search

WhiteCoatAcademy / icd10

Dead simple, incredibly fast ICD-10 diagnosis code searching.
https://icd10.clindesk.org/
Other
26 stars 7 forks source link

Data structure 2.0 #3

Open leoshmu opened 10 years ago

leoshmu commented 10 years ago

We need a revamp of the data structure to support a variety of things:

If we are doing ICD10->ICD9 conversion, or the other way around, what is the data structure. Like a hash table, where icd10 codes are the hash and icd9 is the match?

{e110: {icd9 code for e110}, e119: {icd9 code for e119}}

If we are doing keyword deduplication, where does that data go?

{code: "", keyword_array: ["",""], other keys?}

Please create a spec for the kinds of data that you will make available for the client to use! Then I can wire them up and have the site be ready for the full data

semenko commented 10 years ago

Hope to do tonight!

I'm not sure about the 9/10 usefulness, or how we'll fit that in a search.

As for keyword mapping/dedupe, I was thinking something like:

(Filtering out common words, e.g. with without like and or disease, etc.)

parent: {code:"E101", description: "Type 1 diabetes mellitus", children:[E1010, E1011, E1012] keywords: "type 1 diabetes mellitus ketoacidosis coma diabetic nephropathy chronic kidney complication retinopathy macular edema"}

children: {E1010: {description: "T1dm with ketoacidosis coma}, E1011: {description: T1dm with diabetic nephropathy}}

leoshmu commented 10 years ago

Sweet. In terms of the 9/10 use case, I was thinking that if we are providing a popup of detialed info regarding a parent's children, we could provide conversion info there too. Usefullness of conversion info would come up for practices that have established protocols for a given code in icd9, and are wondering what the corresponding 10 code is. Going the other way is probably less useful, so if this is a searchable thing for icd10 codes only, perhaps the conversion is not all that useful. The only thing could be to remind people as they go through and explore icd10, what the more familiar icd9 codes are and how they map. I suspect the answer is that icd9 maps well to parents but is too coarse grained to map well to the end nodes of the icd10 tree

On Tue, Apr 15, 2014 at 5:04 PM, Nick Semenkovich notifications@github.comwrote:

Hope to do tonight!

I'm not sure about the 9/10 usefulness, or how we'll fit that in a search.

As for keyword mapping/dedupe, I was thinking something like:

(Filtering out common words, e.g. with without like and or disease, etc.)

parent: {code:"E101", description: "Type 1 diabetes mellitus", children:[E1010, E1011, E1012] keywords: "type 1 diabetes mellitus ketoacidosis coma diabetic nephropathy chronic kidney complication retinopathy macular edema"}

children: {E1010: {description: "T1dm with ketoacidosis coma}, E1011: {description: T1dm with diabetic nephropathy}}

— Reply to this email directly or view it on GitHubhttps://github.com/WhiteCoatAcademy/icd10/issues/3#issuecomment-40540264 .

semenko commented 10 years ago

Agreed. I'll look more at the icd9<->10 stuff in a bit. Code almost done for new json structure