22 - Applications of item response theory to improve health outcomes assessment: developing item banks, linking instruments, and computer-adaptive testing

Summary

Introduction

For the field of health outcomes assessment to move forward, researchers must integrate advances in measurement theory with improvements in computer technology. We need health outcomes researchers and patient advocates to guide the development of tailored health-related quality-of-life (HRQOL) instruments that meet the criteria for validity, reliability, and sensitivity to change in health status while minimizing the burden of questions for the cancer patient. Also, to accurately characterize a patient's HRQOL after disease onset or treatment (a call for more specific than generic measures), we need the ability to crosswalk scores from one instrument to another despite groups of patients receiving different sets of questions.

In Chapter 21, this volume, Reise discusses the shortcomings of traditional (i.e., classical test theory, CTT) methods that have been used to direct the development and analysis of HRQOL instruments. CTT item statistics are dependent on the particular sample of respondents, while patient scores are dependent on the particular choice of items. Such dependencies hinder the ability of researchers to combine results from studies using different HRQOL measures or to create tailored instruments by choosing specific questions from other instruments. Reise also explains how item response theory (IRT), with the properties of item parameter invariance (over samples of respondents) and respondent parameter invariance (over samples of items or questions), provides a flexible tool for instrument developers to link sets of questions measuring the same domain on a common metric and to create computerized adaptive assessments.