Vibration-based structural health monitoring method has been widely utilized for operational condition monitoring of offshore wind turbines (OWTs). In this approach, damage to OWTs is often inferred from the changes in their modal parameters (i.e., natural frequency, damping ratio, and mode shape) extracted from the measured vibration responses. However, the modal-based approach may be less effective for detecting damage of OWT systems due to long-term effects (e.g., scour, corrosion). In this regard, this study proposes a statistical subspace-based method to detect damage of the OWTs due to long-term structural deterioration. The test statistics are determined by correlating the dynamic responses of the OWTs under reference and those under damaged conditions. Using the simulated responses of the OWTs under different operational conditions, the impact of key factors associated with the monitoring plan, namely, data length, noise level, and etc., on the damage detection performance is discussed and compared. In comparison to the modal parameter-based method, the proposed damage statistics appear to be more effective in detecting typical damage to OWT systems induced by long-term adverse effects. It is also observed that decreasing the measurement noise level and increasing the data length tend to enhance the damage identifiability of the proposed method.