LAUSR

DOI: 10.5935/abc.20170059 To investigate the relationship between future or unknown outcomes and baseline health states among people with specific conditions, prediction models are an interesting strategy used to assist diagnosis, prognosis and treatment.1 They estimate the likelihood of clinical events taking into account clinical relevant measures and complementary tests.2 These predictors and their importance vary between the different events of interest and their prediction ability varies when considered singly or in combination with other predictors.2 They may facilitate simple and direct comparisons of risks, individualize treatment regimens, and may refine prognosis stratification of patients, especially when many prognostic factors are known. Models have to be simple, easy to use and lead the clinician to make decisions which are more likely to bring benefit to the patients.