Personalized Event-Based Surveillance and Alerting Support for the Assessment of Risk

TL;DR

This paper presents a personalized approach to event-based disease surveillance, combining filtering and re-ranking techniques to help public health officials efficiently assess risks from large streams of disease reports.

Contribution

It introduces a novel two-step process integrating filtering and re-ranking using collaborative filtering and personalization for improved risk assessment support.

Findings

Enhanced filtering reduces irrelevant signals.

Re-ranking improves relevance based on user feedback.

Supports tailored risk assessment for public health officials.

Abstract

In a typical Event-Based Surveillance setting, a stream of web documents is continuously monitored for disease reporting. A structured representation of the disease reporting events is extracted from the raw text, and the events are then aggregated to produce signals, which are intended to represent early warnings against potential public health threats. To public health officials, these warnings represent an overwhelming list of "one-size-fits-all" information for risk assessment. To reduce this overload, two techniques are proposed. First, filtering signals according to the user's preferences (e.g., location, disease, symptoms, etc.) helps reduce the undesired noise. Second, re-ranking the filtered signals, according to an individual's feedback and annotation, allows a user-specific, prioritized ranking of the most relevant warnings. We introduce an approach that takes into account this two-step process of: 1) filtering and 2) re-ranking the results of reporting signals. For this, Collaborative Filtering and Personalization are common techniques used to support users in dealing with the large amount of information that they face.