Trauma patients differ widely in injury pattern, physiological derangement, age and mechanism of injury. Conventional summaries such as the Injury Severity Score, or categories based on a single body region or physiological threshold, capture only part of this heterogeneity. This complicates comparisons between cohorts, the design of clinical studies, education and quality improvement. Data-driven stratification can describe how these dimensions combine in practice, but earlier work has been limited by small cohorts, restricted injury mechanisms or a lack of external evaluation. This PhD project aims to develop and evaluate a reproducible, clinically interpretable stratification of trauma presentations. We use the US National Trauma Data Bank (American College of Surgeons Trauma Quality Programs Participant Use File, admission years 2018–2019), with approximately 1.4 million development records of patients aged 15 years or older with documented injuries. A separate part of the data is held out for final assessment. Each presentation is described by its injury pattern derived from Abbreviated Injury Scale codes (anatomical location, severity and number of injuries), arrival physiology (systolic blood pressure, respiratory rate and Glasgow Coma Scale), age and injury mechanism. Physiological values are missing for 3–11% of records, partly for clinical reasons such as intubation before arrival. Missing values are handled with multiple imputation, with sensitivity analyses for data missing not at random. Candidate stratifications are fitted with distance-based clustering (partitioning around medoids) across alternative injury representations and numbers of groups. Candidates are evaluated on stability under resampling and imputation, coverage of clinically important presentations, and clinical interpretability, rather than on a single statistical index. The selected procedure is then frozen and applied to the held-out data. The expected result is a documented, reproducible stratification of trauma presentations, with openly available code, that can support cohort description and comparison in trauma research and quality improvement. The project forms part of a compilation thesis at Karolinska Institutet. Principal investigator: Jonatan Attergrim, doctoral student, Department of Global Public Health, Karolinska Institutet. Main supervisor: Martin Gerdin Wärnberg, Department of Global Public Health, Karolinska Institutet.