Evaluation of Prediction Models and New Markers

• Activity Number: 574
• Type: Invited
• Date/Time: Wednesday, August 12, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Ewout Steyerberg*
• Companies: Erasmus Medical Center
• Keywords: prediction ; regression analysis

Abstract

Prediction research receives increasing attention. In this presentation, I aim to discuss ongoing research that links prediction, treatment decision-making and cost-effectiveness for existing models and new markers. I will use a classic example as an illustrative case study: the prognosis of patients suffering from a myocardial infarction. 1) Prognostic markers are well able to predict 30-day mortality in patients suffering from a myocardial infarction [1]; 2) A prediction model can subsequently be used to direct decision-making on treatment [2]; 3) The cost-effectiveness of a new therapy is commonly considered at an average level [3]; 4) A more comprehensive perspective for individualized decision-making should consider prognostic factors for treatment benefits (mortality risk reduction x life expectancy) and costs [4]. I will review the work done previously and discuss directions for future research, highlighting challenges in quantifying incremental value of prediction models over not using such models, and novel markers.