forecasting of influenza activities while also suffered from instability and linguistic noise. Few previous studies have focused on unofficial online news articles, which are abundant in their numbers, rich in information, and relatively low in noise. This study examined whether monitoring both official and unofficial online news articles can improve influenza activity forecasting accuracy during influenza outbreaks. Data were retrieved from a Chinese commercial online platform and the website of the Chinese National Influenza Center. We modeled weekly fractions of influenza-related online news articles and compared them against weekly influenza-like illness (ILI) rates using autoregression analyses. We retrieved 153,958,695 and 149,822,871 online news articles focusing on the south and north of mainland China separately from 6 October 2019 to 17 May 2020. Our model based on online news articles could significantly improve the forecasting accuracy, compared to other influenza surveillance models based on historical ILI rates (p = 0.002 in the south; p = 0.000 in the north) or adding microblog data as an exogenous input (p = 0.029 in the south; p = 0.000 in the north). Our finding also showed that influenza forecasting based on online news articles could be 1–2 weeks ahead of official ILI surveillance reports. The results revealed that monitoring online news articles could supplement traditional influenza surveillance systems, improve resource allocation, and offer models for surveillance of other emerging diseases.

Modeling influenza seasonality in the tropics and subtropics.

Climate drivers such as humidity and temperature may play a key role in influenza seasonal transmission dynamics. Such a relationship has been well defined for temperate regions. However, to date no models capable of capturing the diverse seasonal pattern in tropical and subtropical climates exist. In addition, multiple influenza viruses could cocirculate and shape epidemic dynamics. Here we construct seven mechanistic epidemic models to test the effect of two major climate drivers (humidity and temperature) and multi-strain co-circulation on influenza transmission in Hong Kong, an influenza epidemic center located in the subtropics. Based on model fit to long-term influenza surveillance data from 1998 to 2018, we found that a simple model incorporating the effect of both humidity and temperature best recreated the influenza epidemic patterns observed in Hong Kong. The model quantifies a bimodal effect of absolute humidity on influenza transmission where both low and very high humidity levels facilitate transmission quadratically; the model also quantifies the monotonic but nonlinear relationship with temperature. In addition, model results suggest that, at the population level, a shorter immunity period can approximate the co-circulation of influenza virus (sub)types. The basic reproductive number R0 estimated by the best-fit model is also consistent with laboratory influenza survival and transmission studies under various combinations of humidity and temperature levels. Overall, our study has developed a simple mechanistic model capable of quantifying the impact of climate drivers on influenza transmission in (sub)tropical regions. This model can be applied to improve influenza forecasting in the (sub)tropics in the future.

Steering a Historical Disease Forecasting Model Under a Pandemic: Case of Flu and COVID-19

Forecasting influenza in a timely manner aids health organizations and policymakers in adequate preparation and decision making. However, effective influenza forecasting still remains a challenge despite increasing research interest. It is even more challenging amidst the COVID pandemic, when the influenza-like illness (ILI) counts are affected by various factors such as symptomatic similarities with COVID-19 and shift in healthcare seeking patterns of the general population. Under the current pandemic, historical influenza models carry valuable expertise about the disease dynamics but face difficulties adapting. Therefore, we propose CALI-Net, a neural transfer learning architecture which allows us to 'steer' a historical disease forecasting model to new scenarios where flu and COVID co-exist. Our framework enables this adaptation by automatically learning when it should emphasize learning from COVID-related signals and when it should learn from the historical model. Thus, we exploit representations learned from historical ILI data as well as the limited COVID-related signals. Our experiments demonstrate that our approach is successful in adapting a historical forecasting model to the current pandemic. In addition, we show that success in our primary goal, adaptation, does not sacrifice overall performance as compared with state-of-the-art influenza forecasting approaches.

Improving influenza surveillance based on multi-granularity deep spatiotemporal neural network

Influenza is a common respiratory disease that can cause human illness and death. Timely and accurate prediction of disease risk is of great importance for public health management and prevention. The influenza data belong to typical spatiotemporal data in that influenza transmission is influenced by regional and temporal interactions. Many existing methods only use the historical time series information for prediction, which ignores the effect of spatial correlations of neighboring regions and temporal correlations of different time periods. Mining spatiotemporal information for risk prediction is a significant and challenging issue. In this paper, we propose a new end-to-end spatiotemporal deep neural network structure for influenza risk prediction. The proposed model mainly consists of two parts. The first stage is the spatiotemporal feature extraction stage where two-stream convolutional and recurrent neural networks are constructed to extract the different regions and time granularity information. Then, a dynamically parametric-based fusion method is adopted to integrate the two-stream features and making predictions. In our work, we demonstrate that our method, tested on two influenza-like illness (ILI) datasets (US-HHS and SZ-HIC), achieved the best performance across all evaluation metrics. The results imply that our method has outstanding performance for spatiotemporal feature extraction and enables accurate predictions compared to other well-known influenza forecasting models.

Public Health Risk Forecasting with Multiple Machine Learning Methods Combined:Case Study of Influenza Forecasting in Lanzhou, China

Public Health Risk Forecasting with Multiple Machine Learning Methods Combined:Case Study of Influenza Forecasting in Lanzhou, China

Machine Learning-Based Two-Stage Data Selection Scheme for Long-Term Influenza Forecasting

Machine Learning-Based Two-Stage Data Selection Scheme for Long-Term Influenza Forecasting

Influenza Virus: Tracking, Predicting, and Forecasting.

Influenza is a common respiratory infection that causes considerable morbidity and mortality worldwide each year. In recent years, along with the improvement in computational resources, there have been a number of important developments in the science of influenza surveillance and forecasting. Influenza surveillance systems have been improved by synthesizing multiple sources of information. Influenza forecasting has developed into an active field, with annual challenges in the United States that have stimulated improved methodologies. Work continues on the optimal approaches to assimilating surveillance data and information on relevant driving factors to improve estimates of the current situation (nowcasting) and to forecast future dynamics.

Forecasting influenza in Europe using a metapopulation model incorporating cross-border commuting and air travel.

Past work has shown that models incorporating human travel can improve the quality of influenza forecasts. Here, we develop and validate a metapopulation model of twelve European countries, in which international translocation of virus is driven by observed commuting and air travel flows, and use this model to generate influenza forecasts in conjunction with incidence data from the World Health Organization. We find that, although the metapopulation model fits the data well, it offers no improvement over isolated models in forecast quality. We discuss several potential reasons for these results. In particular, we note the need for data that are more comparable from country to country, and offer suggestions as to how surveillance systems might be improved to achieve this goal.

Integrating genotypes and phenotypes improves long-term forecasts of seasonal influenza A/H3N2 evolution.

Seasonal influenza virus A/H3N2 is a major cause of death globally. Vaccination remains the most effective preventative. Rapid mutation of hemagglutinin allows viruses to escape adaptive immunity. This antigenic drift necessitates regular vaccine updates. Effective vaccine strains need to represent H3N2 populations circulating one year after strain selection. Experts select strains based on experimental measurements of antigenic drift and predictions made by models from hemagglutinin sequences. We developed a novel influenza forecasting framework that integrates phenotypic measures of antigenic drift and functional constraint with previously published sequence-only fitness estimates. Forecasts informed by phenotypic measures of antigenic drift consistently outperformed previous sequence-only estimates, while sequence-only estimates of functional constraint surpassed more comprehensive experimentally-informed estimates. Importantly, the best models integrated estimates of both functional constraint and either antigenic drift phenotypes or recent population growth.

Applying Machine Learning Models with An Ensemble Approach for Accurate Real-Time Influenza Forecasting in Taiwan: Development and Validation Study.

BackgroundChangeful seasonal influenza activity in subtropical areas such as Taiwan causes problems in epidemic preparedness. The Taiwan Centers for Disease Control has maintained real-time national influenza surveillance systems since 2004. Except for timely monitoring, epidemic forecasting using the national influenza surveillance data can provide pivotal information for public health response.ObjectiveWe aimed to develop predictive models using machine learning to provide real-time influenza-like illness forecasts.MethodsUsing surveillance data of influenza-like illness visits from emergency departments (from the Real-Time Outbreak and Disease Surveillance System), outpatient departments (from the National Health Insurance database), and the records of patients with severe influenza with complications (from the National Notifiable Disease Surveillance System), we developed 4 machine learning models (autoregressive integrated moving average, random forest, support vector regression, and extreme gradient boosting) to produce weekly influenza-like illness predictions for a given week and 3 subsequent weeks. We established a framework of the machine learning models and used an ensemble approach called stacking to integrate these predictions. We trained the models using historical data from 2008-2014. We evaluated their predictive ability during 2015-2017 for each of the 4-week time periods using Pearson correlation, mean absolute percentage error (MAPE), and hit rate of trend prediction. A dashboard website was built to visualize the forecasts, and the results of real-world implementation of this forecasting framework in 2018 were evaluated using the same metrics.ResultsAll models could accurately predict the timing and magnitudes of the seasonal peaks in the then-current week (nowcast) (ρ=0.802-0.965; MAPE: 5.2%-9.2%; hit rate: 0.577-0.756), 1-week (ρ=0.803-0.918; MAPE: 8.3%-11.8%; hit rate: 0.643-0.747), 2-week (ρ=0.783-0.867; MAPE: 10.1%-15.3%; hit rate: 0.669-0.734), and 3-week forecasts (ρ=0.676-0.801; MAPE: 12.0%-18.9%; hit rate: 0.643-0.786), especially the ensemble model. In real-world implementation in 2018, the forecasting performance was still accurate in nowcasts (ρ=0.875-0.969; MAPE: 5.3%-8.0%; hit rate: 0.582-0.782) and remained satisfactory in 3-week forecasts (ρ=0.721-0.908; MAPE: 7.6%-13.5%; hit rate: 0.596-0.904).ConclusionsThis machine learning and ensemble approach can make accurate, real-time influenza-like illness forecasts for a 4-week period, and thus, facilitate decision making.

Ambient temperature, seasonal influenza and risk of cardiovascular disease in a subtropical area in Southern China

BackgroundGiven the regular winter recurrence of influenza epidemics and the biologically plausible association between seasonal influenza and cardiovascular events, researchers assumed a valid and reliable influenza forecast could envision the timing and burden of winter surge in cardiovascular (CVD) hospitalizations. This, however, is well justified only in temperate regions. In this study, we aim to investigate the temporal association between ambient temperature, seasonal influenza and risk of cardiovascular events in a subtropical city. MethodsGeneralized additive model was used in conjunction with distributed-lag non-linear model of quasi-Poisson family to estimate the association of interest with daily CVD admissions as outcome and daily influenza admissions as predictor, while controlling for meteorological factors (i.e. temperature, relative humidity, wind speed and total rainfall) and respiratory pollutants (i.e. nitrogen dioxide, sulphur dioxide, ozone and PM10). Results were expressed in the form of relative risk (RR). ResultsUsing median as the reference value, a U-shaped association was observed between CVD admissions and temperature. A slight decrease in RR was detected mainly towards the lower end of the temperature scale after adjusting for influenza admissions. Risk of CVD admission was found to be positively associated with the number of influenza hospitalization cases; this association remained consistent and statistically significant across subgroups of age except for those aged 5–49 years. ConclusionThe slight reduction in CVD admission risk towards the lower end of the temperature scale after controlling for influenza activity might be attributed to the winter peaks of influenza, meaning that the effect of low temperature on CVD admissions might be partly mediated by influenza infection. In summary, this study reassures us that ambient temperature is independently associated with CVD hospital admissions and offers support for a positive association between seasonal influenza activity and cardiovascular events in Hong Kong.

Forecasting the 2017/2018 seasonal influenza epidemic in England using multiple dynamic transmission models: a case study

BackgroundSince the 2009 A/H1N1 pandemic, Public Health England have developed a suite of real-time statistical models utilising enhanced pandemic surveillance data to nowcast and forecast a future pandemic. Their ability to track seasonal influenza and predict heightened winter healthcare burden in the light of high activity in Australia in 2017 was untested.MethodsFour transmission models were used in forecasting the 2017/2018 seasonal influenza epidemic in England: a stratified primary care model using daily, region-specific, counts and virological swab positivity of influenza-like illness consultations in general practice (GP); a strain-specific (SS) model using weekly, national GP ILI and virological data; an intensive care model (ICU) using reports of ICU influenza admissions; and a synthesis model that included all data sources. For the first 12 weeks of 2018, each model was applied to the latest data to provide estimates of epidemic parameters and short-term influenza forecasts. The added value of pre-season population susceptibility data was explored.ResultsThe combined results provided valuable nowcasts of the state of the epidemic. Short-term predictions of burden on primary and secondary health services were initially highly variable before reaching consensus beyond the observed peaks in activity between weeks 3–4 of 2018. Estimates for R0 were consistent over time for three of the four models until week 12 of 2018, and there was consistency in the estimation of R0 across the SPC and SS models, and in the ICU attack rates estimated by the ICU and the synthesis model. Estimation and predictions varied according to the assumed levels of pre-season immunity.ConclusionsThis exercise successfully applied a range of pandemic models to seasonal influenza. Forecasting early in the season remains challenging but represents a crucially important activity to inform planning. Improved knowledge of pre-existing levels of immunity would be valuable.

Accuracy of real-time multi-model ensemble forecasts for seasonal influenza in the U.S.

Seasonal influenza results in substantial annual morbidity and mortality in the United States and worldwide. Accurate forecasts of key features of influenza epidemics, such as the timing and severity of the peak incidence in a given season, can inform public health response to outbreaks. As part of ongoing efforts to incorporate data and advanced analytical methods into public health decision-making, the United States Centers for Disease Control and Prevention (CDC) has organized seasonal influenza forecasting challenges since the 2013/2014 season. In the 2017/2018 season, 22 teams participated. A subset of four teams created a research consortium called the FluSight Network in early 2017. During the 2017/2018 season they worked together to produce a collaborative multi-model ensemble that combined 21 separate component models into a single model using a machine learning technique called stacking. This approach creates a weighted average of predictive densities where the weight for each component is determined by maximizing overall ensemble accuracy over past seasons. In the 2017/2018 influenza season, one of the largest seasonal outbreaks in the last 15 years, this multi-model ensemble performed better on average than all individual component models and placed second overall in the CDC challenge. It also outperformed the baseline multi-model ensemble created by the CDC that took a simple average of all models submitted to the forecasting challenge. This project shows that collaborative efforts between research teams to develop ensemble forecasting approaches can bring measurable improvements in forecast accuracy and important reductions in the variability of performance from year to year. Efforts such as this, that emphasize real-time testing and evaluation of forecasting models and facilitate the close collaboration between public health officials and modeling researchers, are essential to improving our understanding of how best to use forecasts to improve public health response to seasonal and emerging epidemic threats.

Google Trends and forecasting of influenza epidemics in Lithuania

Abstract Background Influenza epidemics are a major public health concern, causing hundreds of thousands of deaths worldwide each year. Studies suggest a possible beneficial use of online activity monitoring in influenza epidemics surveillance and forecasting. Until the present time Google search user activity has not been assessed for forecasting of influenza epidemics in Lithuania. Though this country has 2.8 million residents and mother language of majority of them is Lithuanian. Thus, we assessed a possible use of Google Trends for forecasting of influenza epidemics in Lithuania. Methods We translated Google queries most commonly analyzed for forecasting of influenza epidemics by other researchers to Lithuanian language and used Google Trends to retrieve data about frequency of their usage in Google Search during the period of last 5 years. We assessed cross-correlation between usage frequency of each query with an incidence rate of acute upper respiratory tract infections (AURTI) and influenza. Providing data available 2 weeks before the first announcement of influenza epidemic, for the query with the strongest correlation we constructed the autoregressive integrated moving average (ARIMA) model and used it to make a forecast for upcoming 9 weeks. Then we assessed the cross-correlation between the forecast and incidence rate of AURTI and influenza. Results We found the strongest correlation with an incidence rate of AURTI and influenza for a keyword ‘gripas’ (correlation coefficients were 0.805 and 0.803 for lags of 0 and 1, respectively, p = 0.001). 75% true values of the query usage frequency fell into 95% confidence interval of the forecast by ARIMA(1,0,0)(0,1,0)[52] model. Incidence rate of AURTI and influenza correlated with the forecast having coefficients of 0.902 and 0.733 for lags of 0 and 1, respectively (p = 0.007). Conclusions Query ‘gripas’ demonstrated the possible beneficial use in forecasting influenza epidemics 2 weeks before their onset. Key messages Google Trends has a potential to be used for forecasting influenza epidemics in Lithuania. Google Search query ‘gripas’ proved the most beneficial for forecasting incidence rate of upper respiratory tract infections and influenza in Lithuania.

Forecasting influenza activity using self-adaptive AI model and multi-source data in Chongqing, China.

BackgroundEarly detection of influenza activity followed by timely response is a critical component of preparedness for seasonal influenza epidemic and influenza pandemic. However, most relevant studies were conducted at the regional or national level with regular seasonal influenza trends. There are few feasible strategies to forecast influenza activity at the local level with irregular trends.MethodsMulti-source electronic data, including historical percentage of influenza-like illness (ILI%), weather data, Baidu search index and Sina Weibo data of Chongqing, China, were collected and integrated into an innovative Self-adaptive AI Model (SAAIM), which was constructed by integrating Seasonal Autoregressive Integrated Moving Average model and XGBoost model using a self-adaptive weight adjustment mechanism. SAAIM was applied to ILI% forecast in Chongqing from 2017 to 2018, of which the performance was compared with three previously available models on forecasting.FindingsILI% showed an irregular seasonal trend from 2012 to 2018 in Chongqing. Compared with three reference models, SAAIM achieved the best performance on forecasting ILI% of Chongqing with the mean absolute percentage error (MAPE) of 11·9%, 7·5%, and 11·9% during the periods of the year 2014–2016, 2017, and 2018 respectively. Among the three categories of source data, historical influenza activity contributed the most to the forecast accuracy by decreasing the MAPE by 19·6%, 43·1%, and 11·1%, followed by weather information (MAPE reduced by 3·3%, 17·1%, and 2·2%), and Internet-related public sentiment data (MAPE reduced by 1·1%, 0·9%, and 1·3%).InterpretationAccurate influenza forecast in areas with irregular seasonal influenza trends can be made by SAAIM with multi-source electronic data.

Reappraising the utility of Google Flu Trends.

Estimation of influenza-like illness (ILI) using search trends activity was intended to supplement traditional surveillance systems, and was a motivation behind the development of Google Flu Trends (GFT). However, several studies have previously reported large errors in GFT estimates of ILI in the US. Following recent release of time-stamped surveillance data, which better reflects real-time operational scenarios, we reanalyzed GFT errors. Using three data sources—GFT: an archive of weekly ILI estimates from Google Flu Trends; ILIf: fully-observed ILI rates from ILINet; and, ILIp: ILI rates available in real-time based on partial reporting—five influenza seasons were analyzed and mean square errors (MSE) of GFT and ILIp as estimates of ILIf were computed. To correct GFT errors, a random forest regression model was built with ILI and GFT rates from the previous three weeks as predictors. An overall reduction in error of 44% was observed and the errors of the corrected GFT are lower than those of ILIp. An 80% reduction in error during 2012/13, when GFT had large errors, shows that extreme failures of GFT could have been avoided. Using autoregressive integrated moving average (ARIMA) models, one- to four-week ahead forecasts were generated with two separate data streams: ILIp alone, and with both ILIp and corrected GFT. At all forecast targets and seasons, and for all but two regions, inclusion of GFT lowered MSE. Results from two alternative error measures, mean absolute error and mean absolute proportional error, were largely consistent with results from MSE. Taken together these findings provide an error profile of GFT in the US, establish strong evidence for the adoption of search trends based 'nowcasts' in influenza forecast systems, and encourage reevaluation of the utility of this data source in diverse domains.

Anatomy of a seasonal influenza epidemic forecast

Bayesian methods have been used to predict the timing of infectious disease epidemics in various settings and for many infectious diseases, including seasonal influenza. But integrating these techniques into public health practice remains an ongoing challenge, and requires close collaboration between modellers, epidemiologists, and public health staff. During the 2016 and 2017 Australian influenza seasons, weekly seasonal influenza forecasts were produced for cities in the three states with the largest populations: Victoria, New South Wales, and Queensland. Forecast results were presented to Health Department disease surveillance units in these jurisdictions, who provided feedback about the plausibility and public health utility of these predictions. In earlier studies we found that delays in reporting and processing of surveillance data substantially limited forecast performance, and that incorporating climatic effects on transmission improved forecast performance. In this study of the 2016 and 2017 seasons, we sought to refine the forecasting method to account for delays in receiving the data, and used meteorological data from past years to modulate the force of infection. We demonstrate how these refinements improved the forecast’s predictive capacity, and use the 2017 influenza season to highlight challenges in accounting for population and clinician behaviour changes in response to a severe season.

Dynamic Bayesian Influenza Forecasting in the United States with Hierarchical Discrepancy (with Discussion)

Timely and accurate forecasts of seasonal influenza would assist public health decision-makers in planning intervention strategies, efficiently allocating resources, and possibly saving lives. For these reasons, influenza forecasts are consequential. Producing timely and accurate influenza forecasts, however, have proven challenging due to noisy and limited data, an incomplete understanding of the disease transmission process, and the mismatch between the disease transmission process and the data-generating process. In this paper, we introduce a dynamic Bayesian (DB) flu forecasting model that exploits model discrepancy through a hierarchical model. The DB model allows forecasts of partially observed flu seasons to borrow discrepancy information from previously observed flu seasons. We compare the DB model to all models that competed in the CDC’s 2015–2016 and 2016–2017 flu forecasting challenges. The DB model outperformed all models in both challenges, indicating the DB model is a leading influenza forecasting model.