Implications of the structural model in the validation of clinical instruments: Reflective model vs. formative model

Abstract

Background: Questionnaires, scales, indices and classification criteria are developed and subjected to validation to measure subjective variables in a standardized manner.

Theoretical bases of structural models: When the questions in a questionnaire originate from the attribute, it corresponds to the reflective model that has been widely used in psychometrics due to the nature of the variables of interest. In Health Sciences the variables of interest are different, when the questions in the questionnaires arise from constitutive elements and are independent of each other, they belong to a formative model.

Implications of the structural model in the design and validation of health measurement instruments: When using the reflective model, many questions are created, and their reduction will be carried out through statistical analysis. The reliability of the test (internal consistency) and structural validity must be determined.

With the formative model, a reduced number of items are developed, ensuring content validity with scientific evidence or, failing that, by formal consensus of experts. To refine them, the level and quality of evidence, the characteristics of the target population and the purpose of the instrument, statistical methods, and the researcher’s judgment and/or formal expert consensus are used. If there is a gold standard, a study of diagnostic tests is carried out, otherwise the construct validity.

Differences of the reflective versus formative model in the instrument development phase: Nature of the construct, chronology of the phenomenon, relationship between the indicators or items and the construct, directionality of the link between indicators and construct, number of indicators identified, items to be developed and possible reduction method, theoretical impact of the elimination of items or indicators on the appropriateness of construct measurement.