11.3.1Measures

11.3.1.1Introduction

The FHIR Clinical Reasoning Module defines the Measure resource and several associated operations. The Measure Resource represents a structured, computable definition of a health-related measure such as a clinical quality measure, public health indicator, or population analytics measure. These Measures can then be used for reporting, analytics, and data-exchange purposes.

Electronic Clinical Quality Measures (eCQMs) in FHIR are represented as a FHIR Measure resource containing metadata and terminology, a population criteria section, and at least one FHIR Library resource containing a data criteria section as well as the logic used to define the population criteria. The population criteria section typically contains initial population criteria, denominator criteria, and numerator criteria subcomponents, among others. This is elaborated upon in greater detail in the CQF Measures IG. An example of an eCQM as defined in FHIR looks like:

{
  "resourceType" : "Measure",
  "library" : [
    "http://hl7.org/fhir/us/cqfmeasures/Library/EXMLogic"
  ],
  "group" : [
    {
      "population" : [
        {
          "code" : {
            "coding" : [
              {
                "code" : "initial-population"
              }
            ]
          },
          "criteria" : {
            "language" : "text/cql-identifier",
            "expression" : "Initial Population"
          }
        },
        {
          "code" : {
            "coding" : [
              {
                "code" : "numerator"
              }
            ]
          },
          "criteria" : {
            "language" : "text/cql-identifier",
            "expression" : "Numerator"
          }
        },
        {
          "code" : {
            "coding" : [
              {
                "code" : "denominator"
              }
            ]
          },
          "criteria" : {
            "language" : "text/cql-identifier",
            "expression" : "Denominator"
          }
        }
      ]
    }
  ]
}

Measures are then scored according the whether a subjects (or subjects) are members of the various populations.

For example, a Measure for Breast Cancer screening might define an Initial Population (via CQL expressions) of "all women", a Denominator of "women over 35", and a Numerator of "women over 35 who have had breast cancer screenings in the past year". If the Measure is evaluated against a population of 100 women, 50 are over 35, and of those 25 have had breast cancer screenings in the past year, the final score would be 50%¹ (total number in numerator / total number in the denominator).

1. There are several methods for scoring Measures, this is meant only as an example.

11.3.1.2Operations

HAPI implements the $evaluate-measure operation. Support for additional operations is planned.

11.3.1.3Evaluate Measure

The $evaluate-measure operation is used to execute a Measure as specified by the relevant FHIR Resources against a subject or set of subjects. This implementation currently focuses primarily on supporting the narrower evaluation requirements defined by the CQF Measures IG. Some support for extensions defined by other IGs is included as well, and the implementation aims to support a wider range of functionality in the future.

Several example Measures are available in the ecqm-content-r4 IG. Full Bundles with all the required supporting resources are available here. You can download a Bundle and load it on your server as a transaction:

POST http://your-server-base/fhir BreastCancerScreeningFHIR-bundle.json

These Bundles also include example Patient clinical data so once posted Measure evaluation can be invoked with:

GET http://your-server-base/fhir/Measure/BreastCancerScreeningFHIR/$evaluate-measure?periodStart=2019-01-01&periodEnd=2019-12-31&subject=numerator&reportType=subject

parameter "Measurement Period" Interval<DateTime>
  default Interval[@2019-01-01T00:00:00.0, @2020-01-01T00:00:00.0)

GET fhir/Measure/<MeasureId>/$evaluate-measure?periodStart=2019-01-01&periodEnd=2019-12-31

periodStart and periodEnd support Dates (YYYY, YYYY-MM, or YYYY-MM-DD) and DateTimes (YYYY-MM-DDThh:mm:ss). DateTime formats of YYYY-MM-DDThh:mm:ss+zz no longer accepted. To pass in timezones to period queries, please see the Headers section below:

Accepted values for this header are documented on the Wikipedia timezones page

ex: Timezone:America/Denver will set the timezone to Mountain Time.

If the client omits this header, the timezone will default to UTC.

Please consult the below table for examples of various combinations of start, end, and timezone, as well as the resulting queried periods:

GET fhir/Measure/<MeasureId>/$evaluate-measure?reportType=subject-list

GET fhir/Measure/<MeasureId>/$evaluate-measure?subject=Patient/123

GET fhir/Measure/<MeasureId>/$evaluate-measure?practitioner=Practitioner/XYZ

The Measure scoring method determines how a Measure score is calculated. It is set with the scoring element on the Measure resource.

The HAPI implementation conforms to the requirements defined by the CQF Measures IG. A more detailed description of each scoring method is linked in the table below.

• The cohort Measure scoring support is partial. The HAPI implementation does not yet return the required Measure observations

An example Measure resource with scoring defined looks like:

{
  "resourceType": "Measure",
  "scoring": {
    "coding": [ {
      "system": "http://terminology.hl7.org/CodeSystem/measure-scoring",
      "code": "proportion",
      "display": "Proportion"
    } ]
  }
}

A composite Measure is scored by combining and/or aggregating the results of other Measures. The compositeScoring element is used to control how composite Measures are scored. HAPI does not currently support any composite scoring method.

The HAPI implementation uses the populations defined by the CQF Measures IG for each scoring type. A matrix of the supported populations is shown in the Criteria Names section of the CQF Measures IG.

The logical criteria used for determining each Measure population is defined by the Measure.group.population.criteria element. The Measure specification allows population criteria to be defined using FHIR Path, CQL, or other languages as appropriate. The HAPI implementation currently only supports using CQL. The relationship between a Measure Population and CQL is illustrated in the Population Criteria section of the CQF Measures IG.

An example Measure resource with a population criteria referencing a CQL identifier looks like:

{
  "resourceType": "Measure",
  "group": [ {
    "population": [ {
      "code": {
        "coding": [ {
          "system": "http://terminology.hl7.org/CodeSystem/measure-population",
          "code": "initial-population",
          "display": "Initial Population"
        } ]
      },
      "criteria": {
        "language": "text/cql-identifier",
        "expression": "Initial Population"
      }
    }]
  }]
}

Supplemental Data Elements are used to report additional information about the subjects that may not be included in the Population criteria definitions. For example, it may be of interest to report the gender of all subjects for informational purposes. Supplemental data elements are defined by the Measure.supplementalData element, and are reported as Observations in the evaluatedResources of the MeasureReport.

Supplemental Data Elements can be specified as either CQL definitions or FHIR Path expressions.

An example Measure resource with some supplemental data elements set looks like:

{
"resourceType": "Measure",
   "supplementalData": [ {
      "code": {
         "text": "sde-ethnicity"
      },
      "criteria": {
         "language": "text/cql-identifier",
         "expression": "SDE Ethnicity"
      }
   }]
}

Stratifiers are used divide Measure populations into segments of interest. For example, it may be of interest to compare the Measure score between different age groups or genders. Each stratum within a stratification is scored the same way as the overall population. Stratifiers are defined using the Measure.group.stratifier element.

An example Measure resource with a stratifier set looks like:

{
  "resourceType": "Measure",
  "group": [ {
      "stratifier": [ {
         "code": {
            "text": "Stratum 1"
         },
         "criteria": {
            "language": "text/cql-identifier",
            "expression": "Stratification 1"
         }
      }]
   }]
}

A FHIR MeasureReport permits referencing the Resources used when evaluating in the MeasureReport.evaluatedResource element. HAPI includes these resources when generating subject reports for a single Patient. Evaluated resources for population or subject-list reports are not included. For large populations this could quickly become an extremely large number of resources.

The evaluated resources will not include every resource on the HAPI server for a given subject. Rather, it includes only the resources that were retrieved from the server by the CQL logic that was evaluated. This corresponds to the data-requirements for a given Measure. As an example, consider the following CQL:

valueset "Example Value Set" : 'http://fhir.org/example-value-set'

define "Example Observations":
   [Observation : "Example Value Set"]

That CQL will only select Observation Resources that have a code in the "Example Value Set". Those Observations will be reported in the Evaluated Resources while any others will not.

11.3.1.4FAQs

Q: I get an error saying HAPI can't locate my library, and I've verified it's on the server.

A: HAPI follows the Library conformance requirements defined by the CQF Measures IG, meaning the Library must have a logic-library type, the name and versions of the FHIR Library and CQL Library must match, and the url of the Library must end in the name of the Library.

FHIR Libraries generated from CQL via the IG Publisher follow these requirements automatically.

Q: Does HAPI support partitions for evaluation?

A: Yes, though the Measure and associated Resources must be in the same partition as the clinical data being used.

11.3.1.5Roadmap

• Complete cohort implementation
• Support for component stratifiers
• Support for FHIRPath expressions in Stratifiers, Supplemental Data Elements, and Population Criteria
• $data-requirements, $collect-data, $submit-data, and $care-gaps operations
• Support for more extensions defined in the CQF Measures, CPG, and ATR IGs