Intelligent use of digital data for surveillance: The case for ontologies

Session organisers: Fernanda Dórea1 and Crawford Revie2
1 National Veterinary Institute, Sweden
2 Atlantic Veterinary College, Univ. of Prince Edward Island, Canada

E-mail: nanda@datadrivensurveillance.org


Issues of data interpretation and data integration are a major bottleneck for the efficient transformation of data into information for health surveillance. Lack of data standardisation is often blamed, but a great number of data standards already exist and are not adopted. Data coding is a time consuming task and it is not feasible – or even desirable – to expect that human resources will be widely employed to create structured datasets.

How do we delegate the job of extracting information from unstructured data to computers?

Ontologies are knowledge models understandable by both humans and machines. Building ontologies, we focus human resources, in particular domain-expert knowledge, into building knowledge models rather than coding data. This allows such knowledge to be reused again and again, creating truly interoperable systems capable of reasoning with data without relying on static and ineffective local data standards.

Learning outcomes

Attending participants can expect to learn:

  • What ontologies are, in the context of information science
  • How to model domain-specific knowledge using ontological models (concepts and semantic relationships)
  • How to deploy an ontology to perform the task of extracting information from data in an automated manner
  • How ontologies can be used for automated reasoning with medical knowledge.

Prerequisites

No prior knowledge of the subject is needed. Participants are expected to have an interest in the automated interpretation of data by computers, such as in the case of classification of records into syndromes for syndromic surveillance.

Content and structure

The goal of the workshop is to allow participants to understand what ontologies are, and how they can be used to enable computers to transform data into actionable information.

The workshop will be centred around the example of the “Animal Health Surveillance Ontology” (AHSO), which is being developed to provide a common surveillance language that is understandable both by humans and machines, and does not rely on the use of coded data.

Following a brief theoretical introduction, participants will be guided through a practical exercise which will demonstrate how ontologies are built. That is, participants will model a specific piece of domain knowledge, in the area of animal health surveillance, using description logics. After a graphical model has been built in this practical exercise, workshop facilitators will demonstrate the process of coding this model into an ontology.

Instructors will then discuss how the ontology would be used in an applied problem. This overview will allow participants to understand the cycle of construction and use of ontologies; and provide some understanding as to who would be the actors and what would be the roles involved in the development and implementation of ontologies to support data-driven surveillance.

Material

Participants will be expected to work on their own laptops. Participants will receive a link, prior to the workshop, to access any reading materials and free (online) software.

Maximum number of participants

30 people


Instructors

Crawford Revie crawfordrevie@gmail.com

Crawford is a computing scientist who came to veterinary epidemiology by way of a doctorate in mathematical modelling. He currently holds the Canada Research Chair in Epi-informatics, roughly translated as the application of a wide range of informatics tools and approaches to epidemiology. Prior to moving to the Atlantic Veterinary College he was based in Glasgow where we was a member of emerging groups in the areas of Veterinary Informatics and Quantitative Epidemiology at the two Scottish vet schools. A major focus of his research over the past decade has been the application of data-driven models to disease control in a range of veterinary contexts. Recently he has also led research projects in the area of syndromic surveillance, working with a range of animal species in both Canada and sub-Saharan Africa.

Fernanda Dórea / nanda@datadrivensurveillance.org

Fernanda is a veterinarian from Brazil. She contributed to various animal health programs in Brazil, as an Epidemiologist in the headquarters of the Brazilian Ministry of Agriculture, Livestock and Food Supply. Driven by her growing interest in quantitative epidemiology, she pursued a Masters degree in Infectious Disease Modeling at the University of Georgia, USA. As part of her PhD (finished in 2013), under the supervision of Drs. Javier Sanchez and Crawford Revie, she developed a syndromic surveillance system based in laboratory submissions. Her research, now at the National Veterinary Institute in Sweden, continues to focus on exploring ways to automatically extract surveillance information from electronically available data in animal health.