Reflective, causal, and composite indicators of quality of life: A conceptual or an empirical distinction?

Items (or indicators) that constitute “quality of life” instruments can be classified as either reflective (manifestations of some underlying construct), causal (the construct is an effect of the indicators), or composite (the construct is an exact linear combination of the indicators). Psychometric methods based on inter-item associations are only appropriate for reflective indicators, whereas other statistical and non-statistical validation methods can be used for composite or causal indicators. Thus, the distinction has important practical, as well as theoretical, implications. Attempts have been made to empirically identify which items of the EORTC QLQ-C30, a cancer-specific instrument, are causal and which are reflective. Such attempts, however, first require commitment to a particular definition of quality of life, of which there are many. Whether an indicator forms a composite, is causal or reflective of quality of life will depend on the definition adopted, and therefore, the reflective–composite–causal distinction is, arguably, best established on conceptual rather empirical grounds, guided by the “mental experiments” suggested by Bollen (Structural equations with latent variables, Wiley, New York, 1989). Conceptual models of health status and quality of life, as well as a cognitive-linguistic approach to quality of life assessment, may make some contribution to this practice. Theoretical consideration of indicator content can guide not only instrument development and validation, but also the selection of an appropriate instrument.