Differences

This shows you the differences between two versions of the page.

--- ivci:complete [2023/09/07 13:20] – [Ideal Measure] fkaag
+++ ivci:complete [2024/04/02 15:38] (current) – fkaag
@@ Line 5: / Line 5: @@
 ===== Rationale =====
 A universally applicable system ensures that any individual's vaccination history can be integrated, regardless of the origin.
-===== Initial Measure =====
+===== Measure =====
-The total number of codes needed to represent all existing legacy vaccines, plus any vaccine trail representing a class of vaccines rather than an actual product, is not known ex ante.
+If NUVA continues to be updated with codes to represent all vaccine concepts currently coded, then it could be used as a baseline for determining if a code set is complete.
-The most comprehensive among the existing code systems, NUVA (enriched continuously over the last decade and field proven at a large scale) currently includes 917 codes. Based on current trends, we can make an educated guess that the total number should be in the range of 1200 codes. We will use this value as the basis for the exhaustiveness indicator, as the percentage of the 1200 codes covered by a given code system.
+A code set is complete when it allows to represent accurately any existing vaccination record. "Accurately" could be interpreted as, at least, with the correct set of target diseases (otherwise a code system with a single code for "Any vaccine" would be considered as complete).
-===== Ideal Measure =====
+The approach could then be to consider within all identified possible vaccines records (that is, all NUVA codes) if some of them cannot be represented at all. For example, there is no ATC code to represent a Diphteria-Tetanus-Typhoid vaccine, although such a vaccine existed (it was administered to every young Frenchman during his military service).
-If NUVA continues to be updated with codes to represent all vaccine concepts currently coded, then it could be used as a baseline for determining if a code set is complete. To do this NUVA would need to identify a minimum set of codes, from the NUVA code system, which a complete vaccine code system should have a concept for. The concept in the target code system should be such that it can be mapped directly and unambiguously to a NUVA code and that NUVA code can be mapped back directly and unambiguously to that same code. Thus demonstrating that this code system has a direct one-to-one equivalent for the concept in NUVA. The measure would calculate the total number of NUVA minimum set codes that can be one-to-one represented in the code set being measured.
+Completeness differs from [[precise|precision]], that requires to be able to discriminate between concepts. A very limited code set could be quite complete, but highly unprecise. Reversely, a list of pharmaceutical codes used for logistics in a given country will be absolutely precise, but most likely very incomplete because it does not take into account historical vaccines nor vaccines administered in other countries.
-> FKa - I am not sure about the one-to-one mapping back from NUVA to the original code. For example, code systems that represent a substance and its packaging can be complete, while there will be several such codes for a given NUVA code (since NUVA does not take into account the packaging, that is useless for tracking received vaccines).
-Measure = [Total number of target code set values that can be mapped directly one-for-one to NUVA code in the denominator] / [Number of NUVA codes that are necessary for supporting a complete and accurate immunization history for any patient living today in any part of the world.]
+The Python program described at [[nuva-utils]] computes the values for completeness and precision, based upon an alignment table between a given code system and NUVA codes. The description pages for each code (see [[ivci:code_systems|Code systems]]) includes these values as computed from the currently known alignments.
-> FKa - When NUVA incorporates a mapping for an external code system, it is by construction a one-to-one from the original code to the NUVA code. Either it was identified that it matched a preexisting NUVA concept (e.g. the national pharmaceutical code for a vaccine that is already known from other countries) or a specific code is created and associated with the relevant valences (sometimes this will also request to create valences and position them in the valences tree). So the numerator corresponds to the number of codes in the code system.