2021-07-07

Owned by Anonymous
Last updated: Jul 07, 2021 by Anonymous
10 min read

High-level notes and Action Items:

  • Kenneth Gersing to provide data to Dr. Campbell

  • Ana Szarfman to write up the post-market use cases

  • Jim Campbell to provide a metadata table for top lab reports in US domain

  • Deloitte to assist with scheduling smaller minutes which will be dedicated on a specific topic/discussion

  • Riki suggested adding Unit of Measure on slide 7 for Qualitative Results

    • Team has to agree on which system to use

    • Missing Method

  • Next meeting should be dedicate of discussion of a session and not brainstorm

    • ex: discussion with vendors regarding their component

    • ex: public health piece

 

 

CHAT HISTORY

From Andrea Pitkus, PhD, MLS(ASCP)CM to Everyone: 12:09 PM
Are the use cases different from those in the grid Riki started?
Thanks for the clarification.
From Jim Campbell to Everyone: 12:10 PM
Those requirements are necessary but not sufficient
From Mary Edgerton to Everyone: 12:11 PM
Even in first case of interoperability individual labs can very in results so even in first case reference material is needed
From Daniel Herman to Everyone: 12:13 PM
Lots of tests are not harmonized/standardized. It will take while. We should aim to do something to help interpret those test results, when useful for particular use cases
From Andrea Pitkus, PhD, MLS(ASCP)CM to Everyone: 12:14 PM
Define harmonization, normalization, standardization as it means different things to different people. ;)
From Mary Edgerton to Everyone: 12:15 PM
Will these slides be distributed?
From Ana Szarfman to Everyone: 12:15 PM
Totally agree with Andrea.
From Riki Merrick | APHL to Everyone: 12:15 PM
HL7 does have rules for communicating UDI in all of it product families - need to have folks adopt them
From Andrea Pitkus, PhD, MLS(ASCP)CM to Everyone: 12:15 PM
Will be challenging to define orders with reflex, panels, etc defined on the local level.
From Mary Edgerton to Everyone: 12:16 PM
Some of this will require vendor engagement to get the codes put into reporting
From Riki Merrick | APHL to Everyone: 12:16 PM
agreed
From Nanguneri Nirmala to Everyone: 12:17 PM
Versioning for historical data: Should we discuss how far back it should be done?
From Nathan Davis to Everyone: 12:17 PM
We should focus on reportable/result codes. Orderables will always be in flux and have a high degree of local use/definitions (panels, batteries, etc.)
From Dan Rutz to Everyone: 12:18 PM
Agreed. Orderables are often assumed to be something that can be standardly codified and normalized, but they're not. And in reality it's not a target worth trying to hit in this project.
From Andrea Pitkus, PhD, MLS(ASCP)CM to Everyone: 12:19 PM
Agree Nathan. Start with results first and perhaps work back to orders as more complex....
From Mary Edgerton to Everyone: 12:19 PM
I think that we need to set standards for going forward, and then backtrack to deal with historical data.
From Carolyn Hiller to Everyone: 12:19 PM
can we include funding for software interface - ie, make it user friendly?
From Riki Merrick | APHL to Everyone: 12:19 PM
I agree - keep orderables out of this - potentially with the exception of fixed multiplex PCRs
From Andrea Pitkus, PhD, MLS(ASCP)CM to Everyone: 12:20 PM
recommend starting with chemistry and hematology and working towards genomics and pathology (using more discrete data/where codes more available)
From Daniel Herman to Everyone: 12:20 PM
How do we ‘automatically’ match UDI to a specific LOINC without consistent orders?
From Jim Campbell to Everyone: 12:21 PM
Orderables are the only data source which define the preparation and specimen processing dimensions and must be integrated with the laboratory data management to fully define a reportable
From Julia Skapik to Everyone: 12:21 PM
love Carolyn's comment about interface-- would hope that there would only need to be one new interface or system that could bridge several of these excellent ideas
From Dan Rutz to Everyone: 12:21 PM
I think we need to assume that the LIS (or instrument manager, from IHE perspective) interprets ordering provider's request and turns it into specific tests for analyzers to perform.
From Andrea Pitkus, PhD, MLS(ASCP)CM to Everyone: 12:21 PM
Start with getting those reporting non discretely/non digital (still paper/fax) to electronic
From Julia Skapik to Everyone: 12:22 PM
does it make more sense to start at the lab vendors and move to their customers? It seems at big academic medical centers they own the whole thing so the interoperability problem isn't well represented
From Andrea Pitkus, PhD, MLS(ASCP)CM to Everyone: 12:23 PM
Dan, there are a few steps here. One is for profiles/physician convenience or reflex orders, the performing lab has these as discrete orders (even if one click in EHR). 2nd part is in the lab, where the discrete LIS orders may result in multiple test requests of a specific or multiple analyzers.
From Riki Merrick | APHL to Everyone: 12:23 PM
@Daniel - you can map the UDI to the LOINC for the performed test - for each lab they will have specifically defined orderables, but across labs those orderables can be very different, so at the national level harmonizing orders is TOO big a bite
From Andrea Pitkus, PhD, MLS(ASCP)CM to Everyone: 12:27 PM
with units, also scale
From Mary Edgerton to Everyone: 12:27 PM
Along with units of measure there would be high and low threshold of normal
From Andrea Pitkus, PhD, MLS(ASCP)CM to Everyone: 12:27 PM
is it mg/dL or mg/L
From Gregory Pappas to Everyone: 12:27 PM
I'm assuming Deloitte can save all these comments. Maybe they can go on the Confluence page. All comments seem relevant and we can address them
From Ana Szarfman to Everyone: 12:27 PM
I would start by targeting usability predictions - i.e., how to better predict sepsis
From Andrea Pitkus, PhD, MLS(ASCP)CM to Everyone: 12:27 PM
or is there an assumption that calculations will be used for result values
From Daniel Herman to Everyone: 12:28 PM
Specimen. How specific is Specimen SNOMED. At minimum need specimen type, but better would be a specific UDI
From Andrea Pitkus, PhD, MLS(ASCP)CM to Everyone: 12:28 PM
also would include Method if applicable or detection limit
Do you mean qualitative results values where the SCT codes are applied
From Riki Merrick | APHL to Everyone: 12:29 PM
yes for qual value sets using SCT codes
From Andrea Pitkus, PhD, MLS(ASCP)CM to Everyone: 12:29 PM
per CLIA the vendor result values are to be used, and there are SCT codes for many
From Riki Merrick | APHL to Everyone: 12:29 PM
and nominal answers for organism lists
From Carolyn Hiller to Everyone: 12:29 PM
Great presentation. There are so many stakeholders - we really are building a marina.
From Andrea Pitkus, PhD, MLS(ASCP)CM to Everyone: 12:29 PM
will organisms be precoordinated or post coordinated
From Mary Edgerton to Everyone: 12:30 PM
Platform, kit, specimen type, specimen transfer conditions, method, result type(qual or quad), result, units (if applicable), Certified Reference Material used, method of harmonization (calibration, normalization), last date harmonized
From Scott Campbell to Everyone: 12:30 PM
Can I have a life jacket in case I get pushed in the water?
From Andrea Pitkus, PhD, MLS(ASCP)CM to Everyone: 12:30 PM
agree with Nathan per CLIA regulations by law and accreditation
From Riki Merrick | APHL to Everyone: 12:30 PM
@Daniel - SNOMED is pretty precise for specimen (we have been working on a specimen cross mapping table)
From Andrea Pitkus, PhD, MLS(ASCP)CM to Everyone: 12:30 PM
Can we send Scott some lifesavers (chew on that... ;) )
From Scott Campbell to Everyone: 12:31 PM
@Andrea LOL! Thanks….I
From Julia Skapik to Everyone: 12:31 PM
agree-- the responses should be standardized as well
From Andrea Pitkus, PhD, MLS(ASCP)CM to Everyone: 12:31 PM
for susceptibility testing, recognize that some labs report mm and then interpretation as S, I , R while others report out just the interp
From Carolyn Hiller to Everyone: 12:32 PM
Does anyone know the folks that decided to make phones/tech plug-n-play? Thinking that decision happened in a bar somewhere.... What were their motivators?
Agree! Forward looking is our only viable path.
From Andrea Pitkus, PhD, MLS(ASCP)CM to Everyone: 12:32 PM
It might be good to identify the "easier" items/low hanging fruit and those that are more challenging/need more time/effort
From Riki Merrick | APHL to Everyone: 12:32 PM
@Mary - that's a good list - question around specimen transfer conditions - is that applicable at the IVD kit level, or is that something we need to track at the instance level
From MariBeth Gagnon to Everyone: 12:33 PM
Agree with Hung, developing a value set and guidance provided to manufacturers, can make results more harmonized moving forward.
From Mary Edgerton to Everyone: 12:35 PM
@Riki instance level. I have thought this out as a set of data elements to identify platforms, kits, specimen types allowable…and a separate one for results data to be aggregated where I include specimen transport conditions. All codes are fixed in background so that there are not multiple users inputting multiple near-synonymous or synonymous codes, i.e. to enforce uniqueness
@Riki-one set of data elements supports the LIVD files and he second for the view of results across institutions and a litmus test of harmonization to consider them clinically interoperable
From Ana Szarfman to Everyone: 12:35 PM
What a shame: We are dealing with scrambled eggs
From Mary Edgerton to Everyone: 12:36 PM
@Szarfman ar you referring to the variable input in the chat or the state of lab data and results reporting across the US?
From Riki Merrick | APHL to Everyone: 12:37 PM
@Mary - do you agree that these are the ones for LIVD:
Platform,
kit,
specimen type
method
result type(qual or quant)
result
units (if applicable)
From Daniel Herman to Everyone: 12:37 PM
Does specimen refer to specimen type or specimen type + container type?
From Ana Szarfman to Everyone: 12:37 PM
the state of lab data
From Riki Merrick | APHL to Everyone: 12:37 PM
Specimen covers: type + source site + collection method + additive (container type)
From Mary Edgerton to Everyone: 12:38 PM
I have formulated a list based on the current LIVD file at CDC-want to all later? But those are some of them that stick in my head
Then there is component and matrix
From Jim Campbell to Everyone: 12:38 PM
NCI metathesaurus is underspecified for this task as metadata resource
From meedgerton to Everyone: 12:38 PM
@Riki all was talk
From Andrea Pitkus, PhD, MLS(ASCP)CM to Everyone: 12:40 PM
@dan's question. Specimen type and specimen source are confused often by laboratories and miscoded (SCT specimen type codes applied to sources such as head, back or neck, which should be from autopsy specimens and not micro cultures)
From Jim Campbell to Everyone: 12:40 PM
Nebraska could provide a metadata table such as specified for top lab reports in US domain
From Mary Edgerton to Everyone: 12:40 PM
One way to deal with coding is to provide a human readable list of data values with codes embedded in the background.
From Daniel Herman to Everyone: 12:41 PM
For prioritization, probably should sample challenging cases also. Assays that are not well harmonized/standardized and complicated specimen/orders
From Andrea Pitkus, PhD, MLS(ASCP)CM to Everyone: 12:41 PM
May wish to scope as in lab needs vs downstream user needs. For example, most all performing lab naming conventions differ in EHRs/LIS downstream systems contributing to variability in naming
From MariBeth Gagnon to Everyone: 12:41 PM
Daily Med DailyMed was a model we previously considered
From Jim Campbell to Everyone: 12:42 PM
The expanded SNOMED CT LOINC metadata includes METHODs
From Andrea Pitkus, PhD, MLS(ASCP)CM to Everyone: 12:42 PM
Are you defining specimen as collected specimen, analyzed/performed specimen and derivatives
From MariBeth Gagnon to Everyone: 12:42 PM
CD would like to know the setting where the test was performed.
From Andrea Pitkus, PhD, MLS(ASCP)CM to Everyone: 12:42 PM
most chemistries collect whole blood via venipuncture, but testing is performed on serum or plasma.
From Carolyn Hiller to Everyone: 12:43 PM
When prioritizing tests - who's priorities? Input from the stakeholder groups could be useful. ie - what would be of interest to payers? CDC/public health? others?
From Mary Edgerton to Everyone: 12:43 PM
What are we prioritizing-by disease type and its importance to public health, or by lab tests currently used in AI/predictive models such as likelihood to develop sepsis?
From Riki Merrick | APHL to Everyone: 12:43 PM
@MariBeth: that would be something on the instance data, not at the LIVD file level
From Andrea Pitkus, PhD, MLS(ASCP)CM to Everyone: 12:45 PM
Micro uses derived specimens such as urine culture for gram stain/ some testing, but then isolate is used for sensitivities, some identifications, other testing (ie. serotyping, pcr)
RE Method: clinically significant results interpretation from bromocresol green vs bromocresol purple for albumin results per literature
From Mary Edgerton to Everyone: 12:45 PM
@Andrea I think that might break dow by specimen type (material collected + container + medium), then components, matrix and finally analyze.
Last word should be analyte
From MariBeth Gagnon to Everyone: 12:46 PM
@Riki, yes reported at instance, but would you need codes in LIVD?
From Riki Merrick | APHL to Everyone: 12:47 PM
@MariBeth: you mean indicating in what setting this IVD can be used?
From MariBeth Gagnon to Everyone: 12:48 PM
@Riki, yes what settings are approved for use
From Mary Edgerton to Everyone: 12:48 PM
I might collect an NP swab in a test tube, use RNALater to extract mRNA, convert to cDNA, then use primers to generate a result of detected or non-detected. My analyte is th cDNA sequence, my specimen in the NP swab in empty container, my components are mRNA and cDNA, and my matrix that goes into the machine iscDNA and whatever solvent it maybe in.
From Riki Merrick | APHL to Everyone: 12:48 PM
@Ana: can you write up the post-market use case you want to look at?
From Mary Edgerton to Everyone: 12:50 PM
(I can spell but I can’t type and Otto Correct often does not speak my language)
Remind us how to access the confluence system
From Daniel Herman to Everyone: 12:50 PM
If we define goal as being able to interpret lab test results in a ‘non-original context’. We need a lot more details. Many of these might be assumed, but worth pointing out. Specimen details — collection time, run time. Result information — reference interval, result status. Indication. ….
From Riki Merrick | APHL to Everyone: 12:51 PM
SHIELD Coordination
From Jim Campbell to Everyone: 12:51 PM
How about an exemplar data set for visualization/discussion?
From Andrea Pitkus, PhD, MLS(ASCP)CM to Everyone: 12:52 PM
agree on what(s). Think about goals, how to measure, goal (how do you know when missed or met)?
Needs do vary depending on where in the total testing process/lab continuum the data are utilized.
From Julia Skapik to Everyone: 12:54 PM
Thank you for that!