Construction of a global score from multi-item questionnaires in epidemiological studies

Perceived health, mood, job demands and social support are common outcome variables in epidemiological studies of the psychosocial work environment, as measured by multidimensional multiitem questionnaires. The Stress-Energy Questionnaire (SEQ) is one such questionnaire and was developed to measure two critical aspects of mood at work. When a variable is measured by more than one item, the construction of a global score for that particular variable is often necessary. The most common way of aggregating items into a total or global score is to sum or average the responses, which requires equidistance scale categories and all items to be equally important. Assessments on many questionnaires, including the SEQ, are made on rating scales, meaning that the data consist of ordered categories irrespective of the type of coding system. These codes do not represent numerical values, but rather only convenient labelling devices for ordering responses from the lowest to the highest amount of the characteristic being measured. They do not have the mathematical properties needed for arithmetic calculations. In this study, different approaches for the construction of global scores are discussed. We have showed that there are alternative methods for the construction of global scores that take into account the nonmetric properties of data from questionnaires. The median and the criterion based approaches are proposed as the appropriate methods to use for ordinal data and these were applied to the empirical dataset.