Change from prior
How a finding compares with the previous exam. Authoring guidance makes it the required second attribute of a finding model, after presence, with values “unchanged, stable, new, resolved, plus clinically appropriate direction-of-change pairs (larger/smaller for masses, worsened/improved for diseases, increased/decreased for quantities).”1
The attribute exists because a finding is not a snapshot. “Findings exist in time. A radiology report is a snapshot, but the findings it describes persist across exams. A pleural effusion seen today may be the same one from last week, now larger.”1 Change from prior is what carries that comparison at the level of a single observation, and the documented FHIR mapping singles it out along with presence as the components an Observation should carry.2
Change from prior is also what distinguishes an attribute value from a finding. “Stable cardiac silhouette” is not a finding, because “stable” is a change-from-prior value on a cardiac finding.1
Synonyms and near-synonyms
- Interval change and temporal change are clinical equivalents.
change_from_prioris the key name in the extraction schema, one of the standard attribute keys alongside size, acuity, severity, count, and morphology.3- Temporal status on an Imaging Problem List is different: it is computed across all observations of a finding rather than asserted on one.
- Tracking identity is a further step again, linking the same physical entity across reports, which OIDM does not currently do.
Identifier form
An ordinary OIFMA_[A-Z]{3,4}_[0-9]{6} attribute identifier with dot-suffixed value codes.
Where it is used
Finding model format, Observation, and Imaging Problem List.
Conflicts
The attribute is required by guidance but missing from part of the corpus. The open definition cleanup plan lists adding missing standard attributes, presence and change from prior, to 124 existing models as an ordered step, so consuming code cannot assume the attribute is present.4 Value sets also vary by design, since direction-of-change pairs are chosen per finding, so no single enumeration covers the corpus.