We are doing a query to identify STEC O157:H7 cases in EHR records. <span style=“color: #000000; font-family: Times New Roman;”>Some of the LOINC descriptions are listed as [Presence] and others as [Identifier]. For example the descriptions of the LOINCs below are identical except for these terms:</span>
<p style=“margin: 0px;”><span style=“color: #000000; font-family: Times New Roman;”>44088-3 Escherichia coli O157:H7 DNA <span style=“background: yellow; margin: 0px;”>[Identifier]</span> in Unspecified specimen by NAA with probe detection </span></p>
<p style=“margin: 0px;”><span style=“color: #000000; font-family: Times New Roman;”>38990-8 Escherichia coli O157:H7 DNA <span style=“background: yellow; margin: 0px;”>[Presence]</span> in Unspecified specimen by NAA with probe detection</span></p>
What would be the difference in the results you would get by querying for either one of these LOINCs? <span style=“margin: 0px; font-family: ‘Times New Roman’,serif; font-size: 12pt;”><span style=“color: #000000;”>I have tried looking through the LOINC code book and cannot find a definitive answer. Does ‘Identifier’ denote the test was ordered and not necessarily that O157:H7 DNA was detected? Are there specific instructions we need to give sites to ensure that they are only pulling cases where the result was positive? (We just want counts for patients infected with O157:H7 and not counts that include everyone tested)</span></span>
<span style=“margin: 0px; font-family: ‘Times New Roman’,serif; font-size: 12pt;”><span style=“color: #000000;”>Any clarity on these terms is greatly appreciated.</span></span>
<span style=“margin: 0px; font-family: ‘Times New Roman’,serif; font-size: 12pt;”><span style=“color: #000000;”>Many thanks!</span></span>
<span style=“margin: 0px; font-family: ‘Times New Roman’,serif; font-size: 12pt;”><span style=“color: #000000;”>Myka </span></span>
Here is my question again: Apologies for the formatting appearing in the post:
We are trying to query EHR with LOINCs for E Coli O157:H7. Some of the LOINCs specify “Identifier” and some say “Presence.” What do these terms denote?
For example codes 44088-3 and 38990-8 are identical except for these terms in the description. Does a LOINC noted as “Identifier” merely capture that the test was performed? While “Presence” denotes that the test was positive (ie O157:H7 was detected/present)?
The scale attribute can tell us how the differences may look in the electronic health record. As you alluded, LOINC 38990-8 is used when the laboratory only reports out positive, negative or the alternate form: detected , not detected.
However, 44088-3 as a nominal scale is going to give answers of non-ranking information. There are many genetic differences within O157:H7, and example answers of 44088-3 may be genetic names or types in format. I’d like to put a link in here, but I think the Forum algorithms frown on that for security sake. E.coli O157:H7 contains 463 phage-associated genes. Example answers may contain toxB gene, stcE gene, etc. You can scholar google dot com for O157 H7 DNA identifiers to see academic papers containing this information.
Did you try pulling from the EHR already as a testing prototype? Did you find the different styles of answering, but just didn’t realize they were on two different levels of detail?
All the best,
Thank you so much for the response! I am not doing the query, just on the receiving end of the responses. This question came up because we received much larger numbers than expected. For example, one of our sites reported 498 cases for 44088-3 code over a two year period. This is far in excess of the incidence of O157:H7 in that area which we would expect to be around 10-20 positive results per year. One site reported well over 1000 cases. Possibly the sites did not de-dup there results and we are seeing the same patient counted in multiple contexts.
However, is it also possible that the sites are merely reporting that the test was performed? IE if you queried solely on these terms does it bring up every record regardless of result? Apologies for my ignorance in this area. Just trying to make sense of these inflated numbers and how we can derive the actual incidence from the EHRs. Any thoughts are much appreciated!!
Let me check in with some fellow LOINC’ers and see if we can get you the best answer.
Thank you so much! I greatly appreciate your time and expertise.
Pam and Myka,
This is a challenging question. I referenced the LOINC User Guide (Page 19) and the Guide for Using LOINC Microbiology terms (Page 14 - 18) to attempt to answer this question.
I think we have to look at the LOINC property in addition to the scale type to determine the intent on how these LOINC codes should be used.
38990-8 | Escherichia coli O157:H7 DNA [Presence]… has a a property of “PrThr” (Presence/Threshold). Since the organism name is specified in the component name and the scale type is ord (Ordinal) I would expect a result value from the lab of “Positive/Negative” Or "Negative,Trace, Positive). Hard to saw what the exact scale should be without having the manufacturer information.
Note if this had been a “generic” LOINC code like 11475-1 | Microorganism identified in Unspecified specimen by Culture with a nominal scale, I would expect the result to be the name of the organism likely encoded using a SNOMED code from the organism hierarchy.
44088-3 | Escherichia coli O157:H7 DNA [Identifier]… has a property of “Prid” (Presence or Identity) with a scale of nominal. Page 14 of the LOINC Microbiology Guide indicates Property Prid , Scale Nom = absence, or, if the analyte is present, its specific identity. Since this is a gene detection test, this LOINC code may have been used by the lab to identify the specific DNA marker for O157:H7 or other identifier. I think this puts me in agreement with Pam who indicated “<span style=“background-color: #fbfbfb; font-family: Lato, Helvetica, Arial, Lucida, sans-serif; font-size: 14.82px;”>genetic differences within O157:H7” in an earlier response.</span>
The last component that needs to be taken into consideration is how the lab is actually using these two LOINC codes. The microbiologist may be using the LOINC codes interchangeably, by accident. If we add in the possibility for outside laboratory results being mixed in (e.g. On site lab versus Quest diagnositic) there may be further dissonance in how the LOINC codes are being used. Looking at the result values associated with each LOINC code may provide some clarity here.
A couple of disclaimers:
I would like to have Swapna or someone from the LOINC team confirm this interpretation
How the lab is using the above terms may different from how how LOINC intended them to be used. Ideally, you should confirm these definitions with the microbiologist.
I hope this helps.
Myka and Pam,
I apologize for misunderstanding the question. Pam clarified that this was more a question of data quantity and not one of data identification and interpretation of the LOINC code meanings.
Since we are in the microbiology domain, it is quite possible that the one source is not restricting the data set to the final result. Most labs go through multiple iterations of a test result with tentative results versus final result and how that appears in the EHR may change by organization. All of these iterations are stored in the LIS system, but are rarely if ever transmitted to the EHR, if you are working in a “best-of-breed environment. (e.g. SunQuest + EPIC). This changes when working in a “1-stop-shop” environment. (e.g. EPIC Beaker + EPIC EHR).
- 1 source is sending all results (tentative and final) while the other groups are only sending final. There should be a result status that the data could be filtered on.
-If that is not the case, then they sender may not be constraining the dataset to just a single facility. Meaning, they could be providing results for an entire health system as opposed to a single facility in the health system.
- The last potential cause is mismapping of data by the sending laboratory. A single mismapping could cause the numbers to be extraordinarily high, even if the mismapping was fixed. It is extremely unlikely that a lab would re-result a laboratory test to the EHR just to fix a LOINC mapping. Unless the EHR performs a bulk update, which most do not like to do because of the legal nature of the medical record. For this reason may mismapping live on in legacy data in perpetuity. it is important to review the data set using both the local code / name and the LOINC Code.
I have found the need to employ mapping from Local Codes to LOINC at different levels of a healthcare organization based on use case. One map may be used to map data accurately for clinical purposes, using the most exact LOINC code possible. A research/analytics map may be required to aggregate laboratory data for analytics. For example, mapping method specific LOINC codes to methodless LOINC codes where the method does not alter the clinical interpretation of the test. This may also be done based on specimen in cases were a test being run on Blood versus Serum/plasma does not impact the clinical interpretation. Unfortunately, I have yet to find a situation where a single map fits every use case even though this is desirable to reduce maintenance.
Based on my experience at Geisinger Health Systems , these are the most likely causes of the numbers not being what you expected.
I hope this helps.
Many thanks for both of your responses! As I am a complete novice with LOINC codes, the first was incredibly useful for orientation. I think all of the sources of error your listed are likely in play here. We will follow-up with the sites that are giving us impossible numbers and see if we can search based on the local lab identifiers as you suggested. The mismapping makes a lot of sense for some of these sites.
Thanks again, and thank you, Pam, for sending this query on to John!
We have some EHR tables summarizing lab results we’d like to surface in FHIR. For instance, there is a Parasitology table, and we thought we might use 42807-8 Parasite identified in Isolate for these results, not having any organism-specific test names.
However, some of the result strings are negative, so using “prid” might be incorrect. Is there a more general test we could use for this? Same question for virology, bacteriology, mycology, mycobacteriology.
The most general micro test we currently have is 41852-5 “Microorganism or agent identified in Specimen”
One more level of granularity would be the components series of Parasite identified, Ova or Parasite identified, Virus identified, Bacteria identified, Fungus identified, Mycobacterium sp identified
The sporadic occurences of negative results within the table don’t override the intent of giving a genus/species when an organism is found.
Let me know how else to help!
Thanks, Pam! Exactly what I needed.