COPE: Childhood Obesity Prevention [Knowledge] Enterprise Arash Shaban-Nejad1, David L. Buckeridge1, and Laurette Dubé2 1 McGill Clinical & Health Informatics, Department of Epidemiology and Biostatistics, Faculty of Medicine, McGill University, 1140 Pine Ave. W., Montreal, Canada, H3A 1A3 {arash.shaban-nejad,david.buckeridge}@mcgill.ca 2 Desautels Faculty of Management, McGill University, 1001 Sherbrooke St. West, Montreal, Canada H3A 1G5
[email protected]
Abstract. This paper presents our work-in-progress on designing and implementing an integrated ontology for widespread knowledge dissemination in the domain of obesity with emphasis on childhood obesity. The COPE ontology aims to support a knowledge-based infrastructure to promote healthy eating habits and lifestyles, analyze children's behaviors and habits associated with obesity and to prevent or reduce the prevalence of childhood obesity and overweight. By formally integrating and harmonizing multiple knowledge sources across disciplinary boundaries, we will facilitate cross-sectional analysis of the domain of obesity and generate both generic and customized preventive recommendations, which take into consideration several factors, including existing conditions in individuals and communities. Keywords: Biomedical ontologies, Childhood obesity prevention, Knowledge modeling, Behavioral analysis.
1 Introduction Obesity is well known [1] as one of the major risk factors for several diseases, including: hypertension or high blood pressure; coronary heart disease; type 2 diabetes; stroke; gallbladder disease; osteoarthritis; sleep apnea and other breathing problems; some cancers, such as breast, colon, and endometrial cancer; and mental health problems, such as low self-esteem and depression. Addressing childhood obesity prevention as one of today’s most complex health and socioeconomic challenges is tied to the need for better alignment between different sources of data on human behavior, the health system, nutrition and nutrition-related health problems (e.g., child and maternal nutrition, food safety), media, and markets at local, national, and global levels. It does not seem feasible to deal with such complexity without a clear understanding of the active components and parameters, their interactions at different levels, and how they shape individual, organizational, and collective choice, at any point in time and under varying conditions. While obesity traditionally has been described as “an excess amount of body fat” in general population, no consensus exists on the description and classification of obesity in children [2]. A prerequisite for decision making M. Peleg, N. Lavrač, and C. Combi (Eds.): AIME 2011, LNAI 6747, pp. 225–229, 2011. © Springer-Verlag Berlin Heidelberg 2011
226
A. Shaban-Nejad, D.L. Buckeridge, and L. Dubé
and policy analysis in a domain is capturing the domain’s knowledge and standardization of terminologies. One way to achieve this in modern knowledge-based systems is through ontologies, which are used as a formal medium for sharing common vocabularies, providing semantic annotation and integration, indexing and reasoning support [3], and capturing behavioral knowledge. In order to facilitate collaborative medicalsocial research in the domain of healthy eating and child obesity prevention, the Childhood Obesity Prevention [Knowledge] Enterprise (COPE) project aims to employ state of the art techniques, tools and skills and generate a consistent multidisciplinary semantic platform to capture the social, environmental, economical and behavioral knowledge in this domain.
2 The COPE Ontology Design Our approach employs the emerging technologies in Knowledge Representation (KR) and Semantic Web, including OWL 2.0 [4], advanced logical reasoning services, and social networking to develop an integrated knowledge base for the obesity domain. We classify the domain into five major categories that describe the areas of a healthy lifestyle in terms of food and nutrition, associated diseases (i.e., obesity and related chronic diseases), social- environmental factors, behavioral parameters and media (Figure 1). Defining appropriate rules and axioms along with different associated relationships between the individuals of those categories enables us to derive nontrivial inferencing from our ontology to support decision making in this area. As an example, the concepts such as “overweight person” should be defined based on several parameters, such as the person’s height, age, body mass, and gender. The ontology defines the semantic relationship between different dependent components and regulates their interactions through a set of rules. As the formal representation language we use OWL 2.0 along with Description Logics (DLs). OWL 2.0 is the extension of OWL to support more expressive knowledge modeling by adding features such as the possibility of defining property chains, richer data types, data ranges, and qualified cardinality restrictions. Using OWL 2.0 as one of the World Wide Web Consortium (W3C) recommendations enables us to model classes, properties, and individuals, and define the new datatypes through more expressive semantics (in comparison to classic OWL) for developing ontologies that are exchanged as RDF (Resource Description Framework) graphs. For the reasoning we have used logical reasoners such as RACER [5], and Pellet [6]. The iterative, collaborative nature of an ontology development life cycle requires that ontologies go through one or more processes, such as matching, mapping, merging, alignment, integration, debugging, and versioning. To develop an integrated ontology that can be reused, in whole or in part, by different tools and algorithms, we study different relationships (explicit or implicit) and potential matching between components of different knowledge sources. When integrating existing controlled vocabularies and ontologies from different domains, we also needed to deal with several mismatches at the language (syntax), and conceptual (semantics) levels.
COPE: Childhood Obesity Prevention [Knowledge] Enterprise
227
Fig. 1. An abstract representation of the major components and their interactions in the COPE ontology
2.1 Target Data Sources Due to the interdisciplinary nature of the domain, the COPE ontology as an integrated knowledge base, is implemented on the basis of several textual resources, existing controlled vocabularies and thesauruses, blogs and databases, including RAMQ1 (physician services and pharmaceutical prescriptions), Canadian Community Health Survey (CCHS2) (population health database that represents information on health status, health care utilization, and health determinants for Canadians), CARTaGENE3, AGROVOC4 (FAO agricultural thesaurus), and more. CARTaGENE provides information on medical history, genealogical data (to study family medical history), lifestyles, laboratory blood tests, biological samples, and physical measurements. Other important resources to analyze and determine consumer behavior outcomes are 1
Régie de l'assurance maladie du Québec: http://www.ramq.gouv.qc.ca/index_en.shtml
2
Canadian Community Health Survey (CCHS): http://www.statcan.gc.ca/concepts/health-sante/index-eng.htm 3 http://www.cartagene.qc.ca/index.php?lang=english 4 AGROVOC : http://aims.fao.org/agrovoc/
228
A. Shaban-Nejad, D.L. Buckeridge, and L. Dubé
Statistics Canada5 and Statistics Quebec (ISQ)6, consisting of the Survey of Household Spending (SHS)7, the National Population Health Survey (NPHS)8 (provides socio-demographic information about health of Canadians through different timestamps and related parameters) and private sectors. Also, several other surveys, such as those on regional transportation, living standards, and physical activities for different communities, will be used for cross-community comparison at different levels of granularity. Moreover we obtain the data on how people get information, consume media and buy goods and services through our collaboration with our partner in private sector. These datasets (e.g. data on retail measurement, demographics and store locations, and media/advertising measurement) can be linked together using demographic “hooks” or geographic coordinates and fused to third party data sets such as the BMI Survey. We will also collect our own primary data through multi-purpose individual-level questionnaires whenever we face a gap in the available resources.
3 Discussion The availability of an integrated consensus knowledge on different (biological, nutritional, geographical, environmental, behavioral and social) factors relating to childhood obesity should be regarded as a vital component of any national/global surveillance system that continuously “tracks progress toward meeting the overall health objective of reducing and/or eliminating health problems across populations” [7]. COPE as an integrated data model consisting of nutrition, obesity and chronic diseases, behaviors, media, and marketing can be used as a basis for data integration and knowledge discovery for multidisciplinary researches. The COPE ontology can be used in various knowledge-based systems at both individual and community levels. It can provide a semantic backbone for a healthy diet recommender system to promote healthy eating habits. The dietary recommendations can be performed not only based on the existing food guides, which mainly focuses on the amount and types of foods, but also on other important factors personalized to the individuals' information (e.g. income, dietary restrictions, behaviors) and environmental parameters (e.g., familyand parent-associated interventions). As an application scenario, from the behavioral point of view, we are currently studying the impact of the new advances in social media on the general dietary behavior of children. In our scenario we emphasize on the dietary patterns associated with selection/consumption of low/high added-sugar food choices. Several types of behaviors (i.e. regulatory, social, habitual, and etc.) involve in a typical food selection/consumption process. The excessive use of social media may give rise to the abnormal behaviors in children (i.e. social media addiction) that affect their ingestive behavior and cause several childhood psychiatric disorders. Many of these mental disorders are also known as the major risk factors for obesity. An individual can regulate her behaviors (i.e. eating behavior) through her choices (e.g. selection of food, life style, etc.) in response to the environment [8]. For example increasing the level of physical activity [9] is recommended as a conscious 5
http://www.statcan.gc.ca/start-debut-eng.html http://www.stat.gouv.qc.ca/default_an.htm 7 http://www.statcan.gc.ca/imdb-bmdi/3508-eng.htm 8 http://www.statcan.gc.ca/concepts/nphs-ensp/index-eng.htm 6
COPE: Childhood Obesity Prevention [Knowledge] Enterprise
229
regulatory behavior to adjust the energy balance and improve behavioral outcome. Our future work will be focused on two aspects: i) enriching the COPE ontological structure with adding more components and more complex axioms to study multifaceted interactions between different components in the domain; ii) incorporating advanced social networking, social marketing and policy development techniques along with Web 2.0 methodologies to support decision making not only at the individual level but also at the community and society level. Acknowledgments. COPE is a part of the Brain-to-Society (BtS) [8] research agenda within the McGill World Platform for Health and Economic Convergence (MWP) [10] initiative.
References 1. OBESITY, Health Canada Catalogue # H13-7/20-2006E-PDF (2006) ISBN # 0-66244192-3 2. HCF Health Report on obesity and weight loss, No 9 (2003), http://www.hcf.com.au/pdf/obesity.pdf 3. Smith, B., Ashburner, M., et al.: The OBO Foundry: Coordinated Evolution of Ontologies to Support Biomedical Data Integration. Nature Biotechnol. 25(11), 1251–1255 (2007) 4. OWL 2.0 Overview, http://www.w3.org/TR/owl2-overview/ 5. Haarslev, V., Möller, R.: RACER System Description. In: Goré, R.P., Leitsch, A., Nipkow, T. (eds.) IJCAR 2001. LNCS (LNAI), vol. 2083, pp. 701–706. Springer, Heidelberg (2001) 6. Pellet: OWL 2 Reasoner for Java, http://clarkparsia.com/pellet/ 7. Singh, G.K., Kogan, M.D., van Dyck, P.C.: Changes in state-specific childhood obesity and overweight prevalence in the United States from 2003 to 2007. Arch. Pediatr. Adolesc. Med. 164(7), 598–607 (2010) 8. Dubé, L., Bechara, A., Böckenholt, U., Ansari, A., Dagher, A., et al.: Towards a brain-tosociety systems model of individual choice. Market Lett 19, 323–336 (2008) 9. Ng, C., Marshall, D., Willows, N.D.: Obesity, adiposity, physical fitness and activity levels in Cree children. Int. J. Circumpolar Health 65(4), 322–330 (2006) 10. McGill World Platform for Health and Economic Convergence, http://www.mcgill.ca/mwp/
