Prediction Models in Chronic Obstructive Pulmonary Disease - Informing Practice and Improving Research

Abstract

Prediction models have the potential to advance personalised medicine through risk-stratified management, which optimises the benefits and harms of preventive and therapeutic interventions for individuals. Thereby, the risk that patients are either underor over-treated can be minimised. In research, prediction models can be used to efficiently identify study subjects, to control for confounding and to study subgroup effects. In the area of chronic obstructive pulmonary disease (COPD), many prediction models have been developed to predict the risk of early death or exacerbations, the course of disease or for other outcomes. A prerequisite for their use is a careful development and external validation, which is, however, often lacking. This article describes the purposes of prediction models as well as their development and validation including novel approaches like multiple score comparison meta-analysis. It also describes research needs like the need for developing accurate and valid prediction models for exacerbations that support risk-stratified prevention of exacerbations. (BRN Rev. 2018;4(3):156-68) Corresponding author: Milo Puhan, miloalan.puhan@uzh.ch