Validated prediction model for in hospital mortality in AECOPD with type 2 diabetes

Acute exacerbation of chronic obstructive pulmonary disease (AECOPD) frequently coexists with type 2 diabetes mellitus (T2DM) and is associated with substantial short-term mortality. However, patients with AECOPD and metabolic comorbidity represent a clinically heterogeneous subgroup, and admission-based tools with independent validation for early in-hospital mortality risk stratification in this population remain limited. In this multicenter retrospective cohort study, hospitalized patients with AECOPD and concomitant T2DM were consecutively enrolled and divided into a training cohort and an institutionally independent validation cohort from a separate hospital within the same city. Candidate predictors routinely available at admission were evaluated using multivariable logistic regression to develop a parsimonious prediction model for in-hospital mortality. Model performance was assessed in terms of discrimination, calibration, and clinical utility, and compared with established bedside scores, including the quick Sequential Organ Failure Assessment (qSOFA) and BAP-65. Elevated arterial carbon dioxide tension (PaCO₂), procalcitonin (PCT), and D-dimer measured early after admission were independently associated with in-hospital mortality. A simplified model incorporating these three variables demonstrated stable discrimination in both the training and validation cohorts, with an area under the receiver operating characteristic curve of approximately 0.79, and showed good calibration. Decision curve analysis indicated higher or non-inferior net clinical benefit across clinically relevant threshold probabilities compared with qSOFA and BAP-65. In hospitalized patients with AECOPD and T2DM, PaCO₂, PCT, and D-dimer were independently associated with in-hospital mortality. An admission-based model integrating these markers showed promising performance in multicenter data with independent hospital-based validation within the same city. Further validation in geographically distinct populations is needed before broader generalizability can be assumed.