Define Stratification Scheme

Measure developers may define a stratification scheme in lieu of risk adjustment by stratifying the population based on their risk for an outcome or procedure. They may also stratify according to a reporting scheme (e.g., if reported data are in strata by age groups). For more information, refer to the Risk Adjustment in Quality Measurement supplemental material, Table 1. Framework for Risk Adjustment Strategies.

Measure developers should always consider stratifying by sociodemographic characteristics. Stratification may effectively detect potential disparities in care/outcomes among populations related to the measure focus. If the measure developer stratifies results by population characteristics, then the measure developer must describe population stratification variables.

When the measure definition includes stratification, the measure developer reports each population in the measure definition both without stratification and by each stratification criteria. For measures with multiple numerators and/or strata, they should consider scoring each patient/episode for inclusion/ denominator exclusion and/or numerator exclusion to every population. For example, if a measure has two numerators, and the patient is included in the first numerator, the patient should also be scored for inclusion/ denominator exclusion and/or numerator exclusion from the populations related to the other numerators e.g., Antidepressant Medication Management (CMIT Member ID 63) (CMS CBE #0105).

Measure developers should stratify measures by organizational characteristics. This is known as peer group stratification, and it is appropriate in any circumstance when there is unmeasured systematic and persistent patient heterogeneity and characteristics of the organizational setting are related to that unmeasured patient heterogeneity (e.g., location). There should be an explicit hypothesis or rationale as to why the characteristic is related to the unmeasured patient heterogeneity, but not the quality construct.

If there is a reporting requirement of multiple rates or stratifications, the measure developer should state this in the specifications. If the measure developer includes the allowable denominator and/or numerator exclusion in the numerator, they should specify the measure reports the overall rate as well as the rate of each denominator and/or numerator exclusion. If stratification of results is by population characteristics, they should describe the variables used.

Examples

Possible stratification schemes

  • Vaccination measure numerator that includes (1) health care worker who received the vaccine; (2) health care worker who was offered the vaccine and declined; or (3) health care worker who has an allergy, a condition, or another medical contraindication to the vaccine, e.g., Influenza Vaccination Coverage Among Healthcare Personnel (CMIT Member ID 390) (CMS CBE #0431)
  • Stratify the measure by a population type (i.e., race, ethnicity, age, social risk factors, income, region, gender, primary language, disability), e.g., Chlamydia Screening for Women (CMIT Measure ID 128) (CMS CBE #0033)
  • For electronic clinical quality measures, include a Reporting Stratification section in the human-readable document. If a measure does not have reporting strata defined, the default display is “None.” If a measure contains reporting stratification, list each of the reporting strata separately under the Population Criteria section. For details, see the Measure Authoring Development Integrated Environment (MADiE) User Guide.
Last Updated: