This content provides information about composite measure specifications and discusses development, measure testing, and evaluation of composite measures intended for quality measurement in accountability programs. The Blueprint content does not cover quality indicator aggregations such as the Nursing Home Compare star rating and other similar collections of quality measures, and instead focuses on the development, implementation, and maintenance of individual quality measures. Composite measures can be useful for pay-for-performance programs and public reporting websites because they take several components and combine them into a single metric summarizing overall performance. This information supplements the content found in the Measure Specification and Measure Testing sections.

CMS defines a composite measure as “a measure that contains two or more individual measures, resulting in a single measure and a single score.” There are three common types of composite measures - all-or-none (person-level), any-or-none (person-level), and linear combinations (entity-level). Below, the first two examples are person-level and the third is entity-level:

• All-or-none (person-level): Measures of two or more individual performance areas scored using an algorithm producing a single score as its only output. With this type of composite, the individual components do produce individual scores, e.g., the measured entity meets all-or-none of the composite components (e.g., Severe Sepsis/Septic Shock: Management Bundle (CMIT Measure ID 0678)).
• Any-or-none (person-level): Measures of two or more individual performance areas scored using an algorithm producing a single score as its only output. With this type of composite, the individual components do produce individual scores, e.g., the measured entity meets any-or-none of the composite components (e.g., Chronic Conditions Composite (CMIT Measure ID 00593)).
• Linear combination (entity-level): Measures with two or more individual component measures assessed separately at the entity-level and then aggregated into one score.  The aggregation is either unweighted or weighted based on some conceptual rationale (e.g. importance) (e.g., Patient Safety and Adverse Events Composite (CMIT Measure ID 0135)).

There are also psychometric constructs used most commonly in survey-based measures.  Reflective constructs assume that the latent variable causes the indicators, making them interchangeable and expected to covary. Composite/formative constructs assume that the indicators collectively form the latent variable, with each indicator contributing uniquely and not necessarily expected to covary. This content does not address these psychometric constructs.

Other names for composite measures are composite performance measure, composite index, composite indicator, summary score, summary index, or scale. Composite measures can evaluate various levels of the health care system such as individual patient data, individual practitioners, practice groups, hospitals, or health care plans.

Purpose of Composite Measures

When measure developers group measures as a composite, they must have a purpose for the use of the composite (e.g., comprehensive assessment of adult cardiac surgery quality of care). There also needs to be a delineated quality construct for measurement (e.g., the four domains of cardiac surgery quality, which include perioperative medical care, operative care, operative mortality, and postoperative morbidity).

Measure development is unique for composites because the measure developer should examine and understand the intended use of the composite and relationships between the component measures. The American College of Cardiology Foundation/American Heart Association Task Force on Performance Measures (2010) provides useful guidance for composite measure development:

• Explicitly state the purpose, intended audience, and scope of a composite measure.
• The individual measures comprising a composite measure should be evidence-based, valid, feasible, and reliable.
• The methods for weighting and combining individual measures into a composite measure should be transparent and empirically tested.
• Demonstrate the scientific properties of these measures, including reliability, accuracy, and predictive validity.
• Composites should be useful for clinicians and/or payers to identify areas for quality improvement.

The American College of Cardiology Foundation/American Heart Association Task Force on Performance Measure implements this guidance when developing new measures and updating existing measures (American College of Cardiology Foundation/American Heart Association Task Force on Performance Measure, 2019; American College of Cardiology Foundation/American Heart Association Task Force on Performance Measure, 2020).