[Predicting chance of disease: calculation using prediction rules]

A prediction rule is a statistical model that can be used to predict the presence or absence of a disease based on a limited number of tests or predictive factors. One of the mathematical methods used to formulate prediction rules is a logistic regression analysis of patient data. The discriminatory power of a model is visualizable using box-whisker plots and ROC curves; calibration plots show the match between the predicted chance and the observed frequency of a disease. These graphs are used to assess whether a model adequately reproduces reality. On publication of prediction rules it is important that the regression function is written out and that the chances of a disease on the basis of diagnostic scores are displayed in a histogram. For the practical significance of the model, it is also important to know how often the predicted low, medium or high probabilities of a disease do actually occur in comparison with the advance chance of occurrence.