Level of Measurement/Analysis

The unit of measurement/analysis is the primary entity upon which to apply the measure. The measure developer should clearly state and justify the procedure for attributing the measure. The measure developer should specify measures with the broadest applicability (e.g., target/initial population, setting, level of measurement/analysis) as supported by the evidence. However, a measure developed for one level may not be valid for a different level. Therefore, the measure should be respecified and retested for reliability and validity in each setting/population.

Examples

• A measure created to measure performance by a facility such as a hospital may or may not be valid to measure performance by an individual physician.
• If developing a claims-based measure for Medicare use, and the literature and guidelines support the measure for all adults, the measure developer should consider expanding the data sources beyond Medicare Parts A and B claims.

Given the circumstance that multiple entities are using a shared electronic health record, measure developers should consider all relevant testing needed to minimize the possibility of quality actions performed by one entity inappropriately attributed to another entity.

Sampling

If allowing sampling, the measure developer should describe the sample size or provide guidance in determining the appropriate sample size and describe any prescribed sampling methodologies explicitly.

Sampling is a useful technique for quality measures that require primary data collection (manual abstraction) from a source such as the patient medical record. Facilities, e.g., hospitals, with a large number of cases in their population may use sampling because there is a need for a fairly large number of sample cases to achieve a representative sample of the population. As the number of chart-abstracted measures decreases, so does the need for sampling.

Sampling is not applicable to electronic clinical quality measures.