NAMCS FHIR Solution

[MatthewVita]

PROBLEM

There is no robust open source NAMCS FHIR data viewer/export api solution available.

SOLUTION

Change this repository to store enhanced NAMCS data, translate it to FHIR resources, and serve it up via a FHIR server. Additional, the solution should support FHIR data export for testing FHIR operations in FHIR-capable EMRs (this will be useful for the OpenEMR project who is starting a FHIR integration project). The solution should be both useful for FHIR-capable EMR testing and classroom use.

TERMS

• MatthewVita/fake-patient-web (here) was originally intended to be an HTTP API (HTML and JSON) for serving up 300,000 NAMCS records ("Fake Patient Data as a Service"). It was using specially processed data from johnschrom/patient-data-generator. However, given recent trends in the industry, it would be more useful if it were FHIR capable and more data rich.
• amesagnew/hapi-fhir (https://github.com/jamesagnew/hapi-fhir/tree/master/hapi-fhir-jpaserver-example) is a module that can be used to create a FHIR server endpoint against an arbitrary data source. This will power FakePatientWeb.com using enhanced NAMCS data FHIR resources.
• smart-on-fhir/sample-patients (https://github.com/smart-on-fhir/sample-patients) is a SMART on FHIR repository that generates FHIR resource files given text documents.
• johnschrom/patient-data-generator (https://github.com/johnschrom/patient-data-generator) is a program that generates structured CSV data from the NAMCS dataset.
• johnschrom/ambulatory-ehr-data (https://github.com/johnschrom/ambulatory-ehr-data) is a program that takes the structured CSV data from johnschrom/patient-data-generator and puts it into an SQLITE database.

MAPPINGS

By executing python generate.py --summary 2169591, relationships in the smart-on-fhir/sample-patients data are easily discoverable because this patient has a bunch of information, documents, labs, etc. While the NAMCS data from johnschrom/patient-data-generator is well sampled and correct in terms of the relationships, there's not enough mappings for the SMART FHIR data requirements. Problems include:

• Lack of modern coding standards such as SNOMED and RxNorm
• Lack of certain fields (These are the available fields ftp://ftp.cdc.gov/pub/Health_Statistics/NCHS/dataset_documentation/namcs/sas/nam13inp.txt)

These issues stem from NAMCS limitations.

TASKS

Research

• Test out smart-on-fhir/sample-patients
• Test out deploying amesagnew/hapi-fhir RestfulServer.
• Hook up Docker image to use MySQL per this tutorial: https://www.openhealthhub.org/t/howto-build-a-health-database-and-fhir-api-server-in-15-mins-using-open-source/155

Initial Data

• Generate 2 NAMCS samples nationally using johnschrom/patient-data-generator.
• Manually translate the provided ICD9 and Multum Lexicon Plus codes into SNOMED and RXNORM.
• Manually fill in FHIR data mapping gaps and put together semi-realistic patient histories including clinical notes.
• Transfer samples into the format that smart-on-fhir/sample-patients expects.
• Run smart-on-fhir/sample-patients generator against samples.
• Manually insert resources into RestfulServer.

Test Deployment

• Purchase FakePatientWeb.com and host the samples with smart-on-fhir/api-server server.
• Build up custom HTTP GET endpoint that takes in a patient ID that generates relevant FHIR CRUD messages that can be sent to an EMR FHIR endpoint.

Production Dataset

• Generate 100 NAMCS samples nationally using johnschrom/patient-data-generator.
• Programmatically translate the provided ICD9 and Multum Lexicon Plus codes into SNOMED and RXNORM.
• Work with various physicians in the open source healthcare space to hand-craft realistic patient histories around these 100 patients. This Includes clinical notes and other more advanced clinical artefacts.

Production Deployment

• Redeploy FakePatientWeb.com with the new samples.

Follow-up Work

• Work with the ONC to list the FHIR server an official test server.
• Market the solution to see if more providers would be interested in enhancing NAMCS data for samples.

[MatthewVita]

After some time, it makes the most sense to stick with https://github.com/smart-on-fhir/sample-patients and the HAPI FHIR solutions. NAMCS is missing too many data elements to pull this off.