Historical Analytics - Transitive closure - last known position
Approach
Create an enhanced transitive closure table containing inactive concepts at their last known position (but not moving children?)
Pros / Cons
Pro | Con |
|---|---|
Possibly the most fully featured | Does not account for the situation where a concept has been retired because it is the wrong place in the hierarchy ie the historical data will continue to feature the 'wrong' parents. We could discuss that the concept could appear at more than one place in the hierarchy, both in its original position and as a sibling of its replacement in the new location. |
Most tractable solution | Does not obviously provide support for a 'confidence' level. Any system using these would probably want to flag up concepts or records that have been included via historical associations and allow the end user to see the logic behind the inclusion. |
Useful where a "good" historical association is not available. Could be used in combination with other approaches to ameliorate this lack of replacement. | Does not make use of improvements in SNOMED CT over time. |
|
|
|
|
Target Use Cases
Better aligned to answer the questions where the answer for that question is needed to make sense for a particular point in time eg forensic checks.
Points to Note
Could also be used to re-populate previously inactive relationships other than IS A.
In the case of restoring a concept to its last known state, only the "most recently" inactivated relationships would be restored (which may include attributes that themselves need to be reactivated!)
The parents of an inactivated concept may have moved around in the meantime, leading to "last known position" not having the same transitive closure despite best intentions - although this result is expected to be "better"
Some integrity check could be done to test the last known position with the position of the historical association suggested replacement to highlight movement.