O34?Item response theory modeling; old kid on the block to improve pediatric drug dosing

摘要

Background Quantification of changes in disease states or clinicalconditions is essential to establish drug effects and dosingguidelines. When single endpoints for direct measurements arelacking, multi-factorial observational scales may be used, likefor instance pain or sedation scales.1 2Methods Analysis of data from multi-factorial observationalscales is commonly based on total scores. This assumes eachitem to be equally informative, which is generally not true.Item Response Theory (IRT), a long existing method in socialsciences and psychology, has only recently been recognized forits ideal applicability to the analysis of multi-factorial observationalscales. In this approach a latent variable is derived fromall item-level data and the information that each item adds toestablishing the latent variable is being weighted appropriately.The basic concepts, assumptions and applications of IRT inpharmacological research are introduced and illustrated withexamples from studies on analgesia and sedation in the PICUand NICU.