Presentation formats

Background

The presentation of a prediction model deserves careful attention. Epidemiologic regression analyses commonly concentrate on estimation of relative effects, and present tables with odds ratios or hazard ratios, and their confidence intervals. Such tables are usually not sufficient to calculate absolute risks, which requires a model intercept (for continuous or binary outcomes) or a baseline hazard (for survival outcomes). We need to separate presentations that generate predictions (“clinical prediction models”) from presentations that generate advice for a decision (“clinical decision rules”). Various presentation formats are possible for prediction models and for decision rules, some of which will be discussed in this chapter. We illustrate the creation of some formats at a technical level for the testicular cancer case study.