Purpose. Despite dramatic increase in the public caregiver support programs, there are few studies to investigate how the public support programs affect individuals' program utilization. Family Caregiver Alliance (FCA) reported that there were large variations in state-funded caregiver support programs between states. Therefore, this study assesses if there is substantial difference in the use of state-funded caregiver support program between states and examines what factors are related to these variations. Method. Using the Andersen model, this study modified the model by adding the factors related to informal caregiver and state program factors. Considering the state-funded programs as extension of enabling domain, this study designed the multi-level model characterized by hierarchical data structures in which caregivers were grouped into state units. The state level predictors were collected from the Survey of Fifteen States' Caregiver Support Programs in 1999 by FCA. For the individual level data, the 1999 National Long Term Care Survey from the elderly aged 65 or older and their informal caregivers was used (n=647). Bayesian estimation methods were used to test the random intercept and random slope models. The Bayesian software WinBUGS 1.4 program was used to fit the multilevel models testing the predictors at the individual level ( the need factors relevant to their health status, family income, insurance type, residential area, and demographic characteristics) and at the state level ( the number of served persons, total state expenditure, eligibility criteria of program beneficiary). Results. A random ANOVA model found the small variation in the average level of utilization of caregiver support program between states versus within state. After 3000 iterations, approximate missing of three parallel simulated chains was conducted. The result of multilevel analysis revealed that individual level predictors had the significant relationship with the use of caregiver support programs. The informal caregivers caring for those with poorer functional status and severer medical condition were more likely to use caregiver support program. Higher level of caregiver burden increased the use of caregiver support program. Among demographic factors, age of caregivers was positively associated with program utilization. In addition, insurance type was estimated to have little predictive power. In case of state level predictors, total state expenditure of program capturing the financial capacity of program was the most significant factor for this particular sample. Implication. The results suggest that caregivers with higher level of needs should be targeted for closer access to caregiver support programs. According to the significant impact of financial capacity of state programs, increased attention should be paid to the role of state government in the activation of caregiver support programs and in the delivery to family caregiver. In light of the recent reinforced National Family Caregiver Support Programs authorized by the Older Americans Act of 2000, further studies will be required to test various state program factors.