FWIW, I rather like the interim solution proposed by PBR, where there is an IS-A relationship to the caused disorder and a DUE TO relationship to the "causing" disorder.

For the future, an alternative to representing concurrency with two IS-A relationships would be to use the temporal relationship attributes in the ASSOCIATED WITH role hierarchy. For example, we could consider re-structuring that hierarchy along the following lines:

ASSOCIATED WITH

    CAUSAL RELATIONSHIP

        CAUSATIVE AGENT

        DUE TO

    TEMPORAL RELATIONSHIP

        BEFORE

        DURING

        DURING AND AFTER

        AFTER

This approach would seem to have the advantage of providing a "clean separation" between causal relationships and temporal relationships:

• If you know about a causal relationship but know nothing about a temporal relationship, then only use one of the causal relationship attributes

• If you know about a temporal relationship but know nothing about a causal relationship, then only use one of the temporal relationship attributes

• If you know about both a causal relationship and a temporal relationship, then use both a causal relationship attribute and a temporal relationship attribute

To query for "concurrent disorders," you would search for values of the DURING and DURING AND AFTER attributes. Presumably, you would not have to worry about causal relationships, since the "clean separation" means that a causal relationship attribute implies nothing about time. If you want to search for "concurrent and due to" disorders, you would search for values of DURING and DURING AND AFTR (for the "concurrent" part) and DUE TO (for the "due to" part).

The TEMPORAL RELATIONSHIP role sub-hierarchy might need to be "tweaked" to better represent Allen's intervals, but that's another discussion.

In any event, the approach described above seems easier to understand and explain, IMHO.