Updated: Jul 5
Needless to say, a line like that doesn’t go without a request for clarification.
But the answer is all too familiar for the data pro trying to normalize data, especially the taxonomies they fall under. The challenge he experiences repeatedly is the matter of every clinician, surgeon, and physician having their own “grandiloquent” name for a procedure they “invented” and implemented daily or sometimes hourly.
The data pro feels strongly that this profession is especially prone to narcissism-driven taxonomies because many doctors like to think that what they have done has a particular uniqueness instead of a more general descriptor. So something as mainstream as a minimally invasive total knee replacement is not something that a groundbreaking knee surgeon would want in their medical records. What would Shakespeare say?
Is there room for ego in health data classification?
The International Classification of Diseases, Tenth Edition (ICD-10), a clinical cataloging system, went into effect for the U.S. healthcare industry on Oct. 1, 2015. It is arguably the bible of every ailment known to (American) humankind. The catalog currently has close to 130,000 procedure and diagnosis classification codes.
In theory, every healthcare procedure should correspond to a specific ICD-10 code entered into the Electronic Health Record platform. Pretty simple, right? Not necessarily.
Because of the Shakespearean nature of many physicians and clinical workers, the current codes might not naturally fit that surgeon’s “branded” procedure. The result can create two problems. First, dissatisfaction on the medical side of being forced to put a more mundane classification into the system. Or second, having the database director add new taxonomies more indicative of the Shakespearean branding for that procedure.
I use healthcare as an example because of the number of documented procedures in ICD-10, but I can assure you the same phenomena occur in other industries. For instance, while not life or death, I know from personal experience that the classification of uses of marketing technologies can be at the mercy of a search engine optimization innovator who has cracked a code and wants no part of having his new process known simply as SEO or SEM.
In an age of data analytics, machine learning, and artificial intelligence, both scenarios above affect insights –– regardless of how they are normalized into the system. If a more innovative procedure is forced into a more mundane classification, the procedure becomes analytically “sanitized.” If, on the other hand, exceptions are made for even minor innovations, then the algorithm must be able to determine when to integrate the exception into the standard classification for broader analysis and insight.
Any of us who have been on a taxonomy and normalization team know that this is hard work –– even in industries where normalization might be, unlike healthcare, across very simple data universes. There is a constant tug of war between those wanting reliable granular insights and those needing to manage the data creek of billions of instances with only subtle differentiators.
Normalizing addresses, company names, and job titles have become relatively simple. But those enterprises expecting their data to drive artificial intelligence or predictive analytics must step up their game and expect the Shakespearean tendencies of their stakeholder to be a factor!
For more insights on Leadership, Patient Experience, Hospital@Home, Burnout, and Equity, log into the ICD Healthcare Network.