There has been a long held view, even mantra, amongst health informaticians that data should be “structured” and “coded” if it is to be harnessed for clinical decision support, adverse event prevention, and for research purposes. One trouble is, the concept of “coded” data means different things to different people.
Australia was at the international forefront of “coding” clinical data when it established the National Health Data Dictionary, first published as the National Minimum Data Set – Institutional Health Care in September 1989. Back then, most of the data collected according to the data dictionary was probably collected manually, with paper forms ( and perhaps that is still true today?). The reason codes were used was to ease the burden of data entry. The purpose of collecting the coded data was not clinical. It was to meet the needs of statistical aggregation for reporting.
Somehow, as electronic collection of data for clinical use became more prevalent, and the notion of e-health that embraces the sharing of clinical data gained acceptance, many of the people and practices associated with “coding” for secondary use found their way into the messaging specifications of HL7 and beyond. There has been some implicit assumption that what is good for the goose is good for the gander. E-health has a long way to catch up to the well-oiled reporting world, but hopefully one day it will catch up.
But if clinical data is to be captured and reused safely, efficiently and effectively for patient care, then the “coding” practices of last century need to be avoided for e-health. The National Health Data Dictionary may serve statisticians well. The codesets that are embodied therein will not serve the clinical community well. Those interested in part of my justification for this stance are welcome to view this presentation on the subject.