Stream: genomics / eMerge Pilot
Topic: FHIR representation of a genetics test with multiple test...
Mullai Murugan (Mar 06 2019 at 00:25):
As detailed in a gforge tracking item #19828, we have multiple panels as part of our test results. As each panel will have its own positive or negative result and corresponding information, having one Diagnostic Result, Genomic Analysis Overall Interpretation and related Observations for the entire test (as opposed to a separate ones for each panel in the test) might not be sufficient to represent the results.
Larry Babb and I are working together on this; we would like to put this topic up for discussion with the group and are happy to walk through a couple of possible options we have put together in discussion with Kevin Power.
Bret H (Mar 06 2019 at 15:24):
please list the options you have already considered. Sounds like a good place to start
Kevin Power (Mar 06 2019 at 18:21):
I think the options were what we have discussed in CG (let's say they have 3 panels)
A) 1 DiagnosticReport -> 3 Genomic Panels -> Panel groups Overall Intep, Findings, and Impacts
-- note, the Panel can refer to 0..* Overall Interp, but we have DiagnosticReport.results -> Overall Interp as 0..1, so we should change that constraint if we go this path.
B) 1 "Summary" DiagnosticReport -> 3 DiagnosticReports (one per panel) -> DiagnosticReport.results group OverallInterp, Findings, Impacts
@Mullai Murugan - Did I summarize correctly? Have you come up with other options?
It might be that both A and B have a place. The CG group is working with O&O, and we need to define guidance on these approaches.
Mullai Murugan (Mar 06 2019 at 23:56):
@Kevin Power yes, that summarizes the options we had considered. Though option B might possibly be more apt for our use cases, we are working on creating our draft spec with option A on the assumption that it might be more in alignment with what the CG currently includes in the draft spec. @Kevin Power please correct me if this is not accurate.
Andrea Pitkus, PhD, MLS(ASCP)CM, CSM (Mar 07 2019 at 04:42):
Just added thread here on panel orders which looks like it might apply here as another option: stream:genomics topic:panel+orders
Have you considered each panel as an order (perhaps a sub order as part of larger super panel/profile)? If the performing lab test is say Genomics Profile X which is comprised of panels A, B and C, each with their own results/observations, why wouldn't this work? From the reporting side, the Diagnostic Report would be comprised of the Profile structure with panel A as one report section with it's individual result components/observations, followed by Panel B in the next section, and Panel C.
Each Panel can be reported separately as it's own Diagnostic Report if the order is structured that way too.
Kevin Power (Mar 07 2019 at 14:18):
To date, we haven't spent much time on the ServiceRequest resource and how it would apply to the genomics space.
To your example @Andrea Pitkus, PhD, MLS(ASCP)CM, CSM , how are you suggesting to use ServiceRequest to model your example Genomics Profile X with Panels A, B, and C?
Andrea Pitkus, PhD, MLS(ASCP)CM, CSM (Mar 07 2019 at 15:35):
https://www.hl7.org/fhir/valueset-request-intent.html
Let me know if there are specific questions from the Service request slide examples on the genomics orders thread, as it gets complex really quickly. Described with text below, but the slides really pull it all together visually.
Service Request with intent of original order would be used for the top tier/original order/physician convenience panel order 4 grandparent order or regular order for 1, 2, 3 below: (bolded top/original orders)
1. a single order that is a single result (i.e. BRCA1 order/BRCA1 result/positive test result value)
2. A panel order BRCA panel order/ BRCA 1 result / BRCA 2 result
3. A reflexive order that is say, BRCA reflexive panel order (order 1, see 2 for results)/ BRCA 1 gene analysis (order 2) or BRCA 2 gene analysis (order 3) (only performed if BRCA 1 or BRCA 2 screening panel are positive. If either, both neg, then not performed)
4. A Profile order that is say, BRCA genetic profile order (grandparent tier)/ BRCA panel order (parent 1 tier) + BRCA sequencing (parent 2 tier)/ BRCA 1 result (parent 1 result child 1) + BRCA 2 result (parent 1 result child 2) + BRCA sequencing results (parent 2 result child 1) + BRCA sequencing interpretation (parent 2 result child 2)
Service Request order instance order would be used to reflect in the Profile each "2nd tier" or additional order such in cases 3, 4 as the original medical/legal request from the provider is still valid with the original order to be used for performing the remainder of the orders.
If any of the orders are panels say for a list of genes, markers, drugs (for PGX), performed then a panel order would be comprised of:
Panel X order
result A observation
result B observation
result C observation
Panel Y order
result D observation
result E observation
result F observation
In the Diagnostic Report Resource would expect to see the "groupings" /report sections as discussed.
For example:
Grandparent profile order is at the top as the name of the test ordered/reported
Panel X order, now is listed as a report section/grouping header/observation (of observations) level
result A represented as observation (A)
result B represented as observation (B)
result C represented as observation (C)
Note each observation value may be short answer/nom(i.e. ISCN, HGVS or other genetic notation); narrative/text blob such as an interpretation/Impact/clinical significance as modeled with the CG Implementation Guide depending on the section/type of genetic/molecular results reported.
It shows how DR results originate and are connected to the order/Service Request. The other aspect not mentioned/listed is the specimen info upon which all the results were generated. It is also needed. Where labs have testing on tissue or swab specimens such as with many genetics, microbiology and pathology results, they will have an Ask at Order Entry (AOE) question requesting the ordering provider to indicate the source of the specimen or when it has been collected from (i.e breast tissue, swab from inner cheek). It should be indicated in specimen resource as to how specimen was collected (in appropriate containers, additives, time in fixative, etc.) too. AOEs are often reported as individual result components such as the observations that would be reported as part of the Diagnostic Report.
Kevin Power (Mar 07 2019 at 16:15):
Thanks Andrea, interesting summary. Have you seen the example report from @Mullai Murugan ? It can be seen here
I did review the slides you shared in the other thread. Slides 11-15 don't mention ServiceRequest at all - it seems totally focused on the catalog side. But is the implication that each instance of CatalogEntry would get a matching ServiceRequest?
Andrea Pitkus, PhD, MLS(ASCP)CM, CSM (Mar 07 2019 at 20:52):
Hi @Kevin Power
No hadn't seen the report. Reviewing. Couldn't find the order in https://www.bcm.edu/research/medical-genetics-labs/tests.cfm
Does look like this order is reflexive with the various report sections added as needed clinically. They are each individually built most likely in the LIS so aspects like quality control can be documented for each as performed as well as dropping billing for additional work needed for more complex cases.
Each report section can be added accordingly depending on what is performed, similar to structured pathology cases (will likely see examples when the CIC folks join Wed). Overall interpretation contains interpreted text blob. Although may wish at some point to add sub sub sections for management recommendations, family screening, etc. to get to more discrete data capture/encoding. References section is like similar labs and also includes the vendor info. Other labs also include the required performance requirements, noted here in the footnote, as another report section, which I'd recommend so it's structured the same way across labs.
The Activity Definitions, create Service Requests when using "$apply" operation, and it works great (according to another thread). We're still working through SpecimenDefinition needs. Results are from ObservationDefinition. All in FHIR R4 resources.
In Theory all genetics, cytogenetics, molecular diagnostics, FISH, etc orders, including those not orderable by the provider, but those that reflex in the lab such as some of Larry Babb's comments indicate, would be model-able. Would be great to get genetics examples into Service Catalog to test at upcoming connectathon so folks can see it in action.
Kevin Power (Mar 07 2019 at 22:13):
Thanks @Andrea Pitkus, PhD, MLS(ASCP)CM, CSM -- will let @Mullai Murugan and/or @Larry Babb comment on your thoughts.
Larry Babb (Mar 08 2019 at 01:39):
@Andrea Pitkus, PhD, MLS(ASCP)CM, CSM I'm not sure I'm following the suggested approach being proposed here. I would like to be able to discuss it with you directly so you can help me understand it and I can make sure I'm clarifying the concerns we are trying to address. Would you be willing to meet in the next week?
Bret H (Mar 08 2019 at 14:05):
@Mullai Murugan does this seem correct to you and your intent:
With A, it seems to say these all were done at the same time. I could see a system using the 3 panel impacts as the thing to put in front of the clinician.
With B, it seems to say here is the overall summation of all these tests done at the same time. I could see a system hiding the 3 'child' diagnostic reports and exposing the 'Summary' as the thing to put in front of the clinician.
There are a lot of details in genetics lab report that should be available at the clinician level - but not necessarily at the top level. The system needs all the information.
I'll be a downstream system trying to infer the importance of the information in the panel. hope it gets organized well.
Larry Babb (Mar 08 2019 at 14:15):
@Andrea Pitkus, PhD, MLS(ASCP)CM, CSM I found the powerpoint presentation you referenced and now a bit more caught up in the references you made to the Catalog work. I am impressed with it and can see that it is helpful in organizing the services, activities and observations from a definitional standpoint. I now need to understand how this can help with the reporting of results.
Andrea Pitkus, PhD, MLS(ASCP)CM, CSM (Mar 08 2019 at 19:43):
Should clarify the slides have been developed by Francois Macaray and Quest Diagnostics from connectathon and call feedback. Seems like testing in surgical pathology, microbiology cultures, Type and screen (when elutions, other testing is needed), and CG follow more of a reflexive order structure/flow, whereby based on the result of initial results, other orders may be added on. For example if screening tests are normal/negative, then adding on WGS may not be needed. In other cases, iWGS may be needed.
Kevin Power (Mar 08 2019 at 22:11):
I think, in the eMERGE case, the 3 "sections" of their report (cardiac gene panel, polyeneic risk score, PGx) are all delivered every time - so there is no 'reflex' in this case. Maybe I took you too literally?
Andrea Pitkus, PhD, MLS(ASCP)CM, CSM (Mar 09 2019 at 00:12):
Would be helpful to confirm as it does indicate the PgX variants were addended to the report. Wonder if that is due to a timing issue (the rest of the items were reported first and then the addendum was issued when PgX results were available?) Would also be helpful to see the order/test compendium info for the test.
Kevin Power (Mar 09 2019 at 02:37):
good catch on the addendum part - @Mullai Murugan should confirm, but I am about 95% sure she said they always do all 3 each time as a single order.
Larry Babb (Mar 12 2019 at 17:50):
Yes these tests are defined to include the delivery of all 3 sections. The official emerge test does not currently include the Polygenic Risk score yet, but they are considering adding it (not as a reflex) but as a redefining of the original service offering.
Kevin Power (Mar 12 2019 at 21:11):
Thanks for confirming @Larry Babb that this is really a single "test" / "order"
Let's start with ServiceRequest: To align with the thinking that @Andrea Pitkus, PhD, MLS(ASCP)CM, CSM outlined in a Panel Orders discussion - would we propose that this be a single ServiceRequest? Or is it multiple ServiceRequests somehow being linked together to represent this panel? Not sure, but perhaps this is an #Orders and Observation WG question?
Larry Babb (Mar 12 2019 at 21:44):
This emerge panel is a single service request. More and more these composite service offerings are showing up in the genomics community almost certainly related to any exome or whole genome assays.
Mullai Murugan (Mar 16 2019 at 00:12):
Sorry, I had somehow missed this thread. Concurring with @Larry Babb , the eMERGE panel is a single service request as is another pilot we are working on. The pilot, HeartCare (sample report attached to gForge) includes three components and associated results i.e. a gene panel, PGx, and PRS.
Andrea Pitkus, PhD, MLS(ASCP)CM, CSM (Mar 18 2019 at 18:20):
@Kevin Power Great questions.
There are 4 different types of orders, 3 of which are panel orders (more than one item). Expect to see a Service Request for the "original order" as coded in FHIR Service Request type, while the individual result elements (aka observations) would be "Observation Definitions" for each one reported (even if a "Diagnosis" with a text blob/narrative response. IF more of a profile, with "nesting" or "tiered" orders, would have a second layer of Service Request orders which link up to the top original Service Request order (the medical/legal requisition for all). We're working through this on 2 different O&O calls
Bret H (Mar 19 2019 at 15:45):
@Mullai Murugan "The pilot, HeartCare (sample report attached to gForge) includes three components and associated results i.e. a gene panel, PGx, and PRS." - This is all in response to one service request for HeartCare? How are the PGx and PRS related? If the report (sorry havn't looked yet) provides a medical assertion based on all the components as supporting tests then sounds like this is a nested diagnostic report....also, this same topic is on another thread within emerge pilot.
Larry Babb (Mar 27 2019 at 12:45):
Can anyone point me at an example of how one might nest DiagnosticReport(s)? Is it even possible. I'd like to begin exploring it as we are unable to make use of the ObsGenePanel in its current state to achieve what we need.
Kevin Power (Mar 27 2019 at 16:51):
I have not seen a specific example of how to next the reports, but the recommendations thus far have revolved around using extensions to relate the DR's:
http://build.fhir.org/diagnosticreport-profiles.html
The specific extensions mentioned as candidates are:
http://build.fhir.org/extension-diagnosticreport-extends.html
http://build.fhir.org/extension-diagnosticreport-summaryof.html
We still have the todo to build some guidance, but would welcome any further input you have?
Jamie Jones (Apr 22 2019 at 16:17):
Ah, looks like this is the thread where I should update/clarify:
In regards to the use of Panel vs nested DiagnosticReports, the group has already voted to add textual guidance to support BOTH methods, so either approach you prefer is aligned with us at this time, and feedback for how either is/isn't working for you is greatly appreciated. (See GF#19937) :)
Mullai Murugan (May 08 2019 at 15:40):
Thank you all for the discussion and feedback. We will go forward with genomic panels and update this group if we run into issues.
Kevin Power (May 09 2019 at 01:07):
FYI @Mullai Murugan - this is the thread with the GF#19937 that changed the name to grouper.
Mullai Murugan (May 09 2019 at 14:32):
Thanks @Kevin Power
Last updated: Apr 12 2022 at 19:14 UTC
