Pages

Wednesday, August 30, 2017

An Ongoing Problem in FHIR

How do you find a problem that was occurring during a particular time span?  This is relevant if you are doing a search for Conditions (problems) that are active within a particular time period for something like a quality measure or clinical decision support rule. As I've previously discussed here, temporal searching is subtle.

So, if you have a time period with start and endpoints, and you want to find those conditions which were happening in that time period.  There are only two rules you need to care about:

  1. You can rule out anything where the onset was after the end of the time period.
  2. You can rule out anything where abatement was before the time period started.
What's left?  In the following analysis, I'm ignoring "things that happen at the boundary points". For the sake of argument, we'll assume that time is infinitely divisible and that no two things occur at "exactly the same time".  Obviously we quantize time, and boundary conditions are inevitable.  But they aren't IMPORTANT to this discussion.
  1. Things that had an onset within the time period, or before the time period started.
    1. For those items that had an onset within the time period, clearly its in the time period!
    2. For those items that had an onset before the time period started, one of three things must have occurred:
      1. The problem abated before the time period started (which is ruled out by rule #2 above).
      2. The problem abated during the time period, in which case it clearly was occurring within the time period for some point in that period.
      3. The problem abated after the time period ended, in which case, the time period is wholly contained within the period in which the problem is active, and therefore was occurring during the time period.
  2. Things that abated within the time period, or after the time period ended.
    1. For those items with an abatement within the time period, the are clearly withing the time period.
    2. For those items abated after the time period ended, one of three things must have occurred.
      1. The problem onset was after the time period that ended, in which case it is ruled out by rule #1 above.
      2. The problem had an onset during the time period, in which case it clearly was occurring within the time period for some point in that period.
      3. The problem onset was before the time period started, in which case, the time period is wholly contained within the period in which the problem is active, and therefore was occurring during the time period.

So your FHIR query is Condition?onset=le$end&abatement=ge$start

Done.  Simple ... err yeah, I'm going to stand by that.

   Keith

P.S. Yeah, so easy I had to come back and reverse the le/ge above.  Duh.

No comments:

Post a Comment