Deep learning driven clustering of post-traumatic stress disorder (PTSD) profiles in student-athletes: implications for precision and stratified mental health support.
Athletes frequently experience potentially traumatic events related to sports injuries, significantly increasing their risk of developing post-traumatic stress disorder (PTSD). However, existing assessment methods often overlook the heterogeneity of symptom presentation and the necessity for individualized mental health management. Furthermore, exploration of mental health support for the specific population of student-athletes remains limited. This study aims to integrate a deep learning-driven clustering paradigm with a stratified care framework to explore distinct PTSD symptom clustering profiles among high-performance college athletes who have experienced severe sports injuries.
A total of 468 high-performance college athletes who had sustained severe sports injuries within the past year were recruited to complete the PCL-C questionnaire. Transformer-based semantic embedding techniques were used to encode PTSD symptom data, followed by Principal Component Analysis (PCA) and K-means clustering to extract latent subgroups. Model performance was compared with traditional clustering approaches to evaluate clustering validity and representational quality.
Deep learning-based semantic clustering paradigms demonstrate efficacy and feasibility in capturing symptom distributions and latent heterogeneous combinations of clustered symptoms. Five distinct PTSD clusters were identified, reflecting a continuous gradient of symptom severity and heterogeneous combinations of intrusive, avoidance, emotional numbness and hyperarousal features. These clusters correspond to varying levels of clinical risk and offer a complementary insights for stratified care and mental support.
This study proposes a novel cluster-based research paradigm for the precise screening of PTSD in student-athletes. By integrating clustered symptom profiles with stratified care principles, the findings provide actionable insights for optimizing resource allocation and enhancing mental health support systems for student-athletes.
A total of 468 high-performance college athletes who had sustained severe sports injuries within the past year were recruited to complete the PCL-C questionnaire. Transformer-based semantic embedding techniques were used to encode PTSD symptom data, followed by Principal Component Analysis (PCA) and K-means clustering to extract latent subgroups. Model performance was compared with traditional clustering approaches to evaluate clustering validity and representational quality.
Deep learning-based semantic clustering paradigms demonstrate efficacy and feasibility in capturing symptom distributions and latent heterogeneous combinations of clustered symptoms. Five distinct PTSD clusters were identified, reflecting a continuous gradient of symptom severity and heterogeneous combinations of intrusive, avoidance, emotional numbness and hyperarousal features. These clusters correspond to varying levels of clinical risk and offer a complementary insights for stratified care and mental support.
This study proposes a novel cluster-based research paradigm for the precise screening of PTSD in student-athletes. By integrating clustered symptom profiles with stratified care principles, the findings provide actionable insights for optimizing resource allocation and enhancing mental health support systems for student-athletes.