The problems with data seem to be due either to the abstraction process involved in summarizing the essential information about a case or to lack of careful specification of findings. These are accentuated by the nature of discharge summaries, but are not unique to them. Summarizing an examination requires the physician to leave out the information that in his or her judgement is not pertinent, which is a very context sensitive process and may be biased (since the physician knows the diagnosis when the summary is generated). The diagnosis method requires knowing more of this information than is typically included in the summary, so the first problem is reconstituting the information that was summarized away.