Background: A surveillance system for self-harm has not been established in Hong Kong. The existing data source has an unknown degree of underreporting, and therefore a capture–recapture method has been proposed to correct for the incompleteness. Aims: To assess the underestimation of the incidence of self-harm cases presenting to hospital in Hong Kong using a capture and recapture method. Method: Two different yet overlapping hospital administrative datasets of self-harm were obtained from all public hospitals in Hong Kong. From 2002 to 2011, 59,473 distinct episodes involving 36,411 patients were identified. A capture–recapture model considering heterogeneous capture probabilities was applied to estimate the number of self-harm episodes. Results: The estimated number of self-harm incidence was 79,923, equally shared by females and males. Cases of self-harm by females were more likely to be ascertained than those by males. The estimated annual incidence rate of self-harm in Hong Kong from 2002 to 2011 ranged from 96.4 in 2010 to 132.7 in 2002. Limitations: The proposed method does not include patients who required no medical attention and those where the patient consulted private doctors. Conclusion: The capture–recapture model is a useful method for adjusting the underestimation of self-harm cases from existing databases when surveillance system is not available and to reveal some hidden patterns.