Transparency.

Acute pulmonary embolism (APE) is a life-threatening condition requiring precise risk stratification. Although numerous prognostic factors have been proposed, redundancy and limited predictive utility often obscure clinical interpretation. To analyze a predefined set of clinical and laboratory variables in patients with APE using both classical statistical models and a novel taxonomic structural analysis, aiming to identify factors associated with early mortality beyond conventional outcome-based associations. We retrospectively analyzed 366 patients diagnosed with APE between 2009 and 2018, of whom 76 died within one year of the acute event. A total of 20 clinical and laboratory variables—including both established prognostic markers and features with no presumed direct impact on mortality—were assessed using Cox and logistic regression models with the concordance index (C-index) and Akaike’s Information Criterion (AIC). A structural analysis based on Marczewski–Steinhaus (M–S) taxonomic distances was applied to all 1,140 unique triads of risk factors to identify clusters of high patient variability. Segmented regression was then used to determine the transition between homogeneous and heterogeneous predictor spaces. Classical regression identified age as the strongest mortality predictor in APE. In contrast, the taxonomic outcome-agnostic approach revealed CRP as the most prominent structural signal, followed by other key inflammatory markers such as D-dimer, high-sensitivity troponin T (hsTnT), and activated partial thromboplastin time (aPTT). Age, along with certain hematological parameters (e.g., hemoglobin) and major electrolytes (Na ⁺ , K ⁺ , Cl⁻), appeared taxonomically insensitive to acute disease-related changes, reflecting more stable background characteristics. Several other variables, including renal biomarkers (urea, creatinine, and GFR), showed no significant role in APE, with their levels varying randomly between patients. Within this framework, CRP exhibits the highest structural variability among the analyzed factors, suggesting prognostic relevance beyond classical outcome-based associations (such as age). The proposed taxonomic approach complements traditional methods by reducing redundancy, enhancing interpretability, and improving the identification of truly relevant prognostic factors.