The apparent performance is optimistic because the same data are used for development and testing.

Model development usually includes defining the model structure based on findings in the data (e.g. strong, ‘independent’, predictors are selected). Model parameters are usually estimated as if the model structure was pre-specified. Modern estimation techniques are available that take uncertainty in model structure into account, but these are not yet applied routinely in medical research.

The apparent model performance may be satisfactory, since the model was fine-tuned for the development data. Testing in new patients may show disappointing results.