Google Cloud healthcare data platform and FHIR APIs by Kalyan Pamarthy
1 like405 views
The document discusses the Google Cloud Healthcare API, which supports FHIR capabilities and accelerates healthcare solution development. It outlines various functionalities including data management, compliance with regulations, integration with Google Cloud applications, and examples of accessing and managing FHIR resources. Additionally, it highlights customer models and use cases demonstrating the benefits of using the API in clinical data management and machine learning applications.
![{
“resourceType”: “Patient”,
"id": "1e712685-714f-6720-a23a-c90103f70be6",
"identifier":[
{
"use":"usual",
...,
"value":""
}
],
"name":[
{
"use":"official"
"family":"",
"given":["", ""]
}
],
"gender": “unknown”,
...
}
{
"resourceType":"Patient",
"id": "1e712685-714f-6720-a23a-c90103f70be6",
"identifier":[
{
"use":"usual",
...,
"value":"12345",
}
],
"name":[
{
"use":"official",
"family":"Chalmers",
"given":["Peter", "James"]
}
],
"gender": "male",
"birthDate":"1974-12-25",
...
}
Example 2: De-identify - Before/After](https://image.slidesharecdn.com/fhirmeetupapril2020googlecloudhealthcareapi-200425030808/85/Google-Cloud-healthcare-data-platform-and-FHIR-APIs-by-Kalyan-Pamarthy-17-320.jpg)
Google Cloud healthcare data platform and FHIR APIs by Kalyan Pamarthy
- 1. Cloud Healthcare FHIR APIs FHIR Meetup April 23, 2020
- 2. Kalyan Pamarthy, Product Manager, Cloud Healthcare, Google kpamarthy@google.com https://www.linkedin.com/in/kpamarthy/
- 3. Topics for today ● Google Cloud Healthcare API & FHIR capabilities ● Google Cloud Perspectives on FHIR ● Example end-end solution on Google Cloud FHIR APIs Cloud Healthcare API
- 4. Google Cloud Healthcare API
- 5. Accelerate your healthcare solution development Healthcare API ● Primary API surface for healthcare specific data management capabilities. Standards based ○ FHIR (DSTU2, STU3 and R4) ○ Hl7v2 ○ DICOM ● Runs on Google Cloud Platform ● Serverless, REST interface ● Value add capabilities (e.g. De-Id) ● Stores PHI data in compliance with regulations (e.g. HIPAA) ● Integrated with Google Cloud applications (e.g. BigQuery) ● Integrated with Google Cloud components (e.g. PubSub) Cloud Healthcare API - Capabilities
- 7. Google Cloud Healthcare - FHIR APIs
- 8. FHIR as a data model FHIR as an API specification FHIR serves two technical roles in Healthcare IT
- 9. Project Cloud Healthcare API Location Dataset Store (FHIR Datastore) Organization: Project -> Location -> Dataset -> DataStore
- 10. Traditional database stores persist data to disk and offer a API, like Structured Query Language SQL HL7® FHIR With Cloud Healthcare FHIR APIs, you create FHIR data stores that implement healthcare-native APIs FHIR DataStores
- 11. Configuration on FHIR Store { "enableUpdateCreate": false, "disableReferentialIntegrity": true, "disableResourceVersioning": true, "notificationConfig": {"pubsubTopic": "<topic>"} }
- 12. REST Endpoints to manage FHIR Stores * On Our Radar Import Export Streaming* Bulk Streaming Bulk Interact Create Get List Delete Search De Identify FHIR Config Notify Pub/Sub Subscribe FHIR Store
- 14. https://healthcare.googleapis.com/<v>/projects/<P>/locations/<L>/datasets/<D>/<type>Stores/<S> <V> Healthcare API version v1 <P> Project identifier general-hospital <L> Location identifier us-central1 <D> Dataset identifier ehr-production <type> Data type slug hl7v2, dicom, fhir <S> Store identifier city-clinic Accessing FHIR resources with REST
- 15. Apps & services Example 1 : Get Patient data from a Cloud FHIR Store $> curl -X GET https://healthcare.googleapis.com/v1/projects/PROJECT_ID/locations/LOCATION/datasets/DATASET_ID/fhi rStores/FHIR_STORE_ID/Patient/1 -H 'Authorization: Bearer <access-token>' Cloud Healthcare FHIR API
- 16. Cloud Healthcare API curl -X POST -H "Authorization: Bearer $(gcloud auth print-access-token)" -H "Content-Type: application/json; charset=utf-8" --data "{ 'destinationDataset': 'projects/PROJECT_ID/locations/LOCATION/datasets /DESTINATION_DATASET_ID', 'config': { 'fhir': {} } }" "https://healthcare.googleapis.com/v1/projects/P ROJECT_ID/locations/LOCATION/datasets/SOURCE_DAT ASET_ID:deidentify" Cloud Healthcare API Example 2: De-identify an entire healthcare dataset
- 17. { “resourceType”: “Patient”, "id": "1e712685-714f-6720-a23a-c90103f70be6", "identifier":[ { "use":"usual", ..., "value":"" } ], "name":[ { "use":"official" "family":"", "given":["", ""] } ], "gender": “unknown”, ... } { "resourceType":"Patient", "id": "1e712685-714f-6720-a23a-c90103f70be6", "identifier":[ { "use":"usual", ..., "value":"12345", } ], "name":[ { "use":"official", "family":"Chalmers", "given":["Peter", "James"] } ], "gender": "male", "birthDate":"1974-12-25", ... } Example 2: De-identify - Before/After
- 18. BigQuery curl -X POST -H "Authorization: Bearer $(gcloud auth print-access-token)" -H "Content-Type: application/json; charset=utf-8" --data "{ 'bigqueryDestination': { 'datasetUri': 'bq://PROJECT_ID.DATASET_ID' } }" "https://healthcare.googleapis.com/v1/projects/PROJECT_ID/locations/L OCATION/datasets/DATASET_ID/fhirStores/FHIR_STORE_ID:export" Example 3: Generate a SQL projection of your FHIR data
- 19. SELECT observation.subject.reference.value AS patient_id, AVG(CAST(observation.value.quantity.value.value as FLOAT64)) AS avg_hr FROM <D>.observation o, UNNEST(observation.code.coding) c GROUP BY observation.subject.reference.value HAVING avg_hr > 130 ORDER BY avg_hr DESC Example 3: Run SQL queries on the FHIR projection in BigQuery
- 21. ● Have no existing infrastructure for managing clinical data ○ No on premises EHR to connect to ● Do not what to have to define a clinical semantic model ○ FHIR is a godsend as a primary CDR ● Want to write the minimum amount of new code ○ Expect standards & APIs to exist and leverage open source ● Do not want to have to operate or scale infrastructure ● Mobile first approach to applications ● Build machine learning in from day one ● Subject to the same regulatory controls ○ HIPAA, GDPR etc. ○ CMS & ONC interoperability proposals Clinical data is automatically in the right format and location to benefit from the force multiplication of Google Cloud Customer model 1 : “Born Cloud” Digital Healthcare Organizations
- 22. ● Have to address the EHR challenges ○ EHRs were not built to be a data platform servicing applications ○ Data is held in a proprietary format with low semantic interoperability ○ What FHIR APIs that do exist are “one way” and out of date ○ Can’t keep up with new data types (Omics, SDOH, Patient Generated) ○ Rarely only one EHR (average is 10) ○ Typically lots of decommissioned EHR data left after adoption ● BUT - EHRs are here to stay... ● Common answer is “Secondary Longitudinal CDR” ○ Hydrated from the on premises EHR data ○ Kept in near real time synch with EHR data Customer model - 2 : :Traditional” Healthcare Organizations
- 23. Building “Serverless” apps on Google Cloud FHIR APIs
- 24. Example : Putting The Pieces Together with Google Cloud From Data To Prediction : Customer scenario ● Have configured Google Healthcare API in the Google Cloud ● Have trained a “Risk Assessment” ML model which requires Heart Rate as one of its “features” ● Patient John Smith is having his vital signs monitored as an inpatient ● The EHR generates an Observation as HL7 message ● We want to store this as a FHIR resource ● We want to run the ML model against this new data ● We want to feedback ML model results into the EHR
- 25. Cloud Pub/Sub Healthcare API EHR Relay 4. Notify Write Observation Google Healthcare API- Putting The Pieces Together: Case Study 2. Write Observation as FHIR resource 3. Stores Observation Observation Heart Rate 1. HL7 message 5. Trigger Cloud Function 6. Read Patient$everything ML Engine 7. Predict 8. Prediction 9. Write RiskAssessment
- 27. Takeaways 1. Cloud Healthcare API :Managed service for FHIR 2. Accelerate healthcare solution development with Cloud Healthcare API and Google Cloud Ecosystem 3. Import, FHIR Operations, Export, De-Identification, AnalyticsCloud Healthcare API Get Started at cloud.google.com/healthcare