The question "Is this data a document" is unfortunately a harmful question. It implies that the question can be answered, and that getting that answer will be beneficial. I think that it cannot be answered, and that a search for an answer as phrased will in fact be harmful.
The proper question is "what is the best form for my intended use of this data, and how do I get it into this form without destroying its utility for other purposes?" This does not mean that data can be in only one form. In fact, efforts to do that are destined to fail. Data will be in many forms for many purposes, and there will be extensive metadata usage to bridge these diferences. This metadata exists to allow the data formats to be tailored to particular purposes without destroying utility for other purposes.
I'll explain by examples from my experiences with environmental work. Weather data is captured by data loggers, such as stripcharts, and by observers in logbooks. The instant continuous readings are displayed and may be used in real time, but they are lost. Only the strip charts and logbooks survive. Of course there are also calibration records, perhaps in the same logbooks, perhaps in different logbooks. There are attendence records for observers. There are credit card records for travel. (Why are credit card records relevant? There is falsification of logbooks, often discovered by a suspicious data analyst checking observer attendence and travel records. It was a common practice to take unauthorized absences and fill the log books retrospectively to conceal this.)
Are all these documents? Yes, or maybe not.
There are also petabytes, exabytes, and whatever is bigger than that of sounding balloon data, radar data, microwave soundings, radio limbing records, satellite images, etc.
Are all these documents? Yes, or maybe not.
For ground water pollution you must add field notebooks, photographs, samples, cores, photomicrographs, lab notebooks, equipment calibration records, purchase records for supplies, standards tracability records, ......
Are all these documents? Yes, or maybe not.
The weather data must be gridded for analysis, as must geological data. Suddenly all those documents are abstracted into gridded data. Is this another document? You also must document the gridding process. Gridding will reveal questionable data (like falsified observer records). This must be examined. Perhaps the original logbooks must be examined. When done, you've got corrected data, audit reports, correction reports, traceability.
You now wonder about the sensitivity of the data analysis to those errors. So you've got eigenfunction analyses. While you're at it, you perform eigenvector analysis of the data. This can reveal both more erroneous data and give insights into the physical processes being observed. For operational work you must grid and analyze at regular intervals. Most of this error analysis becomes metadata that is incorporated into the routine gridding and analysis process as procedural changes.
Are all these documents? Yes, or maybe not.
A gridded temperature dataset can't have more than just a number, like 18, in a grid cell. That's all the computer models can deal with. If you make that data any more complex, even just by adding some ancillary pointers, the computer models will be unable to run fast enough to deliver useful answers when they are needed. So all of the relationships between the grid and the tens of thousands of source elements has to live in metadata. If you do not define that metadata separately, the computer models will not be able to run fast enough. If you do not capture that metadata, you lose the ability to analyze for errors or to make any process improvements or to make any retrospective analyses for other purposes. You need to maintain the gridded data, the metadata, the orginial records, intermediate derived records, audit records, and calibrations records. But these must be maintained in the appropriate forms and appropriate locations for their intended uses.
Are all these documents? Yes, or maybe not.
The operational gridding may need to be performed every 5 minutes, or hourly, or daily, etc. The processes are updated at a much lower rate. Errors will be corrected at a slower rate and incorporated into the operational process. Researchers may need to perform retrospective analysis to incorporate error correction into their research databases, but the original analyses must be preserved so that post-facto operational reviews can be based on the data that was available to the people at the time they made decisions.
The question of document is irrelevant and distracts from the real issue.
The proper question is "what is the best form for my intended use of this data, and how do I get it in this form without destroying its utility for other purposes?". Asking whether this is a document takes you down the wrong path, into the wrong questions, and the wrong arguments.
I've used examples from environmental work, but all of these issues apply equally well to medical records.
Comments