Rasch measurement is widely used for measurement of latent variables in psychologicaltesting, education, health status measurement, and other fields. The Raschmodel expresses ideal measurement requirements and much research has dealt withtesting whether these assumptions are met in real data. Latent variables are often ofinterest in terms of their relation to other variables, some examples being self ratedhealth as predictor of mortality or psychosocial work environment factors as predictorsof job turnover. This chapter deals with prediction of a binary outcome variableusing latent variables measured using the Rasch model. Three approaches are compared:(i) prediction using the sufficient score, (ii) prediction using the estimated valuesfor each person, and (iii) prediction based on a joint model. Extensions of the Raschmodel including uniform differential item functioning and uniform local dependencebetween items are also discussed. The approaches are illustrated and motivated usingan example from occupational epidemiology.