In Table 4 of their recent paper , Moos et al. state that they provide the ‘sensitivity, specificity and accuracy’ of various levels of alcohol consumption for predicting drinking problems. However, in reviewing this table for an online synopsis of current research (Alcohol, Other Drugs, and Health: Current Evidence, http://www.bu.edu/aodhealth/index.html), we noticed that sensitivity increases and specificity decreases with more stringent criteria, which would be unusual. The table footnote incorrectly defines sensitivity as ‘percentage who exceed guidelines who have drinking problems’. This definition describes the positive predictive value of consumption exceeding guidelines . Similar confusion regarding specificity and negative predictive value are also noted.
Although clinicians are most interested in the predictive value of a positive or negative test, epidemiologists have well-established definitions of sensitivity and specificity, i.e. the probability that those with disease will have a positive test (sensitivity) and those without disease will have a negative test (specificity) . Those parameters are particularly useful as they do not vary with the prevalence of disease, as do predictive values. For clinicians working with older adults in settings with different prevalences (i.e. pretest probabilities) of alcohol problems than those in the population studied, publication of an erratum with the correct sensitivities and specificities would be useful.