To be applied for as many cases possible, SD0005 looks for uniqueness of --SEQ value within a combination of USUBJID, POOLID and SPDEVID variables value, if any of these 3 variables is present.
USUBJID and POOLID are mutually exclusive variables. So, the check works correctly for them.
SPDEVID may be utilized in addition to USUBJID. This case is not covered.
A bug will be fix in the next release of the validator.
According to the SDTM IG for Medical Devises V1.0, DISEQ should be unique within a parameter within a device (SPDEVID) within dataset for DI domain, but for other domains, --SEQ should be unique within every subject/device combination within dataset.
Why does SD0005 always fire when SPDEVID is not unique wuithin a device, but unique within every subject/device combination?