Decoding dynamic lipase trajectory patterns and in-hospital mortality in acute pancreatitis: insights from machine learning in intensive care units.

Background: Serum lipase levels are crucial biomarkers in acute pancreatitis (AP), yet their dynamic patterns and prognostic implications remain incompletely understood. This study aimed to identify distinct lipase trajectory phenotypes and evaluate their association with in-hospital mortality in AP patients.

Methods: We conducted a retrospective analysis of 834 AP patients from the MIMIC-IV database using latent class trajectory modeling (LCTM) to identify distinct lipase trajectory phenotypes. Cox regression models, adjusted for demographics, comorbidities, clinical therapies, and critical illness markers, were employed to assess the association between trajectory classes and in-hospital mortality.

Results: Three distinct lipase trajectory phenotypes were identified: Class 1 (n = 543) with consistently low levels, Class 2 (n = 51) with extremely high and variable levels, and Class 3 (n = 240) with moderately elevated levels. Class 2 patients were significantly older (66.8 ± 17.6 years) and had higher comorbidity burden (CCI: 5.6 ± 3.0). In-hospital mortality rates were 12.2%, 17.6%, and 19.2% for Classes 1, 2, and 3, respectively. After comprehensive adjustment, both Class 2 (HR: 2.21, 95% CI 1.04-4.71, p = 0.042) and Class 3 (HR: 1.61, 95% CI 1.08-2.40, p = 0.022) showed significantly higher mortality risk compared to Class 1.

Conclusions: Dynamic lipase trajectory patterns in AP patients demonstrate distinct phenotypes with significant prognostic value for in-hospital mortality. These findings suggest that monitoring lipase trajectories may enhance risk stratification and guide clinical management in AP patients.