WIP: Synthea Data

apps/bfd-model/bfd-model-rif-samples/dev/design-sample-data-sets.md

@@ -449,3 +449,39 @@ It was created via the following process:
 2. Randomly select 11 beneficiaries from the synthetic data set. Modify their field values (only) as needed to correspond to the values required by MCT, as detailed on [NGD/MBP/BB/MCT Test Case Data](https://jira.cms.gov/browse/BFD-326).  Adjust identifiers `BENE_ID` so that they don't conflict/collide with any other data sets.
 3. For each of those beneficiaries, randomly select Part D events from the synthetic data sample-mct-update-5-pde.txt set to associate with them. Modify their field values (only) as needed to correspond to the values required by MCT, as detailed on [NGD/MBP/BB/MCT Test Case Data](https://jira.cms.gov/browse/BFD-326).
 4. Adjust all other identifiers (`BENE_ID`, `PDE_ID`, and `CLM_GRP_ID`) so that they don't conflict/collide with any other data sets.
+
+### `SYNTHEA`: Synthetic Data Generated by Synthea
+
+[Synthea](https://synthetichealth.github.io/synthea/) is an open source tool
+  for generating large volumes of realistic, but synthetic, health data.
+It was created and is maintained by [MITRE](https://www.mitre.org/),
+  a not-for-profit organization,
+  which operates federally funded research and development centers (FFRDCs).
+The best way to think of Synthea is like this:
+  "what if we paid for a metric ton of academic papers on disease prevalence, progression, etc.
+  and then paid a bunch of scientist-engineers to turn those papers into code
+  to generate realistic health data based on models derived from the papers?"
+That's Synthea: sure, it generates data,
+  but it's _how_ it generates the data that's the interesting, and tricky, part.
+
+Synthea models a wide range of populations, health conditions, etc.
+  and generates a similarly wide range of FHIR data,
+  including `Patient`s, and `ExplanationOfBenfit` resources.
+That's great, but not super useful for BFD: BFD _produces_ FHIR;
+  it doesn't need or _consume_ FHIR data as input.
+
+In 2020, CMS engaged Synthea to add an output mode
+  that produced data in BFD's RIF input file formats.
+The initial engagement ended with these accomplishments:
+
+* Synthea added a `--exporter.bfd.export=true` option to produce RIF.
+* It produces beneficiary RIF records:
+    * These are valid records with all 192 columns and convert to FHIR without errors.
+    * Most of those columns are optional. Synthea currently populates 21 of them.
+    * Of those populated columns, 16 appear to have useful, realistic data.
+* It produces inpatient claim RIF records:
+    * These are valid records with all 272 columns and convert to FHIR without errors.
+    * Most of those columns are optional. Synthea currently populates 52 of them.
+    * Of those populated columns, 25 appear to have useful, realistic data.
+* _Note: All column counts above are estimates produced by eyeballing the data;
+    we need to get more accurate counts from the Synthea folks._

apps/bfd-model/bfd-model-rif-samples/pom.xml

@@ -41,6 +41,14 @@
 			<artifactId>metrics-core</artifactId>
 		</dependency>
 
+		<dependency>
+			<!-- Used in builds/tests to run and manage external processes. (Can't
+				be marked as test-scoped, as the build/test code needs to live in src/main/java.) -->
+			<groupId>org.zeroturnaround</groupId>
+			<artifactId>zt-exec</artifactId>
+			<version>1.12</version>
+		</dependency>
+
 		<dependency>
 			<!-- Used to run our unit and integration tests. -->
 			<groupId>junit</groupId>