Exploring conversation of Consolidated CDA to JSON for web apps
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
app
config
fixtures/keys
lib
public
py
samples
smart-apps
tasks
test
value_sets
.gitignore
README.md
cloudfoundry.json
launch.js
npm-shrinkwrap.json
package.json
server.js

README.md

Exploratory mapping of Consolidated CDA data to JSON, with a goal of preserving CCDA's data model (but applying friendly names to properties whenever possible).

Inspired by popHealth's approach to mapping C32 (+ more Linked Data flavor)

For automated setup, see https://github.com/jmandel/ansible-ccda

Manual Setup

$ git clone https://github.com/chb/json_ccda
$ cd json_ccda
$ npm install
$ cd smart-apps
$ npm install

Launching API server:

$ node launch.js

Launching App server:

$ cd son_ccda/smart-apps
$ node launch.js

Convert a CCD to JSON, without loading into DB:

$ node lib/ccda/import.js -f CCD.sample.xml -p 123  --stdout > sample_ccda.json

Convert and load into MongoDB:

$ node lib/ccda/import.js -f CCD.sample.xml -p 123 --mongo-db

Load a CCDA via HTTP POST:

$ curl -X POST -d @CCD.sample.xml http://localhost:3000/patients/123/documents/ccda

In Browser: `http://localhost:3000/patient/12345/

Testing:

$ ./node_modules/.bin/mocha -R spec

Lots of outstanding mapping questions, including:

  • How to deal with negation indicators? (Would like separate 'AllergyNegation' items, rathern than just a flag)

  • How to deal with nullFlavors? (Currently dropping for simple attributes; this seems okay, otherwise JSON gets very cumbersome. Will want to treat nullFlavor differently in special cases, like problem codes that weren't mappable to SNOMED / ICD10)

  • How to effectively deal with elements that have multiple IDs (should be able to lookup an item by any of its IDs; possibly just receiving a list of URIs for that item).

  • How to map document and entry IDs to stable URIs (considering using three hierarchies in paralle: documents, sections and entries corresponding fairly closely to these concepts in CCDA; except an entry will be constrained.)

  • How to assign IDs to sections (which have none in CCDA)

TODO:

  • Get coding systems in DB (out of flat file)
  • POST receiver to write JSON to DB (current output is files only)
  • Example Mongo queries (e.g. "All patients w/ abnormal lab results since yesterday")

[x] Consider a separate 'links' collecition to offload many:many relations. Keep the fact documents cleaner. But still need patient links for filtering without joins... (or can actually do this by convention + regix on _id:regex:^http:.../patient/123 (but implies we need a slash or huffman-coded pids. patient link is explicit -> better.)

[ ] Batch request mode to issue multiple individual-URI GETs in a swell foop. [ ] Standalone and then integrated Validation of expected coding systems / constraints (explicit valuesets, snomed subsumptions, ucum, ?countries) [ ] dates should be live DB dates and serialized in the REST server. [ ] Use Mongolabs' syntax for queriese (across the board!) [ ] use async.auto or similar to handle import.js flag interactions. [ ] strip sourceIDs from the JS rep andn include only in links... ? [ ] apply consistent units at import-time for vital signs? --> consider code + unit mapping in the import layer, or pushing out to apps

[ ]how to handle results that don't really belong to panels? all-in-one, or one-per?

[ ] Add UCUM validation to physical quantities [ ] Dynamically subclass ConceptDescriptor / SimpleCode to declare valid or obligatory valuesets. e.g. ["productName","0..1", ".//h:manufacturedMaterial/h:code", ConceptDescriptor.shall({valueSet: "1.2.3"})]

[ ] Sweep over the entire tree for validations to bulk-execute.

[x] email struct doc WG to ask about IDs:

1.  Stable over time for an entry in a medicaiton list, or problem list, or vital sign?
2.  Stable over time of an entry /changes/ (e.g. I upgrade asthma by adding a severity)?
3.  Stable over time over change in status (inactive -> active) 
4.  Are there current *actual* implementations that assign multiple identifiers to entries (meds, problems, labs; not people)

Data questions:

If normalizing displayNames, lincensing terms ?require the value come from the SDO?

  • "doseQuantity, if present, SHOULD contain zero or one [0..1] @unit="1", which SHALL be selected from ValueSet UCUM Units of Measure (case sensitive) 2.16.840.1.113883.1.11.12839 DYNAMIC (CONF:8842)." -- how does one select the value "1"? How is UCUM involved?

  • what is manufacturerOrganization or a manufacturedProduct? No datatype in RIM or CDA or CCDA

(datatypes should be copied over from CDA spec! This is a serious usability issue --> easy errors)

  • Severity free-text lives in a "text" element while medication free-text and immunization free-text are in a "code's" "original."

BP centiles app: will have to leave out encounterType. (Semantics for vitals in CCDA are only implicit -- which encoutner did a given value come from? who was reponsible for it? etc.)

no consistent codes required for vital signs

What is the meaning of a vitals organizer? does it imply that systolic and diastolic are matched, if they appear in the same organizer?

reported:

  • What does it mean when a text has text and a reference, and they both are non-emtpy and non-identical?

reported:

  • ActReason code in examples is 2.16.840.1.113883.11.19725 ; but in the published table it is listed as 2.16.840.1.113883.5.8 . Appears examples are wrong in the CCDA 1.1 spec + accompanying XML files.

reported:

  • what is a Medication Activity's or Immunization Activity's "code" element? (CDA and RIM don't seem to say -- different from routeCode, or approachSiteCode).