"patient-friendly" descriptions of LOINC/SNOMED/etc codes.

[jaspermayone]

Creating this issue to stash some resources we have found to find "patient-friendly" descriptions of LOINC/SNOMED/etc codes.

I will post resources here shortly.

[jaspermayone]

Both https://github.com/IHTSDO/snomed-database-loader and https://github.com/IHTSDO/snowstorm look like good SNOMED options, provided by https://github.com/IHTSDO

[jaspermayone]

We know that snomed content is licensed - we are not sure how it is for open-source, but commercial usage of it definitely requires the vendor to have a license

[jaspermayone]

We know that snomed content is licensed - we are not sure how it is for open-source, but commercial usage of it definitely requires the vendor to have a license

See https://www.snomed.org/snomed-ct/get-snomed

[jaspermayone]

From the Fasten discord....

Vadi: snomed content is licensed - I'm not sure how it is for open-source, but commercial usage of it definitely requires the vendor to have a license see https://www.snomed.org/snomed-ct/get-snomed

Analog J: @Vadi my understanding is that the HL7 Fhir foundation has a agreement/partnership with SNOMED to officially support their codes, which is why they show up in the FHIR exports and terminology server. Since there's no "patient friendly" description of codes, I don't think we need a SNOMED license, since we won't be using the raw code dataset

Vadi: Yeah, codes which are included in the FHIR spec itself are okay. Specifically, the Global Patient Set: https://terminology.hl7.org/3.1.0/ValueSet-snomed-intl-gps.html More info: https://www.snomed.org/news-and-events/articles/broadening-intl-patient-care-interoperability- gps

[jaspermayone]

Also from @AnalogJ in the Discord.....

Hi,

I'm working on a patient facing application and I'd like to display LOINC/SNOMED/etc codes with additional context -- basically patient friendly descriptions of the codes.

eg. when searching LOINC LP14890-5 we arrive at https://loinc.org/LP14890-5/ which has 2 text blocks under the Descriptions header - one from Wikipedia and one from Lab Tests Online

However when searching the LOINC dataset CSV files (or the FHIR terminology server https://tx.fhir.org/loinc/

[jaspermayone]

Also from @AnalogJ in the Discord.....

Hi, I'm working on a patient facing application and I'd like to display LOINC/SNOMED/etc codes with additional context -- basically patient friendly descriptions of the codes. eg. when searching LOINC LP14890-5 we arrive at https://loinc.org/LP14890-5/ which has 2 text blocks under the Descriptions header - one from Wikipedia and one from Lab Tests Online However when searching the LOINC dataset CSV files (or the FHIR terminology server https://tx.fhir.org/loinc/

That post can be found here.

or at https://chat.fhir.org/#narrow/stream/179202-terminology/topic/Patient.20Friendly.20LOINC.2FSNOMED.20descriptions/near/320988014

[jaspermayone]

From Carol Macumber In reply...

If the additional information is available in the LOINC release files, they could be represented as codesystem.concept.definition in the tx.fhir LOINC Code System. Rob Hausam looks like the LOINC code system resource at tx.fhir is limited to the axises, code and LOINC name (similar to what's available from Regenstrief directly https://fhir.loinc.org/CodeSystem/$lookup?system=http://loinc.org&code=LP14890-5). If this reference material were to be represented as a definition, there doesn't appear a way to declare it's origin (in this case wikipedia and lab tests online). There is a code system extension for source reference, but that's extending codesystem not codesystem.code http://build.fhir.org/extension-codesystem-sourcereference.html

[jaspermayone]

@AnalogJ#3506 my UMLS License was just Approved.... I will take a look at the data they provide in a little bit....

[jaspermayone]

[jaspermayone]

I think we want Value Set Authority Center (VSAC) and/or SNOMED CT

RxNorm could also be useful to integrate.

[jaspermayone]

https://documentation.uts.nlm.nih.gov/rest/home.html

[jaspermayone]

[jaspermayone]

When you make a get request to the uri inside the first result, you get

[jaspermayone]

The key there is the definitions. When you preform a get request on that, you get

[jaspermayone]

The full object body for that, is

{
    "pageSize": 25,
    "pageNumber": 1,
    "pageCount": 1,
    "result": [
        {
            "rootSource": "MSHCZE",
            "value": "Zlomeniny některé kosti/kostí v kotníku. Bimaleolární fraktura = fraktura obou kotníků, trimaleolární zlomenina = fraktura obou kotníků a zadní hrany tibie.",
            "classType": "Definition",
            "sourceOriginated": true
        },
        {
            "rootSource": "NCI",
            "value": "Traumatic or pathological injury to the ankle joint in which the continuity of an ankle bone is broken. Symptoms include pain, swelling, and difficulty moving the affected leg and foot.",
            "classType": "Definition",
            "sourceOriginated": true
        },
        {
            "rootSource": "MSH",
            "value": "Fractures of any of the bones of the ANKLE.",
            "classType": "Definition",
            "sourceOriginated": true
        },
        {
            "rootSource": "HPO",
            "value": "A fracture or multiple fractures of one or more of three bones in the ankle joint: the tibia (shinbone), the fibula (outer ankle bone), and the talus (which is the bone that connects your leg to your foot). [https://www.hss.edu/condition-list_ankle-fractures.asp]",
            "classType": "Definition",
            "sourceOriginated": true
        },
        {
            "rootSource": "MSHNOR",
            "value": "Brudd i ankelbeina.",
            "classType": "Definition",
            "sourceOriginated": true
        },
        {
            "rootSource": "MSHPOR",
            "value": "Fraturas de quaisquer ossos do TORNOZELO.",
            "classType": "Definition",
            "sourceOriginated": true
        },
        {
            "rootSource": "MSHSPA",
            "value": "Fracturas de cualquiera de los huesos del TOBILLO.",
            "classType": "Definition",
            "sourceOriginated": true
        }
    ]
}

[jaspermayone]

For LONIC, the first 2 steps work, but there is no definition...

Step 1 for a LONIC code looks the same in terms of a response. For the 2nd response, the LONIC res looks like this.

{
    "pageSize": 25,
    "pageNumber": 1,
    "pageCount": 1,
    "result": {
        "ui": "C4739649",
        "name": "Fracture Risk Assessment:-:Pt:^Patient:-:FRAX",
        "dateAdded": "05-06-2019",
        "majorRevisionDate": "01-03-2019",
        "classType": "Concept",
        "suppressible": false,
        "status": "R",
        "semanticTypes": [
            {
                "name": "Clinical Attribute",
                "uri": "https://uts-ws.nlm.nih.gov/rest/semantic-network/2022AB/TUI/T201"
            }
        ],
        "atoms": "https://uts-ws.nlm.nih.gov/rest/content/2022AB/CUI/C4739649/atoms",
        "definitions": "NONE",
        "relations": "NONE",
        "defaultPreferredAtom": "https://uts-ws.nlm.nih.gov/rest/content/2022AB/CUI/C4739649/atoms/preferred",
        "atomCount": 4,
        "cvMemberCount": 0,
        "attributeCount": 0,
        "relationCount": 0
    }
}

[AnalogJ]

Hey @j-dogcoder Thanks for diving into this, this is great!

So it sounds like you were able to find the "Description" for SNOMED codes but not LOINC? Hm. Is there another way that we can pull the LOINC descriptions? maybe using a medical wiki database query?

• https://www.wikidata.org/wiki/Wikidata:Data_access
• https://wikidoc.org/index.php/Main_Page - MediaWiki API
• https://www.wikidoc.org/index.php/WikiPatient - MediaWiki API
• https://en.wikipedia.org/wiki/Health_information_on_Wikipedia - MediaWiki API

[jaspermayone]

Hrm.... none of the suggested links seem to have anything. I will keep looking. The SNOMED description seems to be patent friendly enough

[jaspermayone]

Are you looking for a dataset you can store? Or will API queries be sufficient?

Being able to store the data will allow Fasten to be closer to offline able

[AnalogJ]

@j-dogcoder fixed some of the links above. Basically wikidata is a wiki-like database, with easy access to data in json format. Wikidoc and Wikipedia are both built ontop of MediaWiki, which has an API.

Are you looking for a dataset you can store? Or will API queries be sufficient? Being able to store the data will allow Fasten to be closer to offline able

Agreed, ideally we'd be able to generate a dataset of patient-friendly descriptions for each code, and have that available in the Fasten database, without requiring API access. Most of Wikipedia's content is licensed under Creative Commons, so this should be fair use as long as we reference Wikipedia in the UI.

Since you found SNOMED patient-friendly descriptions already, should we start with them? Write a script to extract the code, description, and any other fields that would be useful for Fasten, and then import it into the database?

[AnalogJ]

I also posted on LOINC's contact-us page, asking them the same question as my FHIR Tulip post

[AnalogJ]

Hell, it might be interesting to use ChatGPT to create summaries of each code as well -- if we had a good way to verify that the results weren't complete lies 😨

[jaspermayone]

Hell, it might be interesting to use ChatGPT to create summaries of each code as well -- if we had a good way to verify that the results weren't complete lies 😨

That could be interesting, but without a way to verify that, we should steer clear for now.

[jaspermayone]

@j-dogcoder fixed some of the links above. Basically wikidata is a wiki-like database, with easy access to data in json format. Wikidoc and Wikipedia are both built ontop of MediaWiki, which has an API.

Are you looking for a dataset you can store? Or will API queries be sufficient? Being able to store the data will allow Fasten to be closer to offline able

Agreed, ideally we'd be able to generate a dataset of patient-friendly descriptions for each code, and have that available in the Fasten database, without requiring API access. Most of Wikipedia's content is licensed under Creative Commons, so this should be fair use as long as we reference Wikipedia in the UI.

Since you found SNOMED patient-friendly descriptions already, should we start with them? Write a script to extract the code, description, and any other fields that would be useful for Fasten, and then import it into the database?

I agree that ideally we would have it in the DB... I will write a script to extract it all... do you just want it as json? or... how are we getting this into the db?

[jaspermayone]

@AnalogJ What kind of db do we use?

[jaspermayone]

I was browsing https://www.nlm.nih.gov/research/umls/licensedcontent/umlsknowledgesources.html and downloaded both the UMLS Metathesaurus Full Subset and MRCONSO.RRF

[jaspermayone]

@AnalogJ what one looks better for our usecase?

[jaspermayone]

[jaspermayone]

Or is there something else on that page that looks better?

[jaspermayone]

would you like be to import it into the db?? or just upload the file for you to use?

[AnalogJ]

sorry for the delay getting back to you @j-dogcoder

• the database type is: SQLite
• MRCONSO.RRF seems like its a more reasonable size, however the sample data/examples don't seem to have a description column -- were you able to find one in the real dataset

[AnalogJ]

on the other hand, the Definitions file - MRDEF.RRF - https://www.ncbi.nlm.nih.gov/books/NBK9685/#ch03.sec3.3.6 does have a description

[jaspermayone]

I can't seem to find that file in the download, and I don't seem to see definitions in the other one. The definitions I found were in the API... maybe I need to write some script to query the API for every single code?

[jaspermayone]

@AnalogJ what are your thoughts on this?

[jaspermayone]

Ugh... when searching with exact, it doesn’t work at all... (doesn’t provide any results) 😠 😠 😠 @AnalogJ plan b?

[AnalogJ]

Just some notes for future me:

---

Here are some medical databases that can be used to find patient-friendly articles by LOINC code:

• MedlinePlus - This is a website run by the US National Library of Medicine that provides patient-friendly information about medical conditions, treatments, and drugs. It can be searched using LOINC codes.
• Healthwise - This is a database of patient education materials that is used by many healthcare organizations. It includes articles and videos on a wide range of topics, and can be searched using LOINC codes.
• UpToDate - This is a clinical decision support resource that includes patient education materials. It can be searched using LOINC codes to find articles on specific topics.
• Mayo Clinic - This is a website that provides patient-friendly information about medical conditions, treatments, and procedures. It includes articles and videos, and can be searched using LOINC codes.
• Health on the Net (HON) - This is a non-profit organization that certifies health websites for accuracy and reliability. Its database includes patient education materials that can be searched using LOINC codes.
• Patient Education Reference Center (PERC) - This is a database of patient education materials that includes articles, videos, and handouts. It can be searched using LOINC codes to find articles on specific topics.
• ClinicalKey - This is a clinical decision support resource that includes patient education materials. It can be searched using LOINC codes to find articles on specific topics.

---

• Unified Medical Language System (UMLS) API: The UMLS API provides access to a collection of medical vocabularies and terminologies, including SNOMED, LOINC, and ICD codes. It is free to use but requires registration.
• MedlinePlus API: The MedlinePlus API provides access to health information and medical dictionary definitions from the National Library of Medicine, and can be searched using SNOMED, LOINC, and ICD codes. It is free to use and requires an API key.
• IBM Micromedex API: The IBM Micromedex API provides access to drug information and medical dictionary definitions, and can be searched using SNOMED and ICD codes. It is available for a fee.
• Healthdirect API: The Healthdirect API provides access to health information and medical dictionary definitions, and can be searched using SNOMED and ICD codes. It is available for a fee.
• Apixio API: The Apixio API provides access to medical dictionary definitions, and can be searched using SNOMED and ICD codes. It is available for a fee.

[AnalogJ]

https://www.wikidata.org/wiki/Q183392

• exact match
• https://disease-ontology.org/?id=DOID:10609

[jaspermayone]

👍🏻

[jaspermayone]

@AnalogJ Should have this done on the repo by the 2nd week in April

[AnalogJ]

I think this is what we're going to want to use:

**Search By Codes*** https://medlineplus.gov/medlineplus-connect/web-service/

https://medlineplus.gov/about/developers/webservices/

examples:

• https://connect.medlineplus.gov/service?mainSearchCriteria.v.cs=2.16.840.1.113883.6.103&mainSearchCriteria.v.c=250.33&knowledgeResponseType=application/json
• https://connect.medlineplus.gov/service?mainSearchCriteria.v.cs=2.16.840.1.113883.6.90&mainSearchCriteria.v.c=E10&knowledgeResponseType=application/json

*Search By Text

https://medlineplus.gov/about/developers/webservices/

It's not perfect, but it supports SNOMED, LOINC and ICD-9/10 codes.

[AnalogJ]

I'm going to try to play around with it today.

Our fallback might be the wikidata database or Unified Medical Language System (UMLS) api.

[jaspermayone]

👍🏻

[jaspermayone]

Hey @AnalogJ, looking at your changes... For best offline usability, perhaps these things can be stored in the lighthouse db? I am planning on updating the script in Vital-Link/patient-friendly-codes, which exports it all to JSON, and periodically updates it if anything changes (will do something similar with Issue #95 (NPI numbers) after I get this finished up. In the long term, I will probably expose the codes over an API as well... Fasten is welcome to use one or both of those when it's complete, so only one of us has to maintain it if that makes sense.

[jaspermayone]

Ah! Keep putting things in the wrong repo!

[AnalogJ]

Hey @j-dogcoder

For best offline usability, perhaps these things can be stored in the lighthouse db? I am planning on updating the script in Vital-Link/patient-friendly-codes, which exports it all to JSON, and periodically updates it if anything changes

Yep, this data needs to be available offline. I think the best place to do this caching is in the Fasten db (locally).

• Medline Plus Connect is already running at scale, and has reasonable rate-limiting (100 requests/min).
• They already understand LOINC/SNOMED/ICD codes, so we don't need to create a lookup/mapping table ourselves.
• if Fasten has to call the Lighthouse (or a different 3rd party API) to pull descriptions for each medical condition code, technically the Lighthouse could map ip-addresses to medical conditions related to a household. Removing that extra step saves us some technical infrastructure and builds more trust (hopefully).

---

I think the next step should be to work on NPI numbers, finding a way to look them up/map them to each Organization.

I've been noodling around with a couple of ideas for how we can build something to allow the community to contribute. I'll update https://github.com/fastenhealth/fasten-onprem/issues/95

[jaspermayone]

Glad this got completed! Left you a comment on the NPI numbers issue.

Will continue to keep this open in Vital-Link/patient-friendly-codes#2 for anyone interested.

[AnalogJ]

[AnalogJ]

http://lite.bmi.utah.edu/OpenInfobuttonDemo.html#