OBJECTIVES: To evaluate the item measurement property of the SF-12v2™ and the merits of the physical and mental health component scales (PCS and MCS) using Rasch model. METHODS: We extracted a sample from the 2005 Medical Expenditure Panel Survey (MEPS). We included respondents who had completed the SF-12v2™, were ≥18 years of age and were diagnosed with any of the top ten most prevalent chronic conditions using the primary ICD-9-CM (n = 5151). Rasch analyses were applied repeatedly in subgroups on all SF-12v2™ items as a 12-item scale as well as on PCS and MCS items separately as two 6-item scales. Items were evaluated using the Rasch model fit statistics to examine 1) whether or not the 12 items measure a unidimensional construct and 2) whether or not PCS items measure a physical health construct and MCS items measure a mental health construct. RESULTS: In the 12-item scale, the two mental health (MH) items “Have you felt calm and peaceful” and “Have you felt downhearted and depressed” consistently demonstrating misfit in all subgroups (infit/outfit mean square > 1.3). By subdividing the PCS and MCS items into two separate 6-item scales, all six MCS items had good model fit including the two MH items. On the other hand, two PCS items concerning respondents' general health and bodily pain showed misfit to the Rasch model. CONCLUSIONS: The SF-12v2™ items do not form a unidimensional scale as the two MH items did not fit with other ten items. However, there are tradeoffs treating PCS and MCS as two distinctive scales as PCS items had a reduced fit. A generic SF-10 for adult population may be a unidimensional component of the SF-12v2™ and could serve as a brief health survey instrument that improves upon the SF-12v2™.