DOI: 10.21276/sasjs.2018.4.3.2

Pages: 59-64

Predictive models are utilised in clinical practice and decision-making processes in cancer patients. The uptake of these models is not as common as other tools used in other clinical settings. This could be due to lack of understanding of the process of building and validating these tools. Aims: To help clinicians to understand the process of building and validating predictive models. Methods: A literature review and practical example from our published work is used to explain the process of building and externally validating a nomogram for colon cancer survival. Results: The nomogram building process should include clear steps to identify the population and outcomes of interest, the predictors, the model used to build the nomogram and the validation process. Externally validating the tool in a different population is important to ensure the tool’s generalizability and its ability to predict the outcomes. There are factors to consider when externally validating any model; one important factor is the number of events in the population used to validate the model.