Due to recent technological advancements, healthcare organizations now have access to large volumes of clinical, financial and consumer information from which to identify patterns and trends.
As with other industries, healthcare is grappling with the best ways to decipher and leverage this “big data,” with the ultimate goals to enhance patient care and improve population health.
The sheer magnitude of available data is both a boon and a hurdle. When interpreting data, for example, more information is not always better-and sometimes it can be just more-unless an organization employs tools to fully characterize the data and discern what is noise and what is not. In fact, to truly optimize big data, organizations must create structural frameworks-otherwise known as ontologies-to organize the information.
What Exactly are Ontologies?
Fundamentally, an ontology is a framework that describes the important concepts that collectively define a topic. Moreover, it formally outlines what exists in a domain, pointing out what data is in scope and what data is out of scope with regards to a particular subject.
For instance, let’s say a hospital is analyzing data about lung cancer, looking across a number of disparate information sources. Although it would be helpful for the hospital to know if individuals with lung cancer are smokers and for how long they have been smoking, it is probably not important what brand of cigarettes they buy and whether they have switched brands in recent months. That information would be more helpful to a cigarette retailer. An ontology would define the data parameters, making sure that demographic, socioeconomic and clinical data are included in the hospital project’s domain while consumer purchasing information is not.
Similarly, an ontology can eliminate apparent non sequiturs. For instance, if a patient with lung cancer has also recently had bunion surgery, the ontology may exclude information about the foot operation from the framework, since that particular clinical data is not directly relevant to the lung cancer diagnosis.
Why Are Ontologies Necessary?
By its very nature, big data comes from a variety of sources, and rarely do these entities define data terms and relationships in the same way. An ontology applies a common semantic model to the data, establishing a lexicon or vocabulary for describing concepts, terms, definitions and relationships. This ensures that the data have the same meaning regardless of source, allowing for apples-to-apples comparisons.
After using an ontology to characterize data, organizations can effectively analyze the information and develop systems that use the data correctly-and in the same way every time. The organization can then identify potential patterns or associations, confident that any correlations are true and real.
Without ontologies, organizations have no reliable way of navigating big data and cannot be certain analyses are based on a complete and accurate information picture. In other words, without ontologies, it can be like taking shots in the dark.
Strategies for Getting Started
While building a detailed ontology takes considerable thought and planning, there are certain steps to getting started that will help lay the groundwork for a strong and comprehensive framework.
Craft a scope statement. The first step is devising a scope statement that gives a high-level view of the topic under study. This is similar to a thesis statement in a research paper or a topic sentence in a paragraph in that it gives a point of reference for the rest of the framework-all further refinements stem from this general idea. For example, a possible scope statement for a project analyzing lung cancer might be, “Information about all patients in a population currently being treated for lung cancer.” Having a general statement like this one gives a rough cut across the data, allowing organizations to quickly make the volume of information more manageable.
Further describe the major concepts. Once an organization has a scope statement in place, the next step is to parse it into subsections, establishing a detailed refinement progression. At this stage, it’s important to clearly delineate what is in scope and out of scope. For instance, are different geographic locations being considered? Socioeconomic groups? Genders? Patients who have had interventions versus those who have not? By deconstructing the various components, an organization can determine how different concepts are related and distinguish key data points from non-relevant information.
Give the framework a structure. After defining a scope statement and supporting concepts, they can be assembled in a structure, such as a lattice or a hierarchy. Although either model is appropriate, a lattice is probably preferable as it can show relationships and interactions without being hierarchical. In the exploratory stage of developing an ontology, an organization may not yet fully know the taxonomic structure and categorization schema of the domain that it is exploring; thus it may be useful to examine the domain under consideration using a lattice structure rather than a hierarchy, especially if it can label the relationships (edges) in the lattice. In a complex domain, such as lung cancer, there may be complex relationships between environmental factors, life style choices, drug interactions, and, perhaps, socio-economic and status (SES) determinants.
Apply the tool. Once developing an ontology, healthcare organizations should consistently apply it in order to ensure accurate and reliable data interpretation. As mentioned before, without this step, they will severely limit their ability to effectively manage big data.
Ontologies Can Aid Big Data Interpretation
By carefully fashioning ontologies to shape and refine data constructs, organizations can lasso big data and use it to improve patient care, streamline business processes and elevate the total patient experience.
If an organization is planning to leverage big data, it must make the commitment to building ontologies-otherwise there is risk of grossly misinterpreting the data and wasting critical resources on analyses that prove to be off base. In the end, establishing ontologies can increase the value of the massive amounts of information resources now available and provide opportunity to better understand and manage population health.
Erik Kuiler, PhD, is an informatics consultant at Systems Made Simple, a leading provider of IT systems and services to support critical architecture, data, and application challenges in the healthcare industry.