Unsupervised Extraction of Epidemic Syndromes from Participatory Influenza Surveillance Self-reported Symptoms

Abstract

Seasonal influenza surveillance is usually carried out by sentinel general practitioners who compile weekly reports based on the number of influenza-like illness (ILI) clinical cases observed among visited patients. This practice for surveillance is generally affected by two main issues: i) reports are usually released with a lag of about one week or more, ii) the definition of a case of influenza-like illness based on patients symptoms varies from one surveillance system to the other, i.e. from one country to the other. The availability of novel data streams for disease surveillance can alleviate these issues; in this paper, we employed data from Influenzanet, a participatory web-based surveillance project which collects symptoms directly from the general population in real time. We developed an unsupervised probabilistic framework that combines time series analysis of symptoms counts and performs an algorithmic detection of groups of symptoms, hereafter called syndromes. Symptoms counts were collected through the participatory web-based surveillance platforms of a consortium called Influenzanet which is found to correlate with Influenza-like illness incidence as detected by sentinel doctors. Our aim is to suggest how web-based surveillance data can provide an epidemiological signal capable of detecting influenza-like illness' temporal trends without relying on a specific case definition. We evaluated the performance of our framework by showing that the temporal trends of the detected syndromes closely follow the ILI incidence as reported by the traditional surveillance, and consist of combinations of symptoms that are compatible with the ILI definition. The proposed framework was able to predict quite accurately the ILI trend of the forthcoming influenza season based only on the available information of the previous years. Moreover, we assessed the generalisability of the approach by evaluating its potentials for the detection of gastrointestinal syndromes. PLOS 1/24We evaluated the approach against the traditional surveillance data and despite the limited amount of data, the gastrointestinal trend was successfully detected. The result is a real-time flexible surveillance and prediction tool that is not constrained by any disease case definition. Author summaryThis study suggests how web-based surveillance data can provide an epidemiological signal capable of detecting influenza-like illness' temporal trends without relying on a specific case definition. The proposed framework was able to predict quite accurately the ILI trend of the forthcoming influenza season based only on the available information of the previous years. Moreover, we assessed the generalisability of the approach by evaluating its potentials for the detection of gastrointestinal syndromes. We evaluated the approach against the traditional surveillance data and despite the limited amount of data, the gastrointestinal trend was successfully detected. The result is a real-time flexible surveillance and prediction tool that is not constrained by ...

Elucidating user behaviours in a digital health surveillance system to correct prevalence estimates

“…Given a uniform prior Beta(1, 1) and a binomial likelihood, the posterior distribution of is then Beta( X w + 1, 𝒩 w − X w + 1). This results in the mode of the posterior distribution of as 𝒩 w / X w , conventional point estimate often used in the literature (See [6, 7, 15, 21]). We define this posterior distribution of as the naïve estimate, which is an estimate for the prevalence of ILI in the population, without consideration of user behaviours.…”

Syndromic Surveillance of Infectious Diseases

“…Another innovative approach consists of involving the public directly in disease surveillance. An example of a participatory disease surveillance is the Influenzanet project, which operates in ten European countries [53]. It is a web-based surveillance system, established in 2011, for monitoring seasonal influenza epidemics.…”

“…It is a web-based surveillance system, established in 2011, for monitoring seasonal influenza epidemics. It relies on volunteers from the general population who monitor and report their symptoms weekly, through internetbased surveys [53]. This methodology can be regarded as a near-real-time-flexible surveillance structure, which can capture data of several groups of syndromes, not restricted to influenza alone [53].…”

“…It relies on volunteers from the general population who monitor and report their symptoms weekly, through internetbased surveys [53]. This methodology can be regarded as a near-real-time-flexible surveillance structure, which can capture data of several groups of syndromes, not restricted to influenza alone [53]. Overall, the ILI incidence that was determined through influenza net data was found to correlate well with the ILI incidence that was reported by the European Center for Disease Prevention and Control Control (ECDC) [54] and by individual countries' traditional surveillance systems.…”

Correlation-based Discovery of Disease Patterns for Syndromic Surveillance

“…The problem of learning syndrome definitions in a datadriven way, without relying on expert knowledge, has for example been addressed by Kalimeri et al [16]. The authors of this work propose an unsupervised, probabilistic framework based on matrix factorization.…”