Abstract Rationale Endobronchial lesions are frequently documented during urgent ICU bronchoscopy, yet their prognostic value remains uncertain. We evaluated whether specific bronchoscopic patterns and rapid detection of carbapenemase-producing Klebsiella pneumoniae (KPC) were associated with mortality in invasively ventilated adults. Methods Single-center retrospective pilot cohort of consecutive ICU patients under invasive mechanical ventilation who underwent bedside flexible bronchoscopy with bronchoalveolar lavage (BAL). Bronchoscopic abnormalities were recorded using a standardized template. Microbiologic evaluation combined culture and a multiplex PCR panel reporting pathogens and resistance determinants (including blaKPC). The primary endpoint was ICU mortality; secondary analyses explored lesion-pathogen relationships. Univariable tests used Fisher’s exact or Mann-Whitney U as appropriate. A two-variable logistic regression model (KPC status; mechanical power ≥18 vs 18) was prespecified. Discrimination was assessed using ROC AUC. Ethics approvals: CEISH-UG-25-014; protocol IESS-HG-BA-DTMC-2019-0263-M; expedited waiver for de-identified analysis. Results Thirty patients were included (median age, 60 years IQR 51-71; 63% male). ICU mortality was 26.7% (8/30). Any pathogen was detected in 73.3%, and KPC in 20.0%. Bronchoscopic findings were ubiquitous: edema (93.3%), mucosal congestion (86.7%), lumen narrowing (76.7%), mucopurulent or copious secretions (96.7%), ulcers (56.7%), and necrosis (36.7%). KPC detection was strongly associated with mortality (deceased 62.5% vs survivors 9.1%; Fisher p = 0.003). Age (OR 1.08 per year; 95% CI 1.01-1.19; p = 0.042) and lower leukocyte count (OR 0.79 per 1 × 10³/µL; 95% CI 0.63-0.97; p = 0.029) were also associated with death, whereas isolated endoscopic lesions were not. Necrosis correlated with mycobacterial detection (post-hoc adjusted p≈0.02). In the prespecified logistic model, KPC remained significant (OR 15.83; 95% CI 2.05-122.07; p = 0.008), while mechanical power was non-significant. Calibration was adequate (Hosmer-Lemeshow p≈0.999) and discrimination acceptable (AUC 0.776; 95% CI 0.587-0.907; p = 0.006). Post-disinfection surveillance cultures were negative after bundle processing. Conclusions In mechanically ventilated ICU patients, the presence of blaKPC in BAL—not individual bronchoscopic lesions—was the strongest mortality signal, together with age and leukopenia. These findings suggest that coupling urgent bronchoscopy with rapid molecular resistance profiling may enhance early risk stratification and guide antimicrobial and infection-control decisions. Endoscopic necrosis appeared more reflective of pathogen class (mycobacteria) than of outcome in this pilot study. Multicenter prospective validation is warranted to test integrated models that combine clinical, endoscopic, microbiologic, and immune markers to guide precision management in severe respiratory failure. This abstract is funded by: no funding
Briones-Zamora et al. (Fri,) studied this question.