OBJECTIVES: The aim of the study was to test items for Differential Item Functioning (DIF) in a condition-specific measure for women having an abnormal screening mammography. METHODS: The draft version of a 33-item condition-specific measure was completed by 184 women who had received an abnormal screening mammography and two weeks later by the same women after having either a known false-positive result, a diagnosed breast cancer or still undiagnosed. The measure was also completed by 240 women with a normal screening result. Data were analysed using the Rasch model. DIF relative to person covariates was checked by analyses of variance examining the degree to which individual residuals for specific items depended on the covariates. Absence of evidence of interaction between the covariates and the estimated trait parameters were taken as evidence of DIF being uniform. If uniform DIF was identified; mean scores of the dimension were estimated from the means of the person locations on the latent trait according to DIF found. RESULTS: One of six items in a dimension measuring “sense of dejection” had uniform DIF in two of four subgroups. After deleting this item the remaining items still fitted the Rasch model with no DIF. In a 2-item dimension describing the impact on breast examination both items had uniform DIF. The first item had uniform DIF relative to women diagnosed with breast cancer and the second item had uniform DIF relative to women with normal screening mammography. There was difference in the mean of the subscales' raw score and the estimated scores calculated according to identified DIF. CONCLUSION: If an item functions differently in subpopulations and it is decided to keep the item in a subscale; special precautions should be taken when calculating scores of this dimension. Otherwise, the difference between subpopulations could be under or overestimated.