The purpose of this study is to investigate the relationship between risk preference, purchase of private health insurance and the utilization of inpatient service based on latent class analysis (LCA) framework. Adopting data from the 2005 Nation Health Interview Survey (NHIS) and linked it with the claim data, the 2006 and 2007 National Health Insurance Research Database (NHIRD), we first utilize latent class analysis (LCA) to identify risk preference classes. Next, we utilize a bivariate probit regression to investigate the relationship between each latent class of risk preference and purchase of private health insurance and the subsequent inpatient service utilization. Three latent class of risk preference were identified: high cautiousness; low cautiousness without wearing a seat belt or helmet; low cautiousness with smoking behavior classes. The bivariate probit empirical results show that compared with the high cautiousness class, low cautiousness with smoking behavior class tends to less likely to purchase private health insurance; however, low cautiousness with smoking behavior class with private health insurance tends to more likely to utilize inpatient service. Moreover, our results also point out an advantageous selection may be presented. It indicate that based on the National Health Insurance Program in Taiwan, the asymmetric information in the supplementary private health insurance market exist after controlling for individuals’ risk preference heterogeneity.