While traditional reliability models have focused on systems with binary states, recent research verifies that multistate degradation models are more appropriate structures to model degradation mechanisms, particularly for mechanical devices. This paper focuses on developing a condition-based replacement policy using online inspection data, which is one of the main problems in prognostic and health management of mechanical devices. This problem is addressed for a type of mechanical device with a multistate degradation process. In addition, one of the practical challenges in failure data analysis using inspection data called interval censoring is addressed. A control-limit policy is presented through a cost-based optimization model, which can be used to determine when to replace a degraded device. The application of the proposed model is demonstrated using a case study on a bearing shell's wear process.