As indicated above, two main groups of cross-references are provided between problems. These are the conventional broader/narrower group and a group of "functional" cross-references. The process of indicating the initial relationships between problems is in effect an extension, if not an integral part, of the naming process. The hierarchical relationships, even if only first approximations, position the problem with respect to others and confirm the distinction from them. Particular attention was therefore given to the relationships between problems.
A problem may have any number of cross-references, but the maximum number of any one type seldom exceeds 20. The 13,167 problems in the database currently have some 80,394 cross-references between them.
1. Hierarchical relationships
In the case of the broader/narrower group, there are three well-established types: "broader" problems, "narrower" problems, and "related" problems. These have the usual meanings, with the related category being used as a catch-all in those exceptional cases when the relationship cannot be more appropriately expressed through any of the other cross-reference types. In contrast to conventional use however, a problem may have several broader problems.
Various series of problems have necessitated some regrouping of problems into problem groups to avoid inclusion of too many problems at too great a level of detail. Different methods of handling this matter and establishing cut-off points were used on an experimental basis. Constructing such problem hierarchies was considerably facilitated when the available documents had attempted some such categorization of the problems.
The number of problems emerging at the top of hierarchies for problems in Sections PB through PF was 308 (PB=45, PC=33; PD=5; PE=3; PF=222). Those at the bottom of hierarchies for that group numbered 4,620. Only 180 were not positioned in a hierarchy, and all but 3 of these were in PF.
2. Functional relationships
In the four types of functional cross-reference, the described problem: aggravates (a cited problem), is aggravated by (a cited problem), alleviates (a cited problem), or is alleviated by (a cited problem). Clearly this group of relationships forms two complementary pairs. In certain cases a problem may both aggravate and be aggravated by the same cited problem.
Relationships between problems, other than hierarchical ones, are included either where they were specifically mentioned in the available documents or where they could be reasonably inferred from such material. It is rare for documents to be systematic in their description of the relationships between problems. Relationship networks have to be built up from several different sources. Often it is not clear whether the relationship applies for the whole of a problem hierarchy or for only some component part. Some effort has been made to "tidy up" such networks, but in general the practice adopted is to include relationships at this time in order that the networks could be more thoroughly criticized with a view to improvement. It is generally easier to criticize errors of commission than to undertake the extra effort to remedy errors of omission.
In the problems from PB through PF, 770 were positioned at one end of a functional chain only, either as aggravating others (but not being aggravated themselves) or as being aggravated by other problems (but not aggravating others themselves).
3. Tentative relationships
It must be strongly emphasized that no cross-reference can be considered "permanent". Cross-references are treated more like pointers. During the editorial process pointers may be modified into a more appropriate configuration. Typically a pointer from Problem A to Problem B may be replaced by one from Problem A to Problem C, plus another from Problem C to Problem B -- if Problem C appears to be an appropriate intermediary. Some pointers may be more obvious and permanent than others that are tentative or only approximate.
Clearly all the different forms of cross-reference interweave to form a very complex network. When indicating functional relationships between problems, the information available may not be sufficiently unambiguous as to whether the pointer should be made to a broader problem or to a narrower problem. Or the information may only mention the relationship to the narrower problem, when the context suggests that it could be more appropriately made to the broader. In this sense whatever indication is given can only be considered tentative, subject to modification later in the editorial process for a future edition.
4. Computer checking
The editorial process of checking relationships is both extremely time consuming and subject to error. Much use is therefore made of software routines both to ensure reciprocity of relationships and to check for various forms of redundancy. Examples of such redundant patterns of linkages include:
- � both Problem A and a narrower problem of Problem A indicated as aggravated by Problem B.
� more subject to query, several narrower problems of Problem A aggravated by Problem B, which could possibly be replaced by Problem A aggravated by Problem B.
As part of the editorial process, it is useful to ask such questions as:
- � which problems are not part of another problem ?
� which problems are not aggravated by other problems ?
� which problems do not aggravate other problems ?
� which problems do not alleviate some other problem ?
From an editorial point of view, it is clearly important to focus on the first question, if only as a means of detecting duplicate narrower problems. This question is also important in the case of highly specific problems (such as some rare disease), which it may not be useful to represent in the system at all, but for which some broader problem can be usefully included, even if no information on it exists (the class of such rare diseases, for example).
In the final published product the distinction between problem perceptions substantiated by information received and those based on interpolations of this kind is quite evident. The interpolated problems are present in skeletal form, with name(s) and cross-references, but without any other descriptive text.