In recent years, the incidence of self-injury in adolescents has shown an upward trend year by year. The high prevalence of self-injury and various accompanied adverse consequences (e.g., declined academic performance and damaged interpersonal relationships) are becoming serious public concerns. To provide insight on prevention and intervention program, this study examines the properties and development of self-injury behaviors in adolescents. It is of interest to study impulsive behaviors in a categorical and latent approach, so that the subgroups of adolescents based on different patterns of observed behavior can be found as well as a single strategy aimed at the underlying trait effective in preventing all forms of self-injury behaviors. Taking a longitudinal perspective enable researchers to understand which individuals are expected to transition to more risky groups in the future and vice versa. Latent transition model was used to examine 10 self-injury behaviors, i.e., self-cutting, burning, biting, punching, carving words or pictures on the skin, scratching skin, inserting objects to the nail or skin, pulling hair out, erasing skin, banging the head or other parts of the body against the wall over three waves in the present study. Questions of stability and change were examined with a model that accounts for possible transitions. Participants were 3,600 students (2,037 girls) from secondary schools (mean age = 14.63 ± 1.25) in Hong Kong, China. Gender was used as a grouping variable to explore the differences in latent structure. Nine covariates, i.e., age, family factors(depression, anxiety, stress), impulsivity, distress tolerance and emotional disorder(parental marital status, parental relationship, parent criticism) were examined as predictors of status of self-injury behaviors. Three distinct groups are found: Non-self-injurers, Experimenters, Repeaters. Non-self-injurers was the most prevalent status and they barely involve in any self-injury behavior; the prevalence of Experimenters and Repeaters showed an decreasing trend across time. Experimenters reported two relative mild self-injury behaviors, i.e., pulling hair out and self-biting, while Repeaters reported a wide variety of self injury behaviors, i.e., self-cutting, biting, carving words or pictures on the skin, scratching skin, pulling hair out, banging the head or other parts of the body against the wall. In both transitions (from T1 to T2 and from T2 to T3), Non-self-injurers was highly stable, while Experimenters and Repeaters tend to move to a less problematic status (Experimenters to Non-self-injurers and Repeaters to Non-self-injurers or Experimenters). Results of multi-nominal logistic regression revealed that adolescents who reported high impulsivity, negative emotions (i.e., depression, anxiety, stress), negative family factors (parental divorce or poor relationship, parental criticism) and low distress tolerance were more likely to be in high-risk status (Experimenters or Repeaters). In addition, high impulsivity was associated with decreased transitions to low-risk status (Non-self-injurers) and increased transitions to high-risk status. Significant gender differences are found both in prevalence and transitions of self-injury behaviors. Compared with boys, girls were more likely to be Experimenters or Repeaters and tend to stay in the high-risk status. Findings of this study emphasizes the importance of intervention program for adolescents who are more vulnerable (adolescents report high impulsivity, negative emotions, negative family factors or low distress tolerance, especially girls).