Clustering reveals two profiles of rewarding experiences after teen self-injury

Adolescents who engage in nonsuicidal self-injury, or NSSI, often describe what happens immediately afterward in strikingly positive terms: they feel better, or they feel that others respond with care. Those subjective rewards are thought to reinforce the behavior over time, yet most research on NSSI motivation has looked at broader motives rather than the moments right after an episode. A new exploratory study published in Child and Adolescent Psychiatry and Mental Health takes that narrower focus, asking whether clinical adolescents differ from one another in how rewarding they find their most recent self-injury episode, and whether those differences track with other clinical features.

The research team, led by Seunghui Han and Jae-Won Kim, assessed 132 adolescents receiving clinical care who had engaged in NSSI. The sample was predominantly female, at 87.9 percent, with a mean age of 15.1 years. To capture the immediate aftermath of self-injury, the researchers used the Suicide Attempt Self-Injury Interview, a structured instrument that asks in detail about the most recent episode, including the subjective experience that followed it. Participants also completed psychological questionnaires and supplied basic demographic and clinical information, giving the team a multidimensional picture of each adolescent.

The central analytical move was a k-means clustering analysis, an unsupervised machine learning method that groups individuals according to similarity on a set of measured features. Here the features were the levels of subjective rewarding experiences reported immediately after the most recent NSSI episode. Because k-means requires the analyst to specify in advance how many groups to form, the researchers used the silhouette method, a technique that scores how well each participant fits their assigned cluster relative to neighboring clusters, to determine the optimal number of subgroups. This is an exploratory, data-driven approach: rather than testing a predefined hypothesis about group membership, the analysis lets the pattern of responses itself suggest how many meaningful profiles exist.

The clustering produced a two-group solution. Adolescents in cluster 1, which contained 62.9 percent of the sample, reported lower levels of rewarding experiences associated with their self-injury. Adolescents in cluster 2, containing 37.1 percent, reported higher levels of these immediate subjective rewards. In other words, even within a single clinical sample of adolescents who self-injure, the felt payoff of the behavior was far from uniform, splitting the group into a larger low-reward profile and a smaller high-reward profile.

The differences did not stop at the reward ratings themselves. The low-reward cluster, despite reporting less subjective payoff from NSSI, showed greater levels of externalizing problems, the constellation of behaviors that involves acting out toward the external environment rather than inward emotional distress. This is a notable inversion of what a simple reward-reinforcement account might predict, and it suggests that the adolescents who find self-injury least rewarding are not necessarily the ones with the mildest clinical profiles.

Cluster 2, the high-reward group, carried a different burden. Compared with cluster 1, these adolescents had higher rates of depressive symptoms, higher rates of stressful life events, and higher rates of receiving medical treatment. The co-occurrence of elevated depression, accumulated life stress, and greater medical involvement paints a picture of adolescents for whom self-injury may function as a more powerfully reinforcing experience, at least in the immediate term, within a context of substantial distress.

The authors frame the findings as identifying adolescents who are vulnerable to the rewards inherent in NSSI. Because subjective rewarding experiences are thought to reinforce self-injurious behavior and have been associated with psychosocial dysfunction, knowing which patients report stronger immediate rewards could, in principle, inform how clinicians think about maintaining factors in each case. The study’s conclusion emphasizes that the two subgroups may benefit from different therapeutic approaches tailored to their situations, though the exploratory design means the clusters are descriptive profiles rather than validated treatment indications.

Several limitations follow from the study’s design. The sample was a clinical one, so the profiles may not generalize to adolescents with NSSI who are not in treatment, and the female predominance of the sample constrains generalization further. The assessment centered on the most recent NSSI episode as recalled and reported by participants, which introduces the possibility of recall and reporting effects. The cross-sectional nature of the data means the cluster differences describe associations at a single point in time; they cannot establish that rewarding experiences cause the observed clinical differences or the other way around.

What the study does deliver is a methodological template for a question that has been underexamined: not why adolescents begin self-injuring, but what they experience in the moments after an episode, and whether those experiences sort patients into clinically meaningful groups. By combining a structured interview about the most recent episode with an unsupervised clustering method and a principled choice of cluster number, the researchers show that immediate subjective reward is a dimension along which clinical adolescents with NSSI genuinely diverge, and that this divergence aligns with distinct patterns of depression, stress, externalizing behavior, and medical treatment. For a behavior whose reinforcement is thought to depend partly on what it feels like in the moment, that kind of heterogeneity is a finding worth taking seriously.

Subject of Research: Heterogeneity of subjective rewarding experiences immediately after nonsuicidal self-injury among clinical adolescents

Article Title: Profiling of subjective rewarding experiences immediately after NSSI in a clinical sample of adolescents using k-means clustering: an exploratory study

Article References: Han, S., Lee, K. H., Seong, E., Ahn, J., Park, M., Chun, J., Jang, M., & Kim, J.-W. (2026). Profiling of subjective rewarding experiences immediately after NSSI in a clinical sample of adolescents using k-means clustering: an exploratory study. Child and Adolescent Psychiatry and Mental Health. https://doi.org/10.1186/s13034-026-01195-1

Image Credits: AI Generated

DOI: 10.1186/s13034-026-01195-1

Keywords: nonsuicidal self-injury, adolescents, k-means clustering, rewarding experiences, depressive symptoms, externalizing problems, stressful life events, clinical sample, SASII, exploratory study, Profiling, subjective