Identifying Features for Keloid Scars Subtyping Using K-Modes Clustering

Keloids are pathological scars that form due to abnormal wound healing, characterized by an overgrowth of collagen. They develop beyond the sites of the original lesion and have a high recurrence rate, are cosmetically deforming, can cause irritation, physical discomfort, and psychological stress, and frequently impact the quality of life. Despite an array of therapeutic approaches, both surgical and non-surgical, no solutions have been able to offer long-term relief. Given the limited understanding of keloid etiology, our study employed a novel data-driven approach using the k-modes clustering algorithm to analyze patterns of ICD-10 medical codes documented in patients’ medical history within a patient cohort with hypertrophic disorders of the skin to assess which risk factors could be more predictive of an individual having keloid scars. The results show that scar conditions and fibrosis of the skin (L905), burns and corrosion of the shoulder and upper limb, except wrist and hand (T22), trunk (T21), and ankle and foot (T25), were the most significant features found in clusters where keloid occurrence was most frequent. This study demonstrates the value of clustering analysis in identifying key factors associated with keloid scars, which can inform targeted prevention and management strategies in susceptible individuals, potentially reducing the incidence and burden of this challenging condition.