Postscript: Applications of Rasch analysis in health care.

Abstract

]here have been several seminal journal supplements devoted to conceptual and methodological advances in health status and quality of life assessment. 1-7 Influential articles on health status, quality of life, and outcomes assessment8-14 have also been published in supplements or special journal issues. This bolus of work has been instrumental in moving the field forward conceptually, methodologically, psychometrically, substantively, and clinically. Health services research has long embraced measurement as one of its foci. Measurement of functional status began in the 1950s, and measurement advances continue to be made. More than 70 distinct measures of functional status now exist.15 More than 2 dozen generic quality of life instruments have been developed,12 and hundreds of disease-specific measures now abound.16 Much of our historical measurement work has relied on theory and methods related to classic test theory (CTT). CTT has served the field well, and it will no doubt continue to do so in the years ahead. Although the methods and theory underlying item response theory (IRT) have been in existence since the 1920s,17 advances in IRT methods grew greatly after 1950 with the work of Birnbaum, Lazarsfeld, and Lord.17 Rasch's work emerged in the early 1960s.'8 In the past 40 years, there has been unparalleled progress in applications of IRT methods to test construction, scaling, and score equating, as well as for identification of item bias and supportive functions such as computerized adaptive testing. It has only been in the last 20 years that IRT methods have been applied in studies of health status. Early work focused on applications in mental health assessment.'9'20 IRT methods, particularly those of the Rasch model, were applied in physical rehabilitation in the 1980s.21 In the 1990s, IRT methods permeated many health sciences.22-30 To date, IRT methods (largely the oneand two-parameter models) have been used in a wide variety of applications in health care, ranging from identification of differential item functioning (DIF)31-34 to test equating,15,35,36 to computerized adaptive testing.37-39 IRT has also been used more directly in the development of de novo health status instruments22,40,41 and in shortening existing instruments.30,'42 This supplement of Medical Care contains 6 diverse pieces of work that discuss or employ the Rasch model. The supplement began originally with 14 articles, 6 of which