Questionnaire layout and wording influence prevalence and risk estimates of respiratory symptoms in a population cohort

One of the more interesting aspects of asthma epidemiology is the challenge of identifying who in a population has the disease. By and large, asthma remains a clinical diagnosis, made on the basis of a characteristic history of symptoms, occasionally supported by physiological measurement. Without a discrete clinical test, epidemiologists have to rely on proxy measures of asthma, such as self-reported history of symptoms or a physician-diagnosis, to determine the prevalence and risk of the disease. Which is where we run into problems.

A new article by Swedish researchers highlights the extent to which the results of epidemiological studies can be influenced by the choice of methodology. The researchers tested whether subtle differences – including questionnaire layout and wording – had a measurable impact on prevalence and risk estimates of asthma.

Their report, published in the Clinical Respiratory Journal, describes how questions with different set-up or differences in wording resulted in significantly different prevalence and risk estimates. In particular, the use of follow-up questions that excluded subjects answering no to the initial question, resulted in 2.9-6.7 percent-units lower prevalence.

These kinds of findings highlight some of the underestimated complexities of chronic disease epidemiology. But they also underscore how little attention developers of mobile and wireless health tools have put into understanding and incorporating previous research on the implications of different methodologies into their applications. Reports like these should motivate us to save ourselves a lot of unnecessary originality and admonish us to greater care and reflection in our work.

Comments are closed.