Questions needs to be presented to the user in a manner that is understandable, efficient, and predictable. This criterion often requires presentation of questions in a manner different that Analysis Normal Form. Analysis Normal Form tries to be optimal for the purposes of data analysis for decision support, and other purposes. It does not try to be optimal for the purpose of data entry.
To have our cake and eat it too, we have to provide for a standard transformation capability that converts answered questions to be transformed into Analysis Normal Form. The following transformation patterns represent a set of necessary transformations identified by this project.
The patterns will assume that the answered questions are in the form of an HL7 FHIR Questionnaire Response. These responses will be converted to analysis normal form by the transformation.
IF-to-ANF Transformation
The Input Form to Analysis Normal Form transformation simply takes the responses elements, and copies them into the proper location in the FHIR Observation resource, with no additional transformation. Serum Sodium may be an example.
This transformation also serves as the foundation for more complex transformations. Multiple transformations can be applied in any order.
Existential-to-ANF Transformation
When the possible responses to a question are present, absent, or unknown, but the code is formed in the same manner that an ANF topic would be formed.
Reason-for-Unknown Transformation
Similar to the Existential-to-ANF transform, but the possible responses include reason for unknown, such as not examined for, or similar.
Extract-SOI-to-ANF Transformation
The code field has an embedded subject of information that is other than the patient. The transformation will extract the subject of information from the code, encode the topic in a manner consistent with the subject of information pattern (Family history: Maternal hip fracture; Age when the first heart attack occurred Father).
Value-to-Attribute Transformation
The value field has an attribute it it (like “blue” in response to “eye color” question), and will encode the attribute as part of the topic (“blue eye color”) which can then be presented as “present”.
LOINC-Expansion Transformation
When a LOINC code contains information that should be in the ANF context, this transformation is used.
Reason Transformation
When a code contains the reason something was done, or is requested, the reason is extracted from the topic, and placed in the proper place in ANF circumstance (I limped because of pain in past 7 days [PROMIS], T1c: Tumor identified by needle biopsy because of elevated prostate specific antigen: PROSTATE: Biopsy/transurethral resection: PROSTATE: Resection).
Onset Transformation
When a code contains an embedded time of onset, this information is extracted into the ANF circumstance (History of angina in last year).
Frequency Transformation
When a code contains an embedded frequency, this information is extracted into the ANF circumstance (Aerobic exercise three or more times per week).
Inverse Form Transformation
In some cases, findings may be inverses of each other (Diabetic-uncooperative patient; Diabetic - cooperative patient), and stating uncooperative patient absent is the same as saying cooperative patient present. Transforming them to one form normalizes representation and retrieval.
Laterality Transformation
Unilateral, bilateral, left sided, and right sided need to be normalized.
Add Comment