[Bayes' Theorem and Base Rate Error: On the Predictive Value of Risk Factors and Screening Tests]

Regularly, risk assessments or decisions are based on dichotomous characteristics or measures. For instance, critical high values in screening tests may indicate an enhanced risk of having a disease. The prevalence of a risk factor (e. g. high blood pressure, depression) may indicate the need of a medical treatment or the occurrence of a negative consequence (e. g. early retirement). In such situations risk assessments are typically based on conditional event rates or conditional event probabilities: What is the risk of having a disease under the condition of having a symptom, risk factor or a critical screening indication? The valid interpretation of conditional event rates poses a considerable challenge for practitioners as well as for rehabilitation patients. Typically, for rare events the risk of having a disease is considerably overestimated if a critical predictor prevails. The Bayes' theorem describes the decision related informations structure appropriately. It is shown, which informations have to be taken into account to ensure valid risk assessments. Using absolute frequencies instead of relative frequencies or proportions may prevent erroneous conclusions. Graphical displays of absolute frequencies by area diagrams or decision trees are recommended to ensure a better comprehensibility of risk information.