Kate – thank you for your responses. Let me first say how much I appreciate the initiative of so many and the time and effort that have gone into the development of LOINC; a truly remarkable and commendable effort.
Let me rephrase my concern. First I understand that UNITSREQUIRED merely specifies the need for UOM when a particular LOINC code is used in an HL7 message. However, I would assume that if the response is ‘Y’ then it would be appropriate or even required that the UOM for that code would be included in a LOINC record. Again that may just be my naiveté.
What is more concerning with regard to what I think is potentially important application of LOINC is the lack of not only UOM specification but also lack of EXAMPLE_UCUM_UNITS in many of the LOINC records. For example, of Pathology related entries with PROPERTY specified as *Ccn (n=19849) there are only 30% of records with EXAMPLE_UCUM_UNITS specified. If we take the subgroup CLASS=’CHEM’ only about 54% have EXMPLE_UCUM_UNITS specified. The absence of EXAMPLE_UCUM_UNITS information is even more important when SUBMITTED_UNITS is NULL. If SUBMITTED_UNITS is NULL and EXAMPLE_UCUM_UNITS is specified then one could use the EXAMPLE _UCUM_UNITS for mapping purposes (assuming there is not more than one UCUM example per record which unfortunately occurs in a number of instances, e.g., LOINC 2161-8). If both are NULL and EXAMPLE_SI_UCUM_UNITS is NULL as well that means any mapping will be ambiguous as far as UOMs is concerned.
I recognize that a fully specified LOINC code is not required to include UOM and that may be satisfactory to address HL7 message specifications and a certain level of semantic interoperability but I am concerned that lack of UOM specificity in LOINC codes limits the ability of LOINC to assist in comparing laboratory results when configuring EMRs to receive and segregate or integrate, as the case may be, results from more than one source laboratory. For example, if laboratory observations from more than one source laboratory can be displayed in a spread-sheet like format and for efficiency of review, trending, etc. it is desirable to integrate similar observations from two or more labs in one row then it becomes essential to accurately map observations on the basis of similar characteristics, e.g., component, property, scale and so on. This is where LOINC could be extremely helpful in the initial mapping and ongoing maintenance process. However, if test characteristics like UOM cannot be specified using a LOINC code then significant misrepresentation of observations can occur, e.g., two observations with different UOM on the same row. Thus one must resort to a more timely and labor intensive exchange of information between labs to achieve a reasonably accurate mapping of observations.
The issue of methodology and UOM seem to me to be critical to not only semantic interoperability for messaging but for all computational use of lab observations. One cannot assume that without knowledge of methodology and/or UOM that it would be appropriate to compare results from different laboratory sources for public health, clinical or other research purposes or for health outcomes research or institutional quality assurance.
Am I among the minority on this issue or are there other folks who share my concerns. What if anything can I do to help improve the current state of LOINC? Thank you in advance for your response.