Add Support for Composite Measure Calculation

[mgramigna]

Summary

The moment we’ve all been waiting for - composite measure support is here! Adds the ability to run measure calculation on a composite measure, and adds the ability to generate a composite measure report.

New behavior

Composite measure calculation

fqm-execution can now calculate detailed results for a composite measure. To do so, a measure bundle must be passed in that contains the composite measure, all component measures, and all their dependencies. For each patient, calculates detailed results for each of the components.

Composite measure report building

fqm-execution can now generate a summary measure report. Running reports with a passed-in measure bundle containing a composite will calculate the measure score based on the composite scoring type that is provided. All-or-nothing, opportunity, and linear are currently supported.

Code changes

Composite measure calculation

• When it comes to Calculator.ts, the best approach for seeing code changes is to go through the file directly rather than looking through the code diff
• Summary of calculation changes are as follows:
  • Check if a composite measure is contained in the bundle - if so, extract all components, otherwise proceed as normal
  • Execution is now performed within a loop over the measure(s)
  • When making results for the patients, in the composite case, we search if execution results already exist for the patient so that we can append to them - each patient’s execution results contain the detailed results for each component
  • In the component case, we append componentResults onto the patient execution results - this is a simplified version of the detailed results that we can use for report building that gives us access to the groupId, componentCanonical, and populationResults
  • Group clause coverage and overall clause coverage are only created for non-composite measures

Composite measure report building

• New abstract class AbstractMeasureReportBuilder. Having this abstract class allows us to cleanly separate the composite functionality from the standard measure report building that we’ve been doing.
• New class CompositeReportBuilder that extends the abstract class:
  • creates a single group to store the numerator and denominator counts - these get populated when getting the report. When looping through patient results, the numerator and denominator in this.compositeFraction get incremented
  • addPatientResults increments the numerator and denominator of this.compositeFraction according to the composite scoring type defined on the composite report builder. all-or-nothing, opportunity, and linear are currently supported.

New integration tests

• Update Makefile to generate an all-or-nothing composite measure that consists of two components —> we restricted integration tests to all-or-nothing because the composite scoring type only affects the measure report content, not detailed results
• Added corresponding integration tests for all-or-nothing composite support

New unit tests

• Unit tests for composite report builder class, report builder factory
• New simple composite measure fixture
• Tests for new composite measure bundle helpers

Other changes

• README - Updated documentation on composite measure calculation support and how to interpret the calculation results
• Updated cql-execution to 3.0.0
• Updated ecqm-bundler to 0.4.1
• The standard MeasureReportBuilder no longer takes in the measure bundle as input - just takes in the measure
• Support for specifying a patient directory as a CLI option instead of needing to list out the path to each patient bundle

Testing guidance

1. Run integration tests for all-or-nothing composite
2. Run unit tests for composite report builder class, changes to standard measure report builder, composite measure bundle helpers, etc.
3. Test composite report building across the three different composite scoring types. This can be done by changing the composite scoring code on the measure.

[github-actions[bot]]

Coverage report

St.:grey_question:	Category	Percentage	Covered / Total
🟢	Statements	86.41% (+0.56% 🔼)	2174/2516
🟡	Branches	75.04% (+0.59% 🔼)	1975/2632
🟢	Functions	88.91% (+0.87% 🔼)	393/442
🟢	Lines	86.74% (+0.57% 🔼)	2099/2420

Show new covered files 🐣

|

St.^{:grey_question:}

| File | Statements | Branches | Functions | Lines | | :----------------------------------------------------------------------------------------------------------------------------------------------------: | :----------------------------------------------------------------------------------------------------- | :--------- | :------- | :-------- | :----- | | 🔴 |

`...` / AbstractMeasureReportBuilder.ts

| 36.36% | 0% | 50% | 33.33% | | 🟢 |

`...` / reportBuilderFactory.ts

| 100% | 100% | 100% | 100% | | 🟢 |

`...` / CompositeReportBuilder.ts

| 100% | 80% | 100% | 100% |

Show files with reduced coverage 🔻

|

St.^{:grey_question:}

| File | Statements | Branches | Functions | Lines | | :----------------------------------------------------------------------------------------------------------------------------------------------------: | :--------------------------------------------------------------------------------- | :------------------------------------------------------------- | :------------------------------------------------------------- | :-------- | :------------------------------------------------------------- | | 🟢 |

`...` / MeasureBundleHelpers.ts

|

98.45% (-0.91% 🔻)

|

90.65% (-1.82% 🔻)

| 100% |

98.88% (-0.43% 🔻)

|

Test suite run success

408 tests passing in 31 suites.

Report generated by 🧪jest coverage report action from ad5ecba3b5cc3ba620c0cd5b0aa2bd104da41105