Revista Brasileira de Epidemiologia
Print version ISSN 1415-790X
REICHENHEIM, Michael E. and MORAES, Claudia Leite. Pillars for assessing validity in epidemiological studies. Rev. bras. epidemiol. [online]. 1998, vol.1, n.2, pp. 131-148. ISSN 1415-790X. http://dx.doi.org/10.1590/S1415-790X1998000200004.
The article outlines several validity issues in epidemiological studies. Six types of validity are covered. Conceptual validity concerns the elements sustaining a theoretical explanatory model, which, by definition, exists only on an abstract and discursive level. Operational validity concerns the processes by which concept-pertaining dimensions at a theoretical level are reduced to variables and indicators, enabling empirical testing. In validity of domain, focus is shed on the exchangeability between subjects of the target population and those of the study population as a basis for generalizing results. Comparison validity concerns the identification of circumstances that rupture the comparability structure between groups, namely, confounding and selective exclusion of subjects. Information validity relates to data collection and ensuing biases. Attention is drawn to the problems of timing and order of measurement related to events and data collection procedures. Finally, focus is placed on some important issues related to specification validity of statistical models, identifying guidelines for selecting those models. In the final part, systematic handling of validity issues is contended as a possible building block for attaining knowledge in epidemiology.
Keywords : Epidemiology; Models, theoretical; Confounding factors; Models, statistical; Research design; Bias; Data collection, methods.