Handling documents with an FHIR data store

Now let’s see what happens when our input for HealthLake is not in FHIR format, but is a document instead. Amazon HealthLake currently only supports data in FHIR format. What if we need to process document-based health data along with FHIR data? Can we still create a centralized scalable FHIR data store for health data?

And the answer is yes. We can use Amazon Textract to get raw text from the document and convert it to an FHIR resource (a DocumentReference resource). This DocumentReference FHIR resource can then be input into Amazon HealthLake with additional FHIR resources.

This is a three-step process:

1. Extracting data from the document with Amazon Textract
2. Creating a DocumentReference FHIR resource from the extracted Textract response
3. Ingesting the DocumentReference FHIR resource to Amazon HealthLake

You can see the architecture in the following figure:

Figure 7.15 – Document...