Catastrophic mortality bonds are designed to hedge against the mortality risks. The payoff at maturity depends on the realized mortality index over the life of the bond, therefore modeling the mortality index is the main concern in our study. Since mortality shocks are detected using outlier analysis, non-Gaussian state space models with a fat-tailed error term are proposed to fit the mortality index and to handle shocks. By comparing several state space models with different fat-tailed distributions, an ARIMA process for the baseline mortality and the t-distribution for capturing mortality shocks are chosen. We obtain the price of the mortality bond using the proposed model and estimate the market price of risk. It appears that the market price of risk is lower than the ones obtained in the literature, which is consistent with the industrial empirical results from Wang (2004). This implies that our model is capable of handling mortality risks.