search

EBISPOT / efo

Github repo for the Experimental Factor Ontology (EFO)
https://www.ebi.ac.uk/efo/
57 stars 13 forks source link

Address problematic mappings for ICD10 and self-reported traits #396

Closed paolaroncaglia closed 5 years ago

paolaroncaglia commented 5 years ago

We have a list of 64 ICD10 codes and patients' self-reported traits for which we haven't found acceptable mappings. We are discussing them with submitters. To keep track of comments and action items for self/EFO editors, I'll keep the list in a spreadsheet and add some notes there (https://docs.google.com/spreadsheets/d/1edkZgAom9bmr2TbZffCda6DYiAuwi8oadl8kPLLKtCA/edit#gid=0).

I added a tab where I sorted traits by their code (i.e. ICD10 code or self-reported "ID"). Note that trait 4195, and ICD10 codes M48 and M53, were duplicated, so I collapsed all related information into 1 row for each.

Whatever edits we do, in the end remember to

paolaroncaglia commented 5 years ago

@gkos-bio (cc @zoependlington ) Could you please help with the following (re. some neoplasm nomenclature I'm not super-confident about). Among the last problematic ICD10 mappings, we have: Secondary malignant neoplasm of respiratory and digestive organs (ICD10:C78) and Secondary malignant neoplasm of other sites (ICD10:C79) We initially thought we should classify these as related to 'metastatic neoplasm', but Wikipedia suggest differently: "Secondary malignant neoplasm is a malignant tumor whose cause is the treatment (usually radiation or chemotherapy) which was used for a prior tumor. It must be distinguished from Metastasis from the prior tumor or a relapse from it since a secondary malignant neoplasm is a different tumor." (https://en.wikipedia.org/wiki/Secondary_malignant_neoplasm) If that's correct, then we should create a new EFO term for 'secondary malignant neoplasm'. Could you please confirm? On the other hand, Wikipedia also says: "Secondary neoplasm refers to any of a class of cancerous tumor that is either a metastatic offshoot of a primary tumor, or an apparently unrelated tumor that increases in frequency following certain cancer treatments such as chemotherapy or radiotherapy." (https://en.wikipedia.org/wiki/Neoplasm#Types). Plus, ICD10 classifies C78 and C79 under 'Malignant neoplasms of ill-defined, other secondary and unspecified sites', and I can't really tell if that refers to metastases or not; if the origin of the tumor is unclear, then perhaps we should map C78 and C79 to generic 'neoplasm'? Thanks. [Note for self: followed up with Gautier (and Ed) by email on 1/4/19, and with Gautier on 9/4.]

After consulting with Gautier, I resolved to:

paolaroncaglia commented 5 years ago

@zoependlington (cc @simonjupp ) The list of pending problematic ICD10 mappings includes 15 ICD10 Z codes, all "children" of the ICD10 category 'Factors influencing health status and contact with health services' (see tab named "ICD10 Z codes" in https://docs.google.com/spreadsheets/d/1edkZgAom9bmr2TbZffCda6DYiAuwi8oadl8kPLLKtCA/edit#gid=260497166, where I grouped the entries based on their ICD10 subcategory). Gautier suggested that it’s up to us to decide if we wish to create new categories in EFO3 for that. These entries don't have appropriate mappings in external ontologies, so I lean towards Gautier's suggestion. The ICD10 Notes for 'Factors influencing health status and contact with health services' (Z00-Z99) say: "Z codes represent reasons for encounters. A corresponding procedure code must accompany a Z code if a procedure is performed. Categories Z00-Z99 are provided for occasions when circumstances other than a disease, injury or external cause classifiable to categories A00-Y89 are recorded as 'diagnoses' or 'problems'. This can arise in two main ways: (a) When a person who may or may not be sick encounters the health services for some specific purpose, such as to receive limited care or service for a current condition, to donate an organ or tissue, to receive prophylactic vaccination (immunization), or to discuss a problem which is in itself not a disease or injury. (b) When some circumstance or problem is present which influences the person's health status but is not in itself a current illness or injury." So I suggest to

Any objection please?

paolaroncaglia commented 5 years ago

@gkos-bio @zoependlington Among the pending problematic mappings, we have the following self-reported traits: Number of self-reported cancers Number of self-reported non-cancer illnesses Number of operations, self-reported Number of treatments/medications taken We thought we could create measurement terms for these, or similar to children of NCIT:C25337 'number', but Gautier wrote "I think there is an issue with ‘number of…’ because I don’t think we even have access to the value (...) so no way to summarize this. However, they all belong to ‘self-reported traits.’" I'd be concerned about creating a term for 'self-reported trait' because some of its children may violate the subclass-of rule. E.g. 'Number of treatments/medications taken' may be self-reported but also recorded by a nurse. The solution would be to append "self-reported" to all labels and to state it clearly in the definition too. The parent 'self-reported trait' could be a subclass of 'quality'. But would these terms be used at all if their value can't be accessed? At least at this stage? Thanks.

Following email exchange with Gautier and Ed (on 1/4/19), I will

paolaroncaglia commented 5 years ago

@zoependlington Please let me know if you have any objection: Re. ‘pulse wave reflection index’ (self-reported field code = 4195): In the original spreadsheet (see my editable copy here https://docs.google.com/spreadsheets/d/1sBlgfl4BlioLmD_ecNg-3hTeyhTl1l2ULIwHz3tKqS0/edit#gid=785509692), there is a note: “reflection index is a measure of reflection A digital volume pulse can be considered to be the summation of a direct and a reflected component. The direct component is due to transmission of a pressure wave from the left ventricle to the finger via the most direct route. The reflected component is formed by pressure transmitted from the heart to the lower body where it is reflected back up the aorta and thence to the finger. if 'b' is the height of direct wave peak and 'a' is the height of reflected wave peak the reflected index = (a/b) * 100%” (this trait is probably measured via a wearable monitoring device). And the initial, non-curated Zooma mapping (score = medium, confidence = none, no other comments) went to EFO_0004724 ‘carotid-femoral pulse wave velocity’. I suggest to

paolaroncaglia commented 5 years ago
paolaroncaglia commented 5 years ago

As for the obscure trait called "Other_fluke_infection", in the original data this was mapped to ICD10 B66, which is a child of 'Helminthiases'. So we should

paolaroncaglia commented 5 years ago

For the trait named "Volume_depletion":

paolaroncaglia commented 5 years ago

For the trait named 'Intrauterine hypoxia': This is actually the same as EFO:1001793 fetal hypoxia, which should therefore be added

paolaroncaglia commented 5 years ago

For the ICD10 trait named 'Other deforming dorsopathies': This is a "child" of ICD10 'Deforming dorsopathies'; all its named siblings are disorders of the spine or joints; therefore, types of skeletal system disease. So,

paolaroncaglia commented 5 years ago

For the ICD10 traits named 'Other inflammatory spondylopathies', 'Other spondylopathies' and 'Spondylopathies_in_diseases_classified_elsewhere': These are all children of ICD10 'Spondylopathies', which are disorders of the vertebrae (not of the vertebral joints), therefore types of bone disease. So,

paolaroncaglia commented 5 years ago

For the ICD10 trait named 'Other dorsopathies, not elsewhere classified': dorsopathies are disorders of the (generic) back or spine (see e.g. https://en.wiktionary.org/wiki/dorsopathy), so

paolaroncaglia commented 5 years ago

For the ICD10 trait named 'Shoulder_lesions': a shoulder includes several bits of different anatomical nature (see http://purl.obolibrary.org/obo/UBERON_0001467), but the ICD10 trait is child of 'Other soft tissue disorders', so

paolaroncaglia commented 5 years ago
paolaroncaglia commented 5 years ago

For the ICD10 traits named 'Other abnormal products of conception', 'Maternal care for other conditions predominantly related to pregnancy', 'Maternal_care_for_known_or_suspected_malpresentation_of_fetus', 'Maternal_care_for_known_or_suspected_disproportion', 'Maternal_care_for_known_or_suspected_abnormality_of_pelvic_organs', 'Maternal care for known or suspected foetal abnormality and damage', 'Maternal care for other known or suspected foetal problems', 'Other_obstetric_trauma': And also for 'Long_labour', because I can't find a medical reference measurement as to what classifies as a "long" labour:

paolaroncaglia commented 5 years ago

For the self-reported trait named 'Number of children fathered':

paolaroncaglia commented 5 years ago

For the self-reported trait named 'Cascot confidence score measurement':

paolaroncaglia commented 5 years ago

For the ICD10 trait named 'Perineal_laceration_during_delivery':

For the ICD10 trait named 'Complications_of_anaesthesia_during_the_puerperium':

Also,

paolaroncaglia commented 5 years ago

For the missing self-reported traits:

paolaroncaglia commented 5 years ago

For the missing ICD10 R code mappings:

TO BE CONTINUED