The availability of race data is essential for identifying and addressing racial/ethnic disparities in the health care system; however, patient self-reported racial/ethnic information is often missing. Indirect methods for estimating race have been developed, yet these perform poorly among racial minorities, usually only consider geocoded and surname data as predictors, and are unable to provide race estimates for subjects missing this information. The objective of this study was to develop a race estimation method that addressed these limitations and achieved higher predictive performance than previous methods. A subset of Rhode Island Medicaid beneficiaries was used to explore multiple imputation methods for race. The general location model was used to implement joint model imputation while Bayesian multinomial regression models were explored for fully conditional specification imputation. Our method outperformed existing ones, especially among American Indians and Asians, using evaluation criteria like area under the curve and racial composition estimates. Family race was identified as an important predictor of race and should be included in race estimation models when possible.