Harmonizing measurements: establishing a common metric via shared items across instruments

Abstract

Abstract Background The proliferation of instruments that define instrument-specific metrics impedes progress in comparative assessment across populations. This paper explores a method to extract a common metric from related but different instruments and transform the original measurements into scores with a standard unit of measurement. Methods Existing data from four assessment instruments of child development, collected from three different samples of children, were used to create “equate clusters” of items that measure the same behaviour in (slightly) different ways. A probability model was formulated to identify best items and groups to serve as anchors linking the instruments, assuming that items in an anchoring or “active” equate cluster are psychometrically equivalent. Quantification and inspection of item characteristic curves were used to resolve which equate clusters should be active. We simulated the impact of various analytic choices. Results Simulation confirmed the feasibility of creating a common metric from data collected with different instruments from respondent samples with different abilities. The method performed as expected in an application in early childhood development. Conclusions The use of equate clusters is an intuitive and flexible way to establish a common metric across instruments and facilitates the transformation of measurements obtained to a standardized scale. Standardizing instrument scores to a common metric allows for population-level comparisons on a global scale.

Keywords

• Rasch model
• Early childhood development
• Global metric
• Concurrent calibration
• Meta-analyses