Measuring the Information Gain of Diagnosis vs. Diagnosis Category Coding.

Coding categories of diseases, injuries, symptoms, findings, etc. with ICD-9-CM necessarily imparts a loss of information vs. coding such entities with a terminology or ontology-a consequence of the nature of classifications. However, to our knowledge, no one has attempted to quantify this information loss or conversely, the information to be gained by coding entities as opposed to categories. We estimated a lower bound on information gain of coding with SNOMED CT instead of ICD-9-CM, as measured by Shannon's information entropy. We found that the nation could gain more than 97 megabytes of information per year by coding diagnoses vs. diagnosis categories, an increase of 10%. This increase is more than that obtained from coding ICD-9-CM at the 5(th) instead of the 3(rd) digit level. We recommend that ICD-9-CM be removed from electronic medical record (EMR) stage 2 and later meaningful use criteria.

The tyranny of the diagnosis code.

The electronic medical record (EMR), while having acknowledged advantages over the paper record and powerful constituencies advocating its adoption, is not in widespread use. One significant obstacle to its acceptance by physicians has not been addressed--its failure to provide easy input for the patients exact diagnoses and for the retrieval of those diagnoses during subsequent patient care. Furthermore, our system designers have failed to respond to the expansion of the use of the medical record from its origin as simply the physician's memory and communication tool to becoming the building block for our Medical Record Health Information System (MRHIS), where it also supplies the justification for payment for care and is the source of fundamental statistics about health and healthcare. These problems reflect a basic flaw in the application of available information technology to EMR design and data management: We use output codes--the category codes from ICD-9-CM--for input of diagnoses. This fact imposes the tyranny. Our medical record must have these ICD-9-CM codes for the reimbursement system. But, to be accepted as the basic record for medical care, and at the same time, to be truly useful for case retrieval and statistics, medical informatics experts recognize that our EMR must have codes for the exact diagnoses of the patient (diagnosis entities). But no practical method for their input and management has been offered This paper proposes a way to provide easy input of diagnosis entities, and their permanent coding as a workable solution to the problem.