nbolser / lets-build

HAPI-FHIR Let's Build

Precursors

• Check this project out to your local hard drive.
• Open it in an IDE
• Have the following tools installed:
  • Java JDK Version 11.x.x (This is the recommended version but Java 8 and above should also work)
  • Apache Maven

Hour 1 - Build a FHIR Data Mapper

Rationale: When adopting FHIR, a common scenario is needing to convert your existing data into the FHIR model. This can be a challenging first step, but if you approach it systematically it can be easy.

Exercise: For this exercise, we will be building a mapper that converts existing data a CSV file into FHIR Patient and Observation resources.

We'll be using a sample data file from the CDC NHANES (National Health and Nutrition Examination Study) publicly available sample data set. The format of the data set is described at this link but we have reworked the format a bit to add fake patient identities and timestamps to the data.

The input CSV file can be found here: https://github.com/hapifhir/lets-build/blob/main/src/main/resources/sample-data.csv

Approach

The input data looks like the following:

SEQN   , TIMESTAMP               , PATIENT_ID, PATIENT_FAMILYNAME, PATIENT_GIVENNAME, PATIENT_GENDER, WBC, RBC, HB
93704.0, 2020-11-13T07:47:35.964Z, PT00002   , Simpson           , Marge            , F             , 7.4, 0.1, 13.1

Note the columns:

• SEQN: This is a unique identifier for the test
• TIMESTAMP: This is the timestamp for the test
• Patient detail columns (note that the patients repeat so you will want to ):
  • PATIENT_ID: This is the ID of the patient
  • PATIENT_FAMILYNAME: This is the family (last) name of the patient
  • PATIENT_GIVENNAME: This is the given (first) name of the patient
  • PATIENT_GENDER: This is the gender of the patient
• Test result columns (each of these will be a separate Observation resource):
  • WBC: "White Blood Cell Count": This is a count of the number of white blood cells in your blood (These cells deliver oxygen)
  • RBC: "Red Blood Cell Count": This is a count of the number of red blood cells in your blood (These cells are an important part of the immune system)
  • HB: "Hemoglobin": This a measurement of the amount of hemoglobin protein in your blood (this iron-rich protein carries oxygen, among other things)

Writing a Mapper

• Open the following class: ca.uhn.fhir.letsbuild.upload.CsvDataUploader.

• Notice how this class has a basic skeleton that reads in a CSV file.

Hints:

• Don't optimize at first! There are lots of ways this code can be made efficient, but it's better to start by getting it working, then worry about performance later.

• You can use the PATIENT_ID column as the resource ID for the Patient resource. Using a client-assigned ID for your resources makes the outcome predictable, and means that you can re-run your uploader as many times as you like without constantly creating new datasets.

• You can use the SEQN column as the resource ID for the Observation resources, but you will need to append something to it since there are 3 Observations per row.

Hour 2 - Upload the Resources to a FHIR Server

• Create a FHIR Client:

  • Using the instructions here: https://hapifhir.io/hapi-fhir/docs/client/generic_client.html
  • Use the public hapi server URL: http://hapi.fhir.org/baseR4

• For each row, create the Patient resource. Use an Update (PUT) operation so that you can control the ID of the patient resource. For example, you might want Marge Simpson to have a resource ID of "Patient/PT00002"

• For each row, create 3 Observation resources, one for each of the 3 tests.

• validate the resources against the server with your client

• validate the resource using the hapi validation stack: https://hapifhir.io/hapi-fhir/docs/client/generic_client.html#extended-operations (4.2.11Extended Operations)

Bonus - Secure Your FHIR Server

In this exercise we will add simple authentication and authorization to our server. To keep things simple:

• We will secure only read operations, and we'll leave write operations unsecured. Obviously this isn't a setup you would use for real, but all of the read security principles work exactly the same for writing and can be used for those operations too.

• We will hardcode our credentials in the server code. This is also not a practice you should ever use in a real system, but it makes it easy to show the concepts and can be replaced with something more robust as needed.

Approach

Open the class ServerAuthorizationInterceptor and