AN EVIDENCE-BASED ONTOLOGY OF BELIEFS AND ATTITUDES TOWARD PROSTATE CANCER SCREENING
John H. Holmes, PhD, University of Pennsylvania School of Medicine, Philadelphia, PA
Purpose. To create a formal knowledge model, or ontology, that describes the concepts related to prostate cancer screening (PCS) beliefs and attitudes and the relationships between them. The ontology is being used to create a machine-readable knowledge base in a computerized behavioral intervention to improve shared decision making about PCS. Methods. Searches were conducted of MEDLINE, PsychINFO, Social Services Abstracts, and Sociological Abstracts databases using the term ((prostate cancer screening) AND (beliefs OR attitudes)). Concept hierarchies were constructed from the key concepts identified in returned items using the Systems Model of Clinical Preventive Care as a theoretical framework (Walsh and McPhee: Health Ed Q 19(2):157-175, 1992). The hierarchies, concepts, terms, and references were entered into Protégé-2000, an ontology software package. Concept hierarchies were graphically mapped using the Ontoviz and Jambalaya visualization tools. Results. A total of 207 unique items were returned by the search. Of these, 160 were deemed relevant to the domain of PCS-related attitudes and beliefs, yielding 77.3% precision for the search. Attitude concepts included ambivalence; antagonism; complacency; embarrassment; fear of PCS procedure, prostate cancer treatment or outcome; worry; fatalism; optimism; mistrust of government, physicians, or health system, any of which may be mediated by physician-patient cultural, racial, and communication discordance. Belief concepts included assumptions about low standard of care in African American community; etiologic myth; low personal health expectation; personal cancer susceptibility; beliefs held by practitioner (efficacy or accuracy of PCS); pain as the first symptom of prostate cancer; absence or presence of PCS as a community standard; sexuality threatened by PCS; and assumptions about peer reaction to having PCS. Conclusions. Until now, no single, computable knowledge formalism exists for describing beliefs and attitudes about PCS. This ontology provides a compact, evidence-based representation of complex PCS domain knowledge that can be used in four ways. First, the knowledge represented in the ontology can be mapped directly to the knowledge base in computerized intervention software. Second, it provides domain knowledge that can be shared between clinical researchers, systems engineers, and multimedia designers. Third, the representation provides a means for effective searching the knowledge base using description logic. Finally, the ontology informs the intervention design by identifying candidate tailoring elements, and providing a rationale for scripting and a structure for evaluating the intervention.