TIMS/N3C: [OMOP <--> ICD10CM, etc] --> SSSOM --> FHIR ConceptMap

[joeflack4]

Overview

Stephanie is doing some mapping work, I think for N3C (I'm not sure the exact nature of it). There is a desire create SSSOM for these mappings. After creating the SSSOM files, we'll convert to FHIR ConceptMap and load onto the TIMS server.

Tasks

• [ ] 1. Complete the map
• [ ] 2. Convert to FHIR and upload to TIMS server

Questions

• [x] 1. mapping_justification: What is the valid range? Is there an enumeration of permissible values somewhere?
• [x] 2. mapping_justification: We know that OMOP has mapped to various terminologies using automation. However, we don't know how it was done. Does semapv:lexicalMatch have an abstract parent class semapv:match that we could use? Or is there something in semapv (or elsewhere) for 'unspecified algorithm'?
• [x] 3. Other mapping details: Similar to mapping_justification, if we don't know the details for "use mapping_tool, mapping_tool_version, confidence , match_string, subject_preprocessing, object_preprocessing,", should we leave these fields out? Or should we include them and leave them blank or populate them with some 'unknown' value or null value?
• [x] 4. What CURIE to use for OMOP?

  Additional information

  Slack Conversation

  looked at the docs, but I’m not 100% sure what fields she would have to fill out, and their range / valid values

  Joe Flack 22 hours ago I believe she needs mapping_justification, which has range of EntityReference, but it hasn’t been properly defined yet. just says str.

  Joe Flack 22 hours ago I see an example here: name: mapping_justification description: A mapping justification is an action (or the written representation of that action) of showing a mapping to be right or reasonable. examples:

  • value: semapv:LexicalMatching
  • value: semapv:ManualMappingCuration

  Joe Flack 22 hours ago So I think for (1) curation, she needs to set mapping_justification column to semapv:ManualMappingCuration for the given rows

  Joe Flack 22 hours ago for (2) generated, you could set mapping_justification to semapv:LexicalMatching (word similarity), if that’s how you are doing the matches. but i don’t know the nature of your generated/automated mappings.

  Joe Flack 22 hours ago In addition, for (1) curation, I’m not sure what other fields she would want to fill out, but I’m guessing they would be: author_id, author_label, reviewer_id, reviewer_label, creator_id, creator_label (although those can also apply to (2) generated as well. (edited)

  Joe Flack 22 hours ago For (2) generated, I think she could use mapping_tool, mapping_tool_version, confidence , match_string, subject_preprocessing, object_preprocessing,

  Joe Flack 22 hours ago as well as semantic_similarity_scoreand semantic_similarity_measure if using a semantic matching algorithm rather than lexical

  Joe Flack 22 hours ago @Stephanie Hong Those are my thoughts so far. we can meet tomorrow to go over if you want. Though would be nice to know if @Nico Matentzoglu is aware of anything I missed

  Nico Matentzoglu 7 hours ago You figured it out quite well. There have been some attempts to connect mapping_justifications to their respective slots a bit more tightly, but nothing definite yet. Just note that in an SSSOM table, you need 1 row per mapping_justification. This means that if you have a manually curated and automatically mapped mapping (two justifications for one mapping), you use two rows.

Related

https://github.com/HOT-Ecosystem/hapi-fhir-jpaserver-starter/issues/59 https://github.com/jhu-bids/TermHub/issues/145

[joeflack4]

@Sigfried This isn't related to TermHub but Stephanie doesn't have any code for this yet, and it uses enclave_wrangelr, so just making an issue here.

@stephanieshong FYI, I created this issue for Nico to respond to.

@matentzn Here you are!

[matentzn]

1. mapping_justification currently allowes these values, but it is extensible and can take any input. (Make pr here).
2. You shouldn't, but you can use semapv:UnspecifiedMatching (highly discouraged) or make a comment on https://github.com/mapping-commons/semantic-mapping-vocabulary/issues/1
3. All of these are optional and should be left blank if unknown.

[stephanieshong]

@matentzn - ok. Thank you. I will give you some example of one to many map that you can comment on. @joeflack4 - there should a input parameter of value sets that specifies what to map which list source code and code system. Where should I grab the input from? input - list of source code /code system pair to map input - target code system for the output output- map in SSSOM format.

[joeflack4]

@matentzn Awesome answers, thanks. I spoke with @stephanieshong and got her what she needs. She was asking about how to construct the curie_map for the most part.

1. Nice. I'll file those mapping_justification values somewhere in my notes. Eventually we'll need to update the documentation w/ these.
2. I think we're going to have to use sempav:UnspecifiedMatching. I do have some thoughts about it, which I will share in that other issue.
3. Got it!

[joeflack4]

@matentzn I added a new question (4): What CURIE to use for OMOP?

For now I randomly picked https://www.ohdsi.org/data-standardization/the-common-data-model/ , but I know that this is wrong. I don't know if there is any URL pattern for pulling up pages that show information about OMOP codes.

[matentzn]

What CURIE to use for OMOP?

In all my projects we use OMOP: https://athena.ohdsi.org/search-terms/terms/

All codes resolve there.

[stephanieshong]

Thank you.

[stephanieshong]

@matentzn - thank you I will use athena link to resolve in the curie map. I am just making a note for myself here:

  - equals_string: semapv:LexicalMatching
  - equals_string: semapv:LogicalReasoning
  - equals_string: semapv:CompositeMatching
  - equals_string: semapv:UnspecifiedMatching
  - equals_string: semapv:SemanticSimilarityThresholdMatching
  - equals_string: semapv:LexicalSimilarityThresholdMatching
  - equals_string: semapv:MappingChaining
  - equals_string: semapv:MappingReview
  - equals_string: semapv:ManualMappingCuration

[joeflack4]

@Sigfried Just FYI TIMS woulda been the better place to put this issue, but w/ the discussion here so far, too late to move it. I'll make a mirror issue and link it.

[stephanieshong]

@joeflack4 - did you remove the sssom_map branch?

[Sigfried]

@joeflack4 Stephanie is asking me where the sssom_mapping branch and files went. I didn't know. I had her add them to the develop branch: 077da0c808a754204ce483a4b31b2d087d2cde1d

[stephanieshong]

@joeflack4 - checked in my sample files under the sample_mapping_files directory

[joeflack4]

Hey @stephanieshong, I know we talked about this on Teams. Yeah, I'm not sure; I don't remember us creating that branch.

I think @Sigfried and I have not yet decided where mapping content should go. Maybe we need a new git submodule? Or we could move them into termhub-csets, if we want to use that submodule for mapping content as well.

For now I agree they are fine there. Could always just add these to a standalone repo as well.

[stephanieshong]

Yes, I created the branch from my machine. But I think I created from the sub-directory and then later removed the .git so i think it got lost. all I need to check-in was the sample SSSOM mapping file for Nico to review.

[stephanieshong]

@joeflack4 - I will create a branch from the main and use that branch for the SSSOM mapping files. This has nothing to do with TermHub for now. Unless you want to use it to map the terminologies from the TermHub which is related. (ICD10CM to SNOMED CT)

[stephanieshong]

A ---[narrowMatch]-->B means that B is narrower than A. So for most part we would be using [boardMatch]