This study examined the validity of a measurement model of nursing home quality by using multilevel confirmatory factor analysis. Based on Mullan and Harrington's (2001) facility-level quality measurement model, a two-level analysis (facility and state) of the measurement model were performed. Two research questions were asked: (1) Can the measurement model developed at the facility-level be applied to state-level nursing home quality measurement? (2) Is the measurement model of nursing home quality stable over time? Panel data of 1997 and 2001, from the national OSCAR database, were used to test the assumptions. The results show that the state-level measurement model fits the data better than the facility-level model does. When the indicator "assessment" was removed from the state-level measurement model, a better-fitted measurement model was found. The two-level measurement model is relatively stable over time, demonstrating the construct validity of this measurement model.