Effective Use of Codes
Tagged: Effective Use of Codes
- This topic has 3 replies, 1 voice, and was last updated 12 months ago by Andrea Pitkus.
June 9, 2021 at 7:23 pm #33469Hera AshrafKeymaster
When a system sends clinical data to another system, it can include references to external “Code Sets”, such as LOINC, CPT, or CVX. This allows the receiving system to map the data, a medication for example, to the local representation of that element, which in turn allows the data to be “understood” by the receiving system. Coded data can be more easily incorporated into clinical decision support and may make reconciliation, tracking, trending and searching easier.
Reply to this post with your answers to these questions:
1. As a recent use case: What has been the most difficult part of integrating outside COVID data – tests, diagnosis, and/or immunizations, etc., into your workflow?
2. What data types(e.g. Labs, radiology, PAMI) from outside sources would be most useful in your practice if they could be used in automated decision support, graphing for trend or other data visualization tools and medical decision making within workflow.
3. Is it valuable to prioritize specific data elements to be more reliably encoded (e.g. common lab tests), if it means getting Clinical Decision Support for those elements more quickly and for easier integration and use at the point of care within clinical workflow?
June 30, 2021 at 10:42 pm #34595Riki Merrick
I have a problem of calling lab tests data types – they are data elements, or if you want to use type, call them types of data (data type defines the format depending on the content, which is not wha you are asking here).
SHIELD (Systemic Harmonization and Interoperability Enhancement for Laboratory Data) is a public provate partnership that is working on a strategic plan with the goal to “Name the same test the same way across the healthcare continuum” – this group is tackling the difficult problem of making sure lab results from differnt performers / vendors / instruments can be compared without the risk to patient safety.
July 8, 2021 at 1:08 pm #36029Tom Bronken
While proper coding of data can facilitate automatic use of the data, this will probably be restricted to automatic alerts and reminders for the foreseeable future. Taking an automatic action on a patient is fraught with risk, and is not something we’ll see soon.
Codes are most useful for locating specific outside data by a user on the receiving end. Our EHRs are designed for efficient hurting down of internal data desired by the user. Unfortunately, data from outside sources arrives piecemeal. Assigning codes (especially if similar codes can be grouped) allows organization of the data and makes searching for and finding what is looked for much more efficient.
Regarding lab values specifically, LOINC coding is well-developed, but reference ranges vary. With accurate lab value LOINC coding accompanied by reference ranges in the metadata, graphing and trending is possible and would be useful.
Currently, documents rarely have accurate and granular LOINC codes when received, even though these are available. This makes finding a document of interest very difficult.
July 8, 2021 at 6:28 pm #36074Andrea Pitkus
Concur with previous posts.
Regarding lab data to be usable as proposed in clinical decision support, etc. it needs to be:
1. electronic (paper doesn’t cut it any more)
2. discretely captured (pdf reports, scans of faxed reports, text blobs may be mapped to a single LOINC at best which may be so generic to not be helpful such as pathology report or reference lab report)
3. encoded accurately (and at the most detailed level) at the source/its origination. Lab data shouldn’t be mapped downstream by those who don’t have access to package inserts or other nuances of testing as they will only be able to encode at a higher level missing key information. Errors are more likely to occur when done by non laboratory/informatics/terminology folks too.
4. Laboratory data comprises over 70% of EHR data and utilized in clinical decision making (older Mayo study published by Dr. Rodney Forsman)
5. With encoding common lab results, it will depend on all the variances in how the test is performed. Again this is best known by the performing lab. LIVD COVID maps on the CDC website provide a great example of the hundred of ways a “COVID test” can be performed. PCR, Antigen and Antibody results are impact decision making differently as so screening, diagnostic and surveillance results.
6. With clinical decision support design/development, often it’s not a single data element or kind of data, but a combination of them. May wish to focus on simpler use cases/fewer data items/high impact use cases/scenarios, but many decisions are complex or involve multiple kinds of data for each decision and clinical care involves many decisions each hour. Decision support should be clinically validated to ensure harm doesn’t result either. Recent U MI study indicates Epic’s Sepsis indicator is missing many cases of sepsis. The question may be which data are best assessed by a health professional/clinician and which can be automated?