Accounting for Healthcare-Seeking Behaviours and Testing Practices in Real-Time Influenza Forecasts

Abstract

For diseases such as influenza, where the majority of infected persons experience mild (if any) symptoms, surveillance systems are sensitive to changes in healthcare-seeking and clinical decision-making behaviours. This presents a challenge when trying to interpret surveillance data in near-real-time (e.g., to provide public health decision-support). Australia experienced a particularly large and severe influenza season in 2017, perhaps in part due to: (a) mild cases being more likely to seek healthcare; and (b) clinicians being more likely to collect specimens for reverse transcription polymerase chain reaction (RT-PCR) influenza tests. In this study, we used weekly Flutracking surveillance data to estimate the probability that a person with influenza-like illness (ILI) would seek healthcare and have a specimen collected. We then used this estimated probability to calibrate near-real-time seasonal influenza forecasts at each week of the 2017 season, to see whether predictive skill could be improved. While the number of self-reported influenza tests in the weekly surveys are typically very low, we were able to detect a substantial change in healthcare seeking behaviour and clinician testing behaviour prior to the high epidemic peak. Adjusting for these changes in behaviour in the forecasting framework improved predictive skill. Our analysis demonstrates a unique value of community-level surveillance systems, such as Flutracking, when interpreting traditional surveillance data. These methods are also applicable beyond the Australian context, as similar community-level surveillance systems operate in other countries.

Early analysis of the Australian COVID-19 epidemic

“…The forecasting method combines an SEEIIR (susceptible-exposed-infectious-recovered) population model of infection with daily COVID-19 case notification counts, through the use of a bootstrap particle filter ( Arulampalam et al, 2002 ). This approach is similar to that implemented and described in Moss et al, 2019b , in the context of seasonal influenza forecasts for several major Australian cities. Briefly, the particle filter method uses post-regularisation ( Doucet et al, 2001 ), with a deterministic resampling stage ( Kitagawa, 1996 ).…”

Resurgence of SARS-CoV-2: Detection by community viral surveillance

“…Data streams based on respiratory symptoms, such as those used for COVID-19 surveillance in most countries are prone to biases that can obscure underlying trends, such as variations in test availability and test-seeking behavior (10). Some countries have augmented these systems with surveys of virus prevalence in the wider population, but these have mostly been one-off activities, for example, in Wuhan, China (11), or were designed explicitly as interventions, for example, in Slovakia (12).…”

Prediction model for COVID-19 patient visits in the ambulatory setting

“…The popular susceptible, infected, and recovered (SIR) epidemiologic model and variations of this model have been used to gauge community spread of a variety of infectious diseases such as in uenza and dengue fever. [8][9][10][11][12] SIR models have also been applied to inpatient settings to predict hospital capacity regarding admissions, ICU beds, and ventilators. 9,11,13,14 In addition to SIR models, the current literature on predicting patient volume varies from descriptive statistics to advanced time series models, with most of the studies that have used time series forecasting models focusing on emergency department and hospital admissions.…”