Nowadays, health indicators play an important role in the studies of accurate remaining useful life estimation. Health indicators can be roughly classified into two categories, one is the integrated health indicators, and the other is non-integrated health indicator. However, few scholars have focused on the which type of health indicator is more adapted to remaining useful life estimation. In this work, first, we take four different health indicators of aforementioned two categories into account, and we discuss accuracies of remaining useful life estimation using four different health indicators. As well, the public XJTU-SY bearing datasets are used for experiments. Finally, we draw a conclusion of what type of health indicators with a smallest error of RUL estimation.