We still have a dramatic gap in the US between what we could do (given some time and space to collaborate, financial support, and endorsement by HITSP) and what we are doing. This is an example where we are so far behind we do need to recalculate our trajectory and start warning people that we will be living in the past for a while longer.
One of the "big" changes, currently scheduled for October, 2013 (originally was slated for 2011) is CMS moving away from the use of ICD9 (International Classification of Disease, 9th edition) for claims submission to ICD10.
For those who are not tracking the fast-paced world of coded classification systems, ICD9 was put into use in 1975 and is the system used to provide information to payers (CMS being the agency behind Medicare) which is used to justify a claim for healthcare services. It provides groups of codes (21669 as of my last count in the US ICD9-Clinical Modification, ICD9CM) to classify diseases into discrete, mutually exclusive pigeon holes. It was replaced by ICD10 some time ago to report the cause of death to public health agencies, but it is otherwise the most common code system found in outpatient clinics which convey information about a diagnosis.
It has some well known shortcomings which make its use in electronic health records, particularly problem lists (or health concern tracking as we are now calling it), unwise. For example, there is no way to say that someone has a headache in ICD9. There are codes for various types of headache, but they all exclude each other in their definitions. In particular the non-specific “visit code” for a head ache specifically excludes migraine head aches and tension head aches. Which is why using ICD9 to convey health information for patient care reasons (as opposed to billing/administrative/public health reporting use) gives me a head ache.
That aside, it is what everyone seems to be using since we have a ton of data from our billing systems which bears some semblance to what we actually diagnose people as having and, voilà, we have ourselves coded data! The problems are it has been shown to have a only a casual correlation with what people are really diagnosed (e.g. by their physician), and that the codes are chosen to help optimize reimbursement, and most systems have a fixed number of codes that can be captured at any given encounter so that the choice of what codes get put into the slots is more often than not motivated by optimizing billing return rather than reflecting what is important to know about a patient.
Back to my original musing about this super huge big deal of switching to ICD10 in four years. ICD10 is marginally better than ICD9. It allows some diagnosis to be in more than one category (well, not really, but it tries, and by “more than one” I mean two, but only if the second category is an infectious disease or cancer) so that a diagnosis of bacterial pneumonia can now be considered both a bacterial disease -and- (drum roll please) a disease of the lung. It also now uses a letter in the code, rather than all numbers. It does have more diagnosis codes, but is otherwise pretty much the same old.
ICD10 has been around since 1994, or four more years since Jim Cimino's seminal Desiderata for Controlled Medical Vocabularies in the Twenty-First Century (Methods of Information in Medicine 1998;37(4-5):394-403), which called for moving away from ICD-like vocabulary systems in favor of more expressive, more flexible and more logical terminologies (such as SNOMED-CT which was released in 2002 by merging SNOMED-RT with the UK NHS Clinical Terms aka the Read Codes. SNOMED traces its roots back to the Systematic Nomenclature of Pathology circa 1965, or back when we were using ICD8!). You can find much more information on SNOMED at the National of Library website (http://www.nlm.nih.gov/research/umls/Snomed/snomed_faq.html) and from the IHTSDO website at (http://www.ihtsdo.org/snomed-ct/).
So the effective result is that in 4 years we are moving from a 1970's era code system, to a 1990's era code system, which we have known for at least the last ten years to be inadequate for use in health record systems.
At some point we need to start making provisions to actually use technology that is more contemporary, such as ICD11, which is slated for release the year before the US makes the leap to ICD10. A more realistic option is to educate clinicians (they are not so stupid as some IT types make them out to be) about the options and the consequences of doing thing right, particularly when we want to start using data to support "pay for performance" where ICD9/10's shortcomings will be very evident or for personalized medicine where using blunt instruments like ICD9/10 will just work when we start talking about trying to customize medication (and other therapy) based upon knowledge about individual's molecular structure.
More later on the rationale, problems, and proposed use of SNOMED-CT for health care information in the United States, circa early 21st century.
Posts
[update]
ReplyDeleteI had a brief discussion with one of the principle designers of ICD-11. He confirmed that ICD-11 was going to be a bigger and better version of ICD-10, with many features similar to those found in clinical terminologies and would be much more useful for the intended use: classifying and reporting diseases, syndromes, and injuries. Ongoing discussions with IHTSDO suggested that there would be mapping between SNOMED-CT disorders and ICD (as there is now), and that the two would continue to play separate but complimentary roles in health informatics.