Testing causal effects in observational survival data using propensity score matching design