Methodological Aspects of Composite Measures: A Comparison of Techniques for Rescaling and Aggregation
Author(s)
Schlick B1, Hoch A2, Dejaco T2, Kiesewetter K1, Rose-Eichberger K2, Scandurra F1, Schwarz C1, Urban M2
1MED-EL Medical Electronics, Innsbruck, 7, Austria, 2MED-EL Medical Electronics, Innsbruck, Austria
Presentation Documents
OBJECTIVES: Composite indicators are powerful practical tools summarizing complex phenomena by aggregating different empirical measures into a single index. This approach might be used to demonstrate the value of health technologies where valid comparative evidence is lacking. However, the quality of such a measure strongly depends on the methodology used in its construction. The validity of such an indicator further requires a high quality of the theoretical framework and the data used.
METHODS: The presented study follows methodical guidelines by the Joint-Research-Center of the European Commission (JRC) and the OECD for the construction of a composite indicator of hearing implant benefit. Based on a meta-analysis of published outcomes for different types of implants, three indicators were selected. Indicators were rescaled using distance to a reference, min-max transformation, and z-score standardization. The dimensions were combined by additive and geometric aggregation, resulting in six alternative models.
RESULTS: In all models an equal weight to the dimensions “audiological benefit”, “subjective benefit” and “safety” was assigned. Interestingly, the choice of rescaling methodology showed a considerable effect on the overall indicator score and the ranking of the different implant types. We observed that this is because z-score standardization and min-max-transformation widen the range of indicators lying within a small interval. This is particularly relevant for the audiological benefit dimension, which barely varies between different implant types. Thus, both methods increase the effect of this dimension on the overall indicator compared to the “distance to a reference” method. In our preferred model, a z-score standardization and geometric aggregation was applied, as the latter does not allow for a full compensation of one dimension for another.
CONCLUSIONS: The proposed composite indicator demonstrates to be sensitive to the techniques used for rescaling and aggregation. A careful assessment of the impact of different construction methods is essential.
Conference/Value in Health Info
Value in Health, Volume 26, Issue 11, S2 (December 2023)
Code
MSR148
Topic
Methodological & Statistical Research
Topic Subcategory
PRO & Related Methods
Disease
Medical Devices, Sensory System Disorders (Ear, Eye, Dental, Skin)