This paper presents an online decentralized parameter estimation approach for structural systems using asynchronous data. This is motivated by the problems encountered in practice where data asynchronism from wireless sensor networks is inevitable. The proposed method performs first online decentralized parameter estimation at each sensor node. Then, only the condensed local model parameter estimation results are transmitted to the base station for fusion in order to obtain reliable global estimations. The proposed approach utilizes the fact that local model parameter estimation is insensitive to asynchronism of different sensor nodes because the local parameter estimation uses only data from its own node. Furthermore, the locally updated model parameters to be fused at the base station are insensitive to asynchronism. The proposed approach provides a simple but reliable framework for online parameter estimation of structural systems using asynchronous data directly. Furthermore, the proposed method requires neither a model of asynchronism nor estimation of the time shifts among different sensor nodes. Finally, examples using truss and bridge models are utilized to demonstrate the proposed framework. Comparison with conventional centralized identification results is also presented.