Methods for detecting seasonal influenza epidemics using a school absenteeism surveillance system

ATQ: alert time quality, an evaluation metric for assessing timely epidemic detection models within a school absenteeism-based surveillance system

BACKGROUND

Wellington-Dufferin-Guelph Public Health (WDGPH) has conducted an absenteeism-based influenza surveillance program in the WDG region of Ontario, Canada since 2008, using a 10% absenteeism threshold to raise an alert for the implementation of mitigating measures. A recent study indicated that model-based alternatives, such as distributed lag seasonal logistic regression models, provided improved alerts for detecting an upcoming epidemic. However model evaluation and selection was primarily based on alert accuracy, measured by the false alert rate (FAR), and failed to optimize timeliness. Here, a new metric that simultaneously evaluates epidemic alert accuracy and timeliness is proposed. The alert time quality (ATQ) metric is investigated as a model selection criterion on both a simulated and real data set.

METHODS

The ATQ assessed alerts on a gradient, where alerts raised incrementally before or after an optimal day were considered informative, but were penalized for lack of timeliness. Summary statistics of ATQ, average alert time quality (AATQ) and first alert time quality (FATQ), were used for model evaluation and selection. Alerts raised by ATQ and FAR selected models were compared. Daily elementary school absenteeism and laboratory-confirmed influenza case data collected by WDGPH were used for demonstration and evaluation of the proposed metric. A simulation study that mimicked the WDG population and influenza demographics was conducted for further evaluation of the proposed metric.

RESULTS

The FATQ-selected model raised acceptable first alerts most frequently, while the AATQ-selected model raised first alerts within the ideal range most frequently.

CONCLUSIONS

Models selected by either FATQ or AATQ would more effectively predict community influenza activity with the local community than those selected by FAR.

Monitoring School Absenteeism for Influenza-Like Illness Surveillance: Systematic Review and Meta-analysis

Background

Influenza causes considerable disease burden each year, particularly in children. Monitoring school absenteeism has long been proposed as a surveillance tool of influenza activity in the community, but the practice of school absenteeism could be varying, and the potential of such usage remains unclear.

Objective

The aim of this paper is to determine the potential of monitoring school absenteeism as a surveillance tool of influenza.

Methods

We conducted a systematic review of the published literature on the relationship between school absenteeism and influenza activity in the community. We categorized the types of school absenteeism and influenza activity in the community to determine the correlation between these data streams. We also extracted this correlation with different lags in community surveillance to determine the potential of using school absenteeism as a leading indicator of influenza activity.

Results

Among the 35 identified studies, 22 (63%), 12 (34%), and 8 (23%) studies monitored all-cause, illness-specific, and influenza-like illness (ILI)–specific absents, respectively, and 16 (46%) used quantitative approaches and provided 33 estimates on the temporal correlation between school absenteeism and influenza activity in the community. The pooled estimate of correlation between school absenteeism and community surveillance without lag, with 1-week lag, and with 2-week lag were 0.44 (95% CI 0.34, 0.53), 0.29 (95% CI 0.15, 0.42), and 0.21 (95% CI 0.11, 0.31), respectively. The correlation between influenza activity in the community and ILI-specific absenteeism was higher than that between influenza activity in community all-cause absenteeism. Among the 19 studies that used qualitative approaches, 15 (79%) concluded that school absenteeism was in concordance with, coincided with, or was associated with community surveillance. Of the 35 identified studies, only 6 (17%) attempted to predict influenza activity in the community from school absenteeism surveillance.

Conclusions

There was a moderate correlation between school absenteeism and influenza activity in the community. The smaller correlation between school absenteeism and community surveillance with lag, compared to without lag, suggested that careful application was required to use school absenteeism as a leading indicator of influenza epidemics. ILI-specific absenteeism could monitor influenza activity more closely, but the required resource or school participation willingness may require careful consideration to weight against the associated costs. Further development is required to use and optimize the use of school absenteeism to predict influenza activity. In particular, the potential of using more advanced statistical models and validation of the predictions should be explored.

Applying the Moving Epidemic Method to Establish the Influenza Epidemic Thresholds and Intensity Levels for Age-Specific Groups in Hubei Province, China

BACKGROUND

School-aged children were reported to act as the main transmitter during influenza epidemic seasons. It is vital to set up an early detection method to help with the vaccination program in such a high-risk population. However, most relative studies only focused on the general population. Our study aims to describe the influenza epidemiology characteristics in Hubei Province and to introduce the moving epidemic method to establish the epidemic thresholds for age-specific groups.

METHODS

We divided the whole population into pre-school, school-aged and adult groups. The virology data from 2010/2011 to 2017/2018 were applied to the moving epidemic method to establish the epidemic thresholds for the general population and age-specific groups for the detection of influenza in 2018/2019. The performances of the model were compared by the cross-validation process.

RESULTS

The epidemic threshold for school-aged children in the 2018/2019 season was 15.42%. The epidemic thresholds for influenza A virus subtypes H1N1 and H3N2 and influenza B were determined as 5.68%, 6.12% and 10.48%, respectively. The median start weeks of the school-aged children were similar to the general population. The cross-validation process showed that the sensitivity of the model established with school-aged children was higher than those established with the other age groups in total influenza, H1N1 and influenza B, while it was only lower than the general population group in H3N2.

CONCLUSIONS

This study proved the feasibility of applying the moving epidemic method in Hubei Province. Additional influenza surveillance and vaccination strategies should be well-organized for school-aged children to reduce the disease burden of influenza in China.

Development and assessment of a hospital admissions-based syndromic surveillance system for COVID-19 in Ontario, Canada: ACES Pandemic Tracker

BACKGROUND

The COVID-19 pandemic has continued to pose a major global public health risk. The importance of public health surveillance systems to monitor the spread and impact of COVID-19 has been well demonstrated. The purpose of this study was to describe the development and effectiveness of a real-time public health syndromic surveillance system (ACES Pandemic Tracker) as an early warning system and to provide situational awareness in response to the COVID-19 pandemic in Ontario, Canada.

METHODS

We used hospital admissions data from the Acute Care Enhanced Surveillance (ACES) system to collect data on pre-defined groupings of symptoms (syndromes of interest; SOI) that may be related to COVID-19 from 131 hospitals across Ontario. To evaluate which SOI for suspected COVID-19 admissions were best correlated with laboratory confirmed admissions, laboratory confirmed COVID-19 hospital admissions data were collected from the Ontario Ministry of Health. Correlations and time-series lag analysis between suspected and confirmed COVID-19 hospital admissions were calculated. Data used for analyses covered the period between March 1, 2020 and September 21, 2020.

RESULTS

Between March 1, 2020 and September 21, 2020, ACES Pandemic Tracker identified 22,075 suspected COVID-19 hospital admissions (150 per 100,000 population) in Ontario. After correlation analysis, we found laboratory-confirmed hospital admissions for COVID-19 were strongly and significantly correlated with suspected COVID-19 hospital admissions when SOI were included (Spearman's rho = 0.617) and suspected COVID-19 admissions when SOI were excluded (Spearman's rho = 0.867). Weak to moderate significant correlations were found among individual SOI. Laboratory confirmed COVID-19 hospital admissions lagged in reporting by 3 days compared with suspected COVID-19 admissions when SOI were excluded.

CONCLUSIONS

Our results demonstrate the utility of a hospital admissions syndromic surveillance system to monitor and identify potential surges in severe COVID-19 infection within the community in a timely manner and provide situational awareness to inform preventive and preparatory health interventions.

A stacked ensemble method for forecasting influenza-like illness visit volumes at emergency departments

Accurate and reliable short-term forecasts of influenza-like illness (ILI) visit volumes at emergency departments can improve staffing and resource allocation decisions within hospitals. In this paper, we developed a stacked ensemble model that averages the predictions from various competing methodologies in the current frontier for ILI-related forecasts. We also constructed a back-of-the-envelope prediction interval for the stacked ensemble, which provides a conservative characterization of the uncertainty in the stacked ensemble predictions. We assessed the accuracy and reliability of our model with 1 to 4 weeks ahead forecast targets using real-time hospital-level data on weekly ILI visit volumes during the 2012-2018 flu seasons in the Alberta Children's Hospital, located in Calgary, Alberta, Canada. Our results suggest the forecasting performance of the stacked ensemble meets or exceeds the performance of the individual models over all forecast targets.