Physicians, like all professionals, have developed their own language over time. Patients frequently complain about their overuse of “medicalese,” and suspiciously complain that we use “big Latin phrases” to justify high fees. I certainly can’t argue that the diagnosis of Hallux Rigidus sounds much less serious when reduced to its English translation of “stiff toe.” None-the-less, medical terminology is essential for exact and effective communication between healthcare professionals. Unfortunately, the translation of this knowledge into computer-understandable vocabulary has been dismal at best. Don’t get me wrong, when it comes getting paid, we have done a masterful job—billing codes and disease classification for claims transmission seem to work just fine. Semantics, or the meaning of terms, is where we have difficulty in the clinical world. As I will explain, it is common for similar terms to have vastly different meanings in other contexts. Clearly, there is enough confusion in medicine today.
Last week, I saw a patient of mine with a persistent mild and non-productive cough that is slowly improving. He contracted a viral illness while onboard a cruise 5 weeks ago. I obtained a chest x-ray which was read by the radiologist as “clear except for possible chronic fibrosis or discoid atelectasis at the right lung base.” This got me thinking about clinical semantics and ontologies. My first dilemma was how to explain to my patient that his x-ray was basically normal, and that the radiologist’s comments were more malpractice-driven rather than clinically helpful. There is, however, a more concerning problem here: how would informatics interpret these results?
For instance, there exists an ICD-9 code for the fibrosis reported above: 515 – Postinflammatory pulmonary fibrosis, but this also includes things like cirrhosis of the lung chronic or unspecified. The same for the atelectasis: 518 – which essentially translates atelectasis to pulmonary collapse. Clearly, ICD doesn’t really represent my patient at all. In fact, it is hard to conceive of a controlled medical vocabulary that represents “schmutz on the x-ray that probably has no clinical significance but the radiologist didn’t want to say ‘normal’ because he doesn’t want to get sued.”
If you think about it, the basic problem here is how the receiver of information imperfectly uses language to reconstruct the patient at presentation. After all, reality is in the eye of the beholder.
Speaking of “eye,” the perception of this term varies markedly depending on whether you are a potato farmer, a hurricane expert, or an ophthalmologist. This is known as semantic heterogeneity and human beings are capable of understanding context and nuance, where computers are not. For instance, a logic based computer model might recognize that the fundus in part of the stomach, the stomach is part of the GI system, therefore the fundus is part of the GI system. However, if a clinician was referring to the fundus of the female uterus, the inferential logic as being part of the GI system is silly.
Here is the role of an ontology—a construct that can delineate the semantics of a particular terminology system in an unambiguous manner and lead to effective communication. This communication is not just human-to-human, but also human-to-computer-to computer-to human.
