The Hospital Anxiety and Depression Scale (HADS) is often used to screen for cancer caregivers' anxiety and depression, despite few studies examining the tool's psychometric performance within this population. The purpose of this article is to use Rasch analysis to assess the psychometric properties of the HADS in a sample of cancer caregivers. HADS was administered to 541 caregivers of a population-based sample of patients diagnosed with one of the eight most incident cancers in Australia. Rasch analysis was conducted using RUMM2020. More than two-thirds of caregivers were women and most participants were married (95.9%) and caring for their spouse/partner with cancer (89.8%). The HADS-Anxiety (HADS-A) subscale showed good fit to the model, with appropriate internal consistency after removal of item 11. There were no disordered thresholds and no differential item functioning (DIF) for sex or age. To achieve satisfactory model fit for the HADS-Depression (HADS-D) subscale item 8 was removed due to DIF for sex and item 14 was rescored to resolve disordered thresholds. Analyses supported the unidimensionality of the individual subscales; however, there was no support for the combination of subscales to form a HADS-Total. The results of Rasch analysis support the use of a modified version of the HADS-A and HADS-D among cancer caregivers. Further research is needed to confirm these findings and identify revised clinical cut-points. Findings reinforce the need for clinicians and researchers to formally test the psychometric properties of the instruments that they intend to use with different samples. Copyright © 2010 John Wiley & Sons, Ltd. Copyright © 2010 John Wiley & Sons, Ltd.