Existing Wording: (Entire page contents)

Proposed Wording: To provide this the resolution team would have to first consider the idea of developing a Profile that contained extensions to the Observation that would permit the use of Observation sets to delineate between the major types of observations performed by genetic testing labs during the multi-step testing processes that they provide. See comments for further explanation.

Comment:

There are testing centers that run genotyping or sequencing tests to simply identify the presence or absence of a given genotype in specific patient tissue (germline, tumor , prenatal) as well as in viral or bacterial organisms found within a patient. There are also labs that perform interpretive services on the findings produced by wet lab and bioinformatic processes. The interpretive services are not always included but can be performed as a secondary service at the lab identifying the genotypes or by a secondary lab that specializes in one or more clinical domain areas or even by the receiving clinician/provider that ordered the test (the treating physician). So, it seems that we would want to define LOINC panels that group logical sets of observations together for the separate types of observations, then use an approach similar to how the Example shown for DiagnosticReport "Lipid Report" using LOINC panels to profile the constraints around providing a set of Observations that comprise a logical observations that are included with every Lipid Panel. This concept and proper use of Profiling should be understood by the authors of the Genetics group before proceeding with an approach that simply adds a dozen or more attributes (as well intentioned as they may be) to a single Observation Resource. I strongly recommend a review of these proper uses of Profile extensions using LOINC codes and panels, as that has historically been an important aspect of Laboratory results reporting. With the current model there is no standard way to associate LOINC observation codes to these new set of geneticsXXX attributes. Hardcoding or mapping is unreasonable.

I suggest the following presentation for a quick helpful picture of how LOINC panels are used to constrain a Profile http://www.slideshare.net/ewoutkramer/authoring-fhir-profiles-extended-version starting at slide 41. In actuality, the entire presentation is useful for understanding how and what is needed to be considered when developing a Profile. Summary: 1) Demonstrate pragmatic examples of how profiles can be developed in FHIR, particularly around the use of LOINC codes and panels to develop Observation sets. 2) Ask the authors to identify the major workflow/service patterns performed by genetic testing labs (not super detailed). I believe they will see that there is a logical and useful separation of wet lab, variant identification/calling services from the clinical analytics/interpretation services. 3) develop the sets of observed values for the components in each phase. For example, GenotypeObservation in the first phase with a presence or absence interpretation. And, possibly, GeneticInterpretationObservation with the targeted set of phenotypes (germline, cancer, condition, pharmacogentic, etc) and the clinical significance as asserted by an expert geneticist. 4) With a logical grouping of the observation sets, I believe it will provide the health care community the opportunity to build up representations of complex genotypes and alleles, when needed in pharmacogentic reporting, as well as structure variants, somatic interpretation, and other state specific use cases yet to be standardized.