Standards for the electronic transfer of clinical data: progress and promises.
Data exchange standards have two components: the message format or syntax and the dictionary of codes (semantics). For many applications, message standards already have been developed. For a few kinds of clinical entities, such as drugs, these code systems (e.g., the National Drug Code) are virtually complete, but a few gaps must be filled and an agreement must be reached about the level of granularity needed. The available codes for clinical descriptors are inadequate but the National Library of Medicine's Universal Medical Language (UML) project will do much to redress this deficiency. Codes for clinical variables such as blood pressure and blood glucose which have methods, units, normal ranges, and physiologic correlates are very inadequate. CPT4 provides some of the needed codes but has huge gaps. An early effort to extend CPT4 is included in ASTM 1238. Work being done by ASTM E31.12 and the Euclides project will offer robust codes for clinical laboratory measurements. If we want to pool data from different institutions for clinical and policy research, universal codes for observations are prerequisite. And agreement of an international coding system for observation-bearing variables should be a major agenda item for standards groups in the next year. Our goal has been to standardize the communication of clinical data between clinical systems, not the systems themselves or their internal operation. In fact, standardizing the internals of clinical application could be counterproductive at the present. It would deflect energy from, and delay the spread of, CDI standards. Moreover, it gives undue attention to computer systems, rather than the data they contain. The data are the most expensive part of any data system. They are the raison d'être for such systems. Computer systems come and go. The data last forever. Yet we have been mesmerized by the computer system while ignoring its contents. As a result, most computer-stored clinical data must live like the tragic boy in the bubble. They cannot "live" outside of the computer system in which they were born. So, we find at every hospital the bizarre rituals of humans reading computer-generated reports so they can type this information in another computer. Electronic (e.g., stored clinical) data should not depend upon the internals of a particular program, language, or machine for its interpretation. The clinical data entered into one computer system should be directly available to any other computer system that now receives them through manual transcription. Data interchange standards give life to our data--independent of the source system.