Prediction and surveillance of influenza epidemics

Enhancing infectious diseases early warning: A deep learning approach for influenza surveillance in China

Objective

This study aimed to develop a universally applicable, feedback-informed Self-Excitation Attention Residual Network (SEAR) model. This model dynamically adapts to evolving disease trends and surveillance system changes, accommodating various scenarios. Thereby enhancing the effectiveness of early warning systems.

Methods

Surveillance data on influenza-like illness (ILI) was collected from various regions including Northern China, Southern China, Beijing, and Yunnan. The reproduction number (Rt) was estimated to determine the threshold for issuing warnings. The Self-Excitation Attention Residual Network (SEAR) was devised employing deep learning algorithms and was trained, validated, and tested. The SEAR model's efficacy was assessed based on five metrics: accuracy rate, recall rate, F1 score, confusion matrix, and the receiver operating characteristic curve.

Results

With an advance warning set at three days, the SEAR model outperformed five primary models - logistic regression, support vector machine, random forest, Extreme Gradient Boosting, and Long Short-Term Memory model - in all five evaluation metrics. Notably, the model's warning performance declined with an increase in the early warning value and the number of warning days, albeit maintaining a ROC value over 0.7 in all scenarios.

Conclusion

The SEAR model demonstrated robust early warning performance for influenza in diverse Chinese regions with high accuracy and specificity. This novel model, augmenting traditional systems, supports widespread application for respiratory disease outbreak monitoring. Future evaluations could incorporate alternative indicators, with the model continuously updating through data feedback, thus enhancing its universal applicability. Ongoing optimization, using iterative feedback and expert judgment, heralds a transformative approach to surveillance-based early warning strategies.

Prediction of influenza outbreaks in Fuzhou, China: comparative analysis of forecasting models

BACKGROUND

Influenza is a highly contagious respiratory disease that presents a significant challenge to public health globally. Therefore, effective influenza prediction and prevention are crucial for the timely allocation of resources, the development of vaccine strategies, and the implementation of targeted public health interventions.

METHOD

In this study, we utilized historical influenza case data from January 2013 to December 2021 in Fuzhou to develop four regression prediction models: SARIMA, Prophet, Holt-Winters, and XGBoost models. Their predicted performance was assessed by using influenza data from the period from January 2022 to December 2022 in Fuzhou. These models were used for fitting and prediction analysis. The evaluation metrics, including Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and Mean Absolute Error (MAE), were employed to compare the performance of these models.

RESULTS

The results indicate that the epidemic of influenza in Fuzhou exhibits a distinct seasonal and cyclical pattern. The influenza cases data displayed a noticeable upward trend and significant fluctuations. In our study, we employed SARIMA, Prophet, Holt-Winters, and XGBoost models to predict influenza outbreaks in Fuzhou. Among these models, the XGBoost model demonstrated the best performance on both the training and test sets, yielding the lowest values for MSE, RMSE, and MAE among the four models.

CONCLUSION

The utilization of the XGBoost model significantly enhances the prediction accuracy of influenza in Fuzhou. This study makes a valuable contribution to the field of influenza prediction and provides substantial support for future influenza response efforts.

Syndromic surveillance to detect disease outbreaks using time between emergency department presentations

OBJECTIVE

Early warning of disease outbreaks is paramount for health jurisdictions. The objective of the present study was to develop syndromic surveillance monitoring plans from routinely collected ED data with application to detecting disease outbreaks.

METHODS

The study involved secondary data analysis of ED presentations to major public hospitals in Queensland and South Australia spanning 2017-2020. Monitoring plans were developed for all major Queensland and South Australian public hospitals using an adaptation of Exponentially Weighted Moving Averages - a process control method used in detecting anomalies in industrial production processes. The methods rely on setting a threshold (control limit) relating to the time between an event of interest (e.g. flu outbreak) using ED presentations as a signal to monitor. An outbreak is flagged as this time gets significantly smaller, and each event offers a decision point on whether an outbreak has occurred. The models incorporate differing levels of temporal memory to cover outbreaks of different sizes.

RESULTS

The novel approach to real-time outbreak detection indicates outbreaks for individual hospitals coinciding with the first wave of the COVID-19 outbreak in Queensland and South Australia as well as the large 2017 and 2019 influenza seasons.

CONCLUSION

Outbreak detection models demonstrate the ability to quickly flag an outbreak based on clinician-assigned ED diagnoses. An implemented syndromic surveillance approach can pick up geographic outbreaks quickly so they can be contained. Such capability can help with surveillance related to the current COVID-19 pandemic and potential future pandemics.

Risk-adjusted zero-inflated Poisson CUSUM charts for monitoring influenza surveillance data

Background

The influenza surveillance has been received much attention in public health area. For the cases with excessive zeroes, the zero-inflated Poisson process is widely used. However, the traditional control charts based on zero-inflated Poisson model, ignore the association between influenza cases and risk factors, and thus may lead to unexpected mistakes when implementing monitoring charts.

Method

In this paper, we proposed risk-adjusted zero-inflated Poisson cumulative sum control charts, in which the risk factors were put to adjust the risk of influenza and the adjustment was made by zero-inflated Poisson regression. We respectively proposed the control chart monitoring the parameters individually and simultaneously.

Results

The performance of our proposed risk-adjusted zero-inflated Poisson cumulative sum control chart was evaluated and compared with the unadjusted standard cumulative sum control charts in simulation studies. The results show that for different distribution of impact factors and different coefficients, the risk-adjusted cumulative sum charts can generate much less false alarm than the standard ones. Finally, the influenza surveillance data from Hong Kong is used to illustrate the application of the proposed chart.

Conclusions

Our results suggest that the adjusted cumulative sum control chart we proposed is more accurate and credible than the unadjusted standard control charts because of the lower false alarm rate of the adjusted ones. Even the unadjusted control charts may signal a little faster than the adjusted ones, the alarm they raise may have low credibility since they also raise alarm frequently even the processes are in control. Thus we suggest using the risk-adjusted cumulative sum control charts to monitor the influenza surveillance data to alert accurately, credibly and relatively quickly.

The development and deployment of a model for hospital-level COVID-19 associated patient demand intervals from consistent estimators (DICE)

Hospitals commonly project demand for their services by combining their historical share of regional demand with forecasts of total regional demand. Hospital-specific forecasts of demand that provide prediction intervals, rather than point estimates, may facilitate better managerial decisions, especially when demand overage and underage are associated with high, asymmetric costs. Regional point forecasts of patient demand are commonly available, e.g., for the number of people requiring hospitalization due to an epidemic such as COVID-19. However, even in this common setting, no probabilistic, consistent, computationally tractable forecast is available for the fraction of patients in a region that a particular institution should expect. We introduce such a forecast, DICE (Demand Intervals from Consistent Estimators). We describe its development and deployment at an academic medical center in California during the ‘second wave’ of COVID-19 in the Unite States. We show that DICE is consistent under mild assumptions and suitable for use with perfect, biased and unbiased regional forecasts. We evaluate its performance on empirical data from a large academic medical center as well as on synthetic data.

Seasonal Activity of Influenza in Iran: Application of Influenza-like Illness Data from Sentinel Sites of Healthcare Centers during 2010 to 2015

This study aimed to predict seasonal influenza activity and detection of influenza outbreaks. Data of all registered cases (n = 53,526) of influenza-like illnesses (ILIs) from sentinel sites of healthcare centers in Iran were obtained from the FluNet web-based tool, World Health Organization (WHO), from 2010 to 2015. The status of the ILI activity was obtained from the FluNet and considered as the gold standard of the seasonal activity of influenza during the study period. The cumulative sum (CUSUM) as an outbreak detection method was used to predict the seasonal activity of influenza. Also, time series similarity between the ILI trend and CUSUM was assessed using the cross-correlogram. Of 7684 (14%) positive cases of influenza, about 71% were type A virus and 28% were type B virus. The majority of the outbreaks occurred in winter and autumn. Results of the cross-correlogram showed that there was a considerable similarity between time series graphs of the ILI cases and CUSUM values. However, the CUSUM algorithm did not have a good performance in the timely detection of influenza activity. Despite a considerable similarity between time series of the ILI cases and CUSUM algorithm in weekly lag, the seasonal activity of influenza in Iran could not be predicted by the CUSUM algorithm.

An exhaustive review and analysis on applications of statistical forecasting in hospital emergency departments

Emergency departments (EDs) provide medical treatment for a broad spectrum of illnesses and injuries to patients who arrive at all hours of the day. The quality and efficient delivery of health care in EDs are associated with a number of factors, such as patient overall length of stay (LOS) and admission, prompt ambulance diversion, quick and accurate triage, nurse and physician assessment, diagnostic and laboratory services, consultations and treatment. One of the most important ways to plan the healthcare delivery efficiently is to make forecasts of ED processes. The aim this study is thus to provide an exhaustive review for ED stakeholders interested in applying forecasting methods to their ED processes. A categorisation, analysis and interpretation of 102 papers is performed for review. This exhaustive review provides an insight for researchers and practitioners about forecasting in EDs in terms of showing current state and potential areas for future attempts.

Monitoring Pertussis Infections Using Internet Search Queries

This study aims to assess the utility of internet search query analysis in pertussis surveillance. This study uses an empirical time series model based on internet search metrics to detect the pertussis incidence in Australia. Our research demonstrates a clear seasonal pattern of both pertussis infections and Google Trends (GT) with specific search terms in time series seasonal decomposition analysis. The cross-correlation function showed significant correlations between GT and pertussis incidences in Australia and each state at the lag of 0 and 1 months, with the variation of correlations between 0.17 and 0.76 (p < 0.05). A multivariate seasonal autoregressive integrated moving average (SARIMA) model was developed to track pertussis epidemics pattern using GT data. Reflected values for this model were generally consistent with the observed values. The inclusion of GT metrics improved detective performance of the model (β = 0.058, p < 0.001). The validation analysis indicated that the overall agreement was 81% (sensitivity: 77% and specificity: 83%). This study demonstrates the feasibility of using internet search metrics for the detection of pertussis epidemics in real-time, which can be considered as a pre-requisite for constructing early warning systems for pertussis surveillance using internet search metrics.

Web-based infectious disease surveillance systems and public health perspectives: a systematic review

BACKGROUND

Emerging and re-emerging infectious diseases are a significant public health concern, and early detection and immediate response is crucial for disease control. These challenges have led to the need for new approaches and technologies to reinforce the capacity of traditional surveillance systems for detecting emerging infectious diseases. In the last few years, the availability of novel web-based data sources has contributed substantially to infectious disease surveillance. This study explores the burgeoning field of web-based infectious disease surveillance systems by examining their current status, importance, and potential challenges.

METHODS

A systematic review framework was applied to the search, screening, and analysis of web-based infectious disease surveillance systems. We searched PubMed, Web of Science, and Embase databases to extensively review the English literature published between 2000 and 2015. Eleven surveillance systems were chosen for evaluation according to their high frequency of application. Relevant terms, including newly coined terms, development and classification of the surveillance systems, and various characteristics associated with the systems were studied.

RESULTS

Based on a detailed and informative review of the 11 web-based infectious disease surveillance systems, it was evident that these systems exhibited clear strengths, as compared to traditional surveillance systems, but with some limitations yet to be overcome. The major strengths of the newly emerging surveillance systems are that they are intuitive, adaptable, low-cost, and operated in real-time, all of which are necessary features of an effective public health tool. The most apparent potential challenges of the web-based systems are those of inaccurate interpretation and prediction of health status, and privacy issues, based on an individual's internet activity.

CONCLUSION

Despite being in a nascent stage with further modification needed, web-based surveillance systems have evolved to complement traditional national surveillance systems. This review highlights ways in which the strengths of existing systems can be maintained and weaknesses alleviated to implement optimal web surveillance systems.

Predicting Fluctuating Rates of Hospitalizations in Relation to Influenza Epidemics and Meteorological Factors

Introduction

In France, rates of hospital admissions increase at the peaks of influenza epidemics. Predicting influenza-associated hospitalizations could help to anticipate increased hospital activity. The purpose of this study is to identify predictors of influenza epidemics through the analysis of meteorological data, and medical data provided by general practitioners.

Methods

Historical data were collected from Meteo France, the Sentinelles network and hospitals’ information systems for a period of 8 years (2007–2015). First, connections between meteorological and medical data were estimated with the Pearson correlation coefficient, Principal component analysis and classification methods (Ward and k-means). Epidemic states of tested weeks were then predicted for each week during a one-year period using linear discriminant analysis. Finally, transition probabilities between epidemic states were calculated with the Markov Chain method.

Results

High correlations were found between influenza-associated hospitalizations and the variables: Sentinelles and emergency department admissions, and anti-correlations were found between hospitalizations and each of meteorological factors applying a time lag of: -13, -12 and -32 days respectively for temperature, absolute humidity and solar radiation. Epidemic weeks were predicted accurately with the linear discriminant analysis method; however there were many misclassifications about intermediate and non-epidemic weeks. Transition probability to an epidemic state was 100% when meteorological variables were below: 2°C, 4 g/m³ and 32 W/m², respectively for temperature, absolute humidity and solar radiation. This probability was 0% when meteorological variables were above: 6°C, 5.8g/m³ and 74W/m².

Conclusion

These results confirm a good correlation between influenza-associated hospitalizations, meteorological factors and general practitioner’s activity, the latter being the strongest predictor of hospital activity.

Algorithms for detecting and predicting influenza outbreaks: metanarrative review of prospective evaluations

OBJECTIVES

Reliable monitoring of influenza seasons and pandemic outbreaks is essential for response planning, but compilations of reports on detection and prediction algorithm performance in influenza control practice are largely missing. The aim of this study is to perform a metanarrative review of prospective evaluations of influenza outbreak detection and prediction algorithms restricted settings where authentic surveillance data have been used.

DESIGN

The study was performed as a metanarrative review. An electronic literature search was performed, papers selected and qualitative and semiquantitative content analyses were conducted. For data extraction and interpretations, researcher triangulation was used for quality assurance.

RESULTS

Eight prospective evaluations were found that used authentic surveillance data: three studies evaluating detection and five studies evaluating prediction. The methodological perspectives and experiences from the evaluations were found to have been reported in narrative formats representing biodefence informatics and health policy research, respectively. The biodefence informatics narrative having an emphasis on verification of technically and mathematically sound algorithms constituted a large part of the reporting. Four evaluations were reported as health policy research narratives, thus formulated in a manner that allows the results to qualify as policy evidence.

CONCLUSIONS

Awareness of the narrative format in which results are reported is essential when interpreting algorithm evaluations from an infectious disease control practice perspective.

Influenza surveillance and incidence in a rural area in China during the 2009/2010 influenza pandemic

BACKGROUND

Most influenza surveillance is based on data from urban sentinel hospitals; little is known about influenza activity in rural communities. We conducted influenza surveillance in a rural region of China with the aim of detecting influenza activity in the 2009/2010 influenza season.

METHODS

The study was conducted from October 2009 to March 2010. Real-time polymerase chain reaction was used to confirm influenza cases. Over-the-counter (OTC) drug sales were daily collected in drugstores and hospitals/clinics. Space-time scan statistics were used to identify clusters of ILI in community. The incidence rate of ILI/influenza was estimated on the basis of the number of ILI/influenza cases detected by the hospitals/clinics.

RESULTS

A total of 434 ILI cases (3.88% of all consultations) were reported; 64.71% of these cases were influenza A (H1N1) pdm09. The estimated incidence rate of ILI and influenza were 5.19/100 and 0.40/100, respectively. The numbers of ILI cases and OTC drug purchases in the previous 7 days were strongly correlated (Spearman rank correlation coefficient [r] = 0.620, P = 0.001). Four ILI outbreaks were detected by space-time permutation analysis.

CONCLUSIONS

This rural community surveillance detected influenza A (H1N1) pdm09 activity and outbreaks in the 2009/2010 influenza season and enabled estimation of the incidence rate of influenza. It also provides a scientific data for public health measures.

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.

Medical and economic burden of influenza in the elderly population in central and eastern European countries

Influenza affects 5-15% of the population during an epidemic. In Western Europe, vaccination of at-risk groups forms the cornerstone of influenza prevention. However, vaccination coverage of the elderly (> 65 y) is often low in Central and Eastern Europe (CEE); potentially because a paucity of country-specific data limits evidence-based policy making. Therefore the medical and economic burden of influenza were estimated in elderly populations in the Czech Republic, Hungary, Kazakhstan, Poland, Romania, and Ukraine. Data covering national influenza vaccination policies, surveillance and reporting, healthcare costs, populations, and epidemiology were obtained via literature review, open-access websites and databases, and interviews with experts. A simplified model of patient treatment flow incorporating cost, population, and incidence/prevalence data was used to calculate the influenza burden per country. In the elderly, influenza represented a large burden on the assessed healthcare systems, with yearly excess hospitalization rates of ~30/100,000. Burden varied between countries and was likely influenced by population size, surveillance system, healthcare provision, and vaccine coverage. The greatest burden was found in Poland, where direct costs were over EUR 5 million. Substantial differences in data availability and quality were identified, and to fully quantify the burden of influenza in CEE, influenza reporting systems should be standardized. This study most probably underestimates the real burden of influenza, however the public health problem is recognized worldwide, and will further increase with population aging. Extending influenza vaccination of the elderly may be a cost-effective way to reduce the burden of influenza in CEE.

Influenza forecasting with Google Flu Trends

BACKGROUND

We developed a practical influenza forecast model based on real-time, geographically focused, and easy to access data, designed to provide individual medical centers with advanced warning of the expected number of influenza cases, thus allowing for sufficient time to implement interventions. Secondly, we evaluated the effects of incorporating a real-time influenza surveillance system, Google Flu Trends, and meteorological and temporal information on forecast accuracy.

METHODS

Forecast models designed to predict one week in advance were developed from weekly counts of confirmed influenza cases over seven seasons (2004-2011) divided into seven training and out-of-sample verification sets. Forecasting procedures using classical Box-Jenkins, generalized linear models (GLM), and generalized linear autoregressive moving average (GARMA) methods were employed to develop the final model and assess the relative contribution of external variables such as, Google Flu Trends, meteorological data, and temporal information.

RESULTS

A GARMA(3,0) forecast model with Negative Binomial distribution integrating Google Flu Trends information provided the most accurate influenza case predictions. The model, on the average, predicts weekly influenza cases during 7 out-of-sample outbreaks within 7 cases for 83% of estimates. Google Flu Trend data was the only source of external information to provide statistically significant forecast improvements over the base model in four of the seven out-of-sample verification sets. Overall, the p-value of adding this external information to the model is 0.0005. The other exogenous variables did not yield a statistically significant improvement in any of the verification sets.

CONCLUSIONS

Integer-valued autoregression of influenza cases provides a strong base forecast model, which is enhanced by the addition of Google Flu Trends confirming the predictive capabilities of search query based syndromic surveillance. This accessible and flexible forecast model can be used by individual medical centers to provide advanced warning of future influenza cases.

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.

Can linked emergency department data help assess the out-of-hospital burden of acute lower respiratory infections? A population-based cohort study

Background

There is a lack of data on the out-of-hospital burden of acute lower respiratory infections (ALRI) in developed countries. Administrative datasets from emergency departments (ED) may assist in addressing this.

Methods

We undertook a retrospective population-based study of ED presentations for respiratory-related reasons linked to birth data from 245,249 singleton live births in Western Australia. ED presentation rates <9 years of age were calculated for different diagnoses and predictors of ED presentation <5 years were assessed by multiple logistic regression.

Results

ED data from metropolitan WA, representing 178,810 births were available for analysis. From 35,136 presentations, 18,582 (52.9%) had an International Classification of Diseases (ICD) code for ALRI and 434 had a symptom code directly relating to an ALRI ICD code. A further 9600 presentations had a non-specific diagnosis. From the combined 19,016 ALRI presentations, the highest rates were in non-Aboriginal children aged 6–11 months (81.1/1000 child-years) and Aboriginal children aged 1–5 months (314.8/1000). Croup and bronchiolitis accounted for the majority of ALRI ED presentations. Of Aboriginal births, 14.2% presented at least once to ED before age 5 years compared to 6.5% of non-Aboriginal births. Male sex and maternal age <20 years for Aboriginal children and 20–29 years for non-Aboriginal children were the strongest predictors of presentation to ED with ALRI.

Conclusions

ED data can give an insight into the out-of-hospital burden of ALRI. Presentation rates to ED for ALRI were high, but are minimum estimates due to current limitations of the ED datasets. Recommendations for improvement of these data are provided. Despite these limitations, ALRI, in particular bronchiolitis and croup are important causes of presentation to paediatric EDs.

Review: a critical evaluation of arguments opposing male circumcision for HIV prevention in developed countries

A potential impediment to evidence-based policy development on medical male circumcision (MC) for HIV prevention in all countries worldwide is the uncritical acceptance by some of arguments used by opponents of this procedure. Here we evaluate recent opinion-pieces of 13 individuals opposed to MC. We find that these statements misrepresent good studies, selectively cite references, some containing fallacious information, and draw erroneous conclusions. In marked contrast, the scientific evidence shows MC to be a simple, low-risk procedure with very little or no adverse long-term effect on sexual function, sensitivity, sensation during arousal or overall satisfaction. Unscientific arguments have been recently used to drive ballot measures aimed at banning MC of minors in the USA, eliminate insurance coverage for medical MC for low-income families, and threaten large fines and incarceration for health care providers. Medical MC is a preventative health measure akin to immunisation, given its protective effect against HIV infection, genital cancers and various other conditions. Protection afforded by neonatal MC against a diversity of common medical conditions starts in infancy with urinary tract infections and extends throughout life. Besides protection in adulthood against acquiring HIV, MC also reduces morbidity and mortality from multiple other sexually transmitted infections (STIs) and genital cancers in men and their female sexual partners. It is estimated that over their lifetime one-third of uncircumcised males will suffer at least one foreskin-related medical condition. The scientific evidence indicates that medical MC is safe and effective. Its favourable risk/benefit ratio and cost/benefit support the advantages of medical MC.