This retrospective cohort study was conducted at a university-affiliated tertiary hospital in China between January 2009 and July 2024. We included hospitalised patients with bronchiectasis exacerbations, defined according to Hill et al. [19] Inclusion criteria were as follows: (a) age ≥18 years; and (b) availability of sputum or bronchoalveolar lavage fluid (BALF) specimens. Exclusion criteria were as follows: (a) poor-quality sputum sample; (b) absence of chest high-resolution computed tomography (HRCT); (c) co-isolation of Enterobacterales and Pseudomonas aeruginosa (P. aeruginosa); (d) incomplete data; and (e) initiation of mechanical ventilation within the first 24h of admission. Each patient was included only once; for patients with multiple hospital admissions during the study period, only the first eligible admission was used as the baseline for analysis.

Clinical data were extracted from the electronic medical records, including demographic characteristics (age and sex), comorbidities (e.g., hypertension, diabetes, COPD, pulmonary aspergillosis, and malignant tumour), baseline laboratory parameters obtained within the first 24h of admission, and microbiological findings based on respiratory specimens collected during the initial 24h of admission. Follow-up information, including vital status, was obtained through telephone contact and outpatient visits up to 31 August 2025, or until loss to follow-up or death, whichever occurred first.

The study was conducted in accordance with the Declaration of Helsinki (as revised in 2013). The study was approved by the Ethics Committee on Biomedical Research at West China Hospital of Sichuan University (approval No. 2022455). The requirement for informed consent was waived because of the retrospective design and the use of de-identified data.

Deep sputum or BALF specimens were collected for microbiological culture within the first 24h of admission. Sputum samples with <10 squamous epithelial cells per low-power field (LPF) and >25 white blood cells (WBCs)/LPF were considered acceptable for analysis [20]. Isolated pathogens included P. aeruginosa, Acinetobacter baumannii, Haemophilus influenzae, Staphylococcus aureus, Streptococcus pneumoniae, Enterobacterales, Aspergillus spp., and other organisms. Co-isolation was defined as the isolation of ≥2 of these microorganisms (excluding Candida spp.) from the same respiratory specimen.

Patients were classified into Enterobacterales and non-Enterobacterales groups according to respiratory culture results obtained during admission. Enterobacterales comprised Escherichia coli (E. coli), Klebsiella pneumoniae (K. pneumoniae), Serratia spp., and Enterobacter cloacae (E. cloacae). In subgroup analyses, Enterobacterales isolates were categorised as E. coli, K. pneumoniae, or other Enterobacterales (including Serratia spp. and E. cloacae).

Primary endpoints were all-cause and cause-specific mortality. Secondary endpoints were invasive mechanical ventilation (IMV) and non-invasive mechanical ventilation (NIMV) initiated >24h after hospital admission. Cause-specific mortality was defined as bronchiectasis-related death, determined from medical records and follow-up information, including death due to massive haemoptysis, pneumonia/respiratory infection, or respiratory failure primarily attributable to bronchiectasis or its exacerbation.

Cause-specific mortality was independently adjudicated by 2 investigators who were blinded to the study variables of interest, based on review of medical records and follow-up data, with disagreements resolved by a third senior investigator. Patients receiving only NIMV during hospitalisation were classified into the NIMV group, whereas those receiving IMV at any time were classified into the IMV group.

Statistical analyses were performed using R (version 4.2.3) and Stata (version 17.0). The Shapiro–Wilk test was used to assess the normality of continuous variables. Normally distributed data are presented as mean±SD, and non-normally distributed data as median (IQR). Between-group differences in continuous variables were assessed using the t test for normally distributed data and the Wilcoxon rank-sum test for non-normally distributed data. Categorical variables are expressed as number (percentage) and were compared using the χ2 test.

Propensity scores for Enterobacterales isolation were estimated using a logistic regression model including baseline demographic, comorbidity, microbiological, and laboratory variables. Patients were matched in a 1:1 ratio by nearest-neighbour matching without replacement, using a calliper width of 0.2 of the SD of the logit of the propensity score. Covariate balance was assessed using standardised mean differences, with an absolute SMD<0.10 indicating adequate balance (Supplementary Appendix, page 1). Propensity score matching was applied only to the mortality analyses to improve baseline comparability for long-term outcome assessment. IMV and NIMV were analysed in the full cohort because the number of ventilation events after matching was too limited for stable regression analyses in the matched sample.

We employed multivariable logistic regression to evaluate associations between Enterobacterales isolation and mechanical ventilation outcomes, and Cox proportional hazards models to assess associations between Enterobacterales and mortality. Overall and cause-specific survival were estimated using Kaplan–Meier curves. Patients lost to follow-up were treated as censored observations, and sensitivity analyses were undertaken to assess the robustness of the findings. Covariates were selected using least absolute shrinkage and selection operator (LASSO) regularisation, informed by clinical relevance (Supplementary Appendix, pages 2–3).

Restricted cubic spline (RCS) analyses were used to assess potential nonlinear associations of PNI with mortality and mechanical ventilation, with adjustment for the same covariates as in the corresponding multivariable models. Nonlinearity was assessed by testing the nonlinear spline terms. All models incorporated robust standard errors to account for potential model misspecification. Model adequacy was evaluated using variance inflation factors to assess multicollinearity, Hosmer–Lemeshow tests for logistic model calibration, and Schoenfeld residuals to test proportional hazards assumptions. Discrimination was assessed using the area under the receiver operating characteristic curve (AUC) for logistic models and Harrell C-index for Cox models. Sensitivity analyses included E-value calculations to quantify robustness to potential unmeasured confounding. A 2-sided P<.05 was considered statistically significant.

A total of 4634 hospitalised patients with bronchiectasis exacerbations were included in the study, of whom 228 (4.9%) were in the Enterobacterales group and 4406 (95.1%) were in the non-Enterobacterales group (Fig. S1 in the Supplementary Appendix). The median age was 58.6 years, and 51.5% of patients were male.

Pulmonary aspergillosis (6.1% vs 3.1%; P=.010) and malignancy (8.3% vs 4.0%; P=.001) were more prevalent in patients with Enterobacterales isolation. In the Enterobacterales group, the neutrophil-to-lymphocyte ratio (NLR), systemic immune-inflammation index (SII), and systemic inflammation response index (SIRI) were higher, whereas the prognostic nutritional index (PNI) was lower (43.77 vs 45.70; P<.001). IMV was more frequent in patients with Enterobacterales isolation (4.8% vs 1.0%; P<.001).

Detailed baseline characteristics are presented in Table 1. After 1:1 propensity score matching, baseline demographics, comorbidities, and laboratory parameters were well balanced between the 2 groups. During follow-up, cause-specific mortality was higher in the Enterobacterales group than in the non-Enterobacterales group (25.4% vs 18.0%; P=.053), although this difference did not reach conventional statistical significance (Table S1 in the Supplementary Appendix).

Pathogens were isolated from sputum in 4070 patients (87.8%) and from BALF in 564 (12.2%). Pseudomonas aeruginosa (P. aeruginosa) was the most frequently isolated bacterium, detected in 796 patients (17.2%), and carbapenem-resistant P. aeruginosa (CRPA) in 151 (3.3%). Other bacteria included H. influenzae in 82 patients (1.8%), Stenotrophomonas maltophilia in 44 (1.0%), Mycobacterium tuberculosis in 38 (0.8%), S. aureus in 32 (0.7%), nontuberculous mycobacteria in 12 (0.3%), and S. pneumoniae in 10 (0.2%).

Enterobacterales were isolated in 228 patients (4.9%). Within this group, the predominant species were Klebsiella pneumoniae (K. pneumoniae) in 116 patients (50.9%), Escherichia coli (E. coli) in 70 (30.7%), E. cloacae in 39 (17.1%), and Serratia spp. in 10 (4.4%). Co-isolation was markedly more prevalent in the Enterobacterales group than in the non-Enterobacterales group (16.7% vs 2.2%; P<.001). Candida glabrata and Candida tropicalis were also more common in the Enterobacterales than in the non-Enterobacterales group (3.5% vs 1.0% and 5.3% vs 1.6%, respectively; P=.001 and P<.001, respectively). Detailed microbiological characteristics are shown in Table 2. After 1:1 propensity score matching, microbiological characteristics in the matched cohort are presented in Table S2.

Multivariable logistic regression showed that Enterobacterales isolation was independently associated with IMV (adjusted odds ratio [aOR], 2.39; 95%CI, 1.21–6.87; P=.022), but not with NIMV (aOR, 1.01; 95%CI, 0.48–2.13; P=.977) (Table 3). Co-isolation (aOR, 4.67; 95%CI, 2.02–10.79; P<.001), mMRC dyspnoea score (aOR, 1.74; 95%CI, 1.49–2.03; P<.001), and NLR (aOR, 1.07; 95%CI, 1.04–1.11; P<.001) were also independently associated with increased odds of IMV, whereas PNI was associated with lower odds (aOR, 0.92; 95%CI, 0.89–0.95; P<.001).

For NIMV, COPD (aOR, 3.73; 95%CI, 2.69–5.17; P<.001), P. aeruginosa isolation (aOR, 1.86; 95%CI, 1.32–2.62; P<.001), mMRC dyspnoea score (aOR, 2.08; 95%CI, 1.87–2.32; P<.001), and higher ALT levels (aOR, 1.00; 95%CI, 1.00–1.01; P=.006) were associated with a higher risk of NIMV, whereas higher PNI (aOR, 0.92; 95%CI, 0.90–0.93; P<.001) and platelet-to-white blood cell ratio (PWR) (aOR, 0.97; 95%CI, 0.96–0.98; P<.001) were independently associated with a lower risk.

RCS analysis showed an inverse, approximately linear association between PNI and IMV (P for overall<.001; P for nonlinear=.159; Fig. S2A), and a nonlinear association between PNI and NIMV (P for overall<.001; P for nonlinear<.001; Fig. S2B). In the 2-piecewise logistic regression analysis, the estimated inflection point was 41. The association between PNI and NIMV was not significant below this threshold (OR, 0.970; 95%CI, 0.934–1.010; P=.122), whereas above this threshold, each 1-unit increase in PNI was associated with significantly lower odds of NIMV (OR, 0.847; 95%CI, 0.805–0.889; P<.001).

Model diagnostics showed no evidence of problematic multicollinearity (all VIFs≤1.35; Table S3). IMV and NIMV models showed excellent discrimination (AUC, 0.882 and 0.881, respectively; Tables S4 and S5). E-values for the associations of Enterobacterales isolation with IMV and NIMV are reported in Table S6. Logistic regression analyses for IMV and NIMV adjusted for treatment-related variables, including maintenance inhaled bronchodilator, macrolide, and systemic antibiotic therapy, are presented in Table S7.

During a median follow-up of 72 months (IQR, 34.5–132), 14 patients in each of the Enterobacterales and non-Enterobacterales groups were lost to follow-up. Overall, 133 patients in the matched cohort (n=456) (29.2%) died, including 99 cause-specific deaths (Table S1).

Kaplan–Meier analysis demonstrated lower cause-specific survival in the Enterobacterales group than in the non-Enterobacterales group (P=.041; Fig. 1). The 1-, 3-, 5-, and 7-year cause-specific survival rates were 94.7%, 86.4%, 80.9%, and 73.8%, respectively, in the Enterobacterales group, compared with 94.7%, 88.4%, 83.9%, and 81.0%, respectively, in the non-Enterobacterales group.

Fig. 1.

All-cause and cause-specific survival probabilities according to Enterobacterales isolation and species. (A) Cause-specific survival of Enterobacterales and non-Enterobacterales; (B) all-cause survival of Enterobacterales and non-Enterobacterales; (C) cause-specific survival stratified by E. coli, K. pneumoniae, other Enterobacterales, and non-Enterobacterales; (D) all-cause survival stratified by E. coli, K. pneumoniae, other Enterobacterales, and non-Enterobacterales.

In species-level analyses, E. coli was associated with the lowest survival. Both cause-specific (P=.043) and all-cause (P=.016) survival were lower in patients with E. coli isolation than in those with K. pneumoniae or other Enterobacterales. For E. coli, the 1-, 3-, 5-, and 7-year cause-specific survival rates were 95.7%, 81.7%, 76.1%, and 66.2%, respectively, and the all-cause survival rates were 95.7%, 77.3%, 67.2%, and 58.4%, respectively.

In multivariable Cox proportional hazards models, Enterobacterales isolation was independently associated with a higher risk of cause-specific mortality (adjusted hazard ratio [aHR], 1.49; 95%CI, 1.13–2.30; P=.035), whereas no significant association was observed with all-cause mortality (Table S8).

Older age (aHR, 1.05; 95%CI, 1.03–1.07; P<.001), COPD (aHR, 1.63; 95%CI, 1.06–2.51; P=.028), mMRC dyspnoea score (aHR, 1.25; 95%CI, 1.06–1.47; P=.007), and malignant tumour (aHR, 1.97; 95%CI, 1.10–3.52; P=.022) were independently associated with a higher risk of cause-specific mortality, whereas higher PNI was associated with a lower risk (aHR, 0.95; 95%CI, 0.92–0.97; P<.001).

In species-level Cox models, E. coli was consistently associated with an increased risk of both cause-specific mortality (aHR, 1.99; 95%CI, 1.13–3.50; P=.017) and all-cause mortality (aHR, 1.84; 95%CI, 1.18–2.87; P=.007), whereas K. pneumoniae and other Enterobacterales were not significantly associated with either outcome (Fig. 2; Tables S9–S11).

Fig. 2.

Univariable and multivariable Cox proportional hazards analyses of Enterobacterales and all-cause and cause-specific mortality. (A) Univariable Cox model for cause-specific mortality; (B) multivariable Cox model for cause-specific mortality; (C) univariable Cox model for all-cause mortality; and (D) multivariable Cox model for all-cause mortality. Multivariable models were adjusted for age, diabetes, COPD, malignant tumour, pulmonary aspergillosis, co-isolation, mMRC dyspnoea score, calendar year, and PNI.

Higher PNI was independently associated with a lower risk of both cause-specific mortality (aHR, 0.95; 95%CI, 0.92–0.97; P<.001) and all-cause mortality (aHR, 0.95; 95%CI, 0.93–0.97; P<.001). RCS analyses showed inverse associations of PNI with cause-specific and all-cause mortality, with no evidence of significant nonlinearity (P for nonlinear=.085 and .055, respectively; Fig. S3). Similar results were observed in the species-level models (P for nonlinear=.074 and .068, respectively; Fig. S4).

The pattern of covariate effects was similar in the species-level models, with older age, malignant tumour, and COPD associated with higher mortality risk. All Cox models satisfied the proportional hazards assumption and showed no evidence of problematic multicollinearity (mean VIF≈1.12–1.15; Tables S12 and S13). Binary and 4-category Enterobacterales models were well calibrated and demonstrated good discrimination, with Harrell C-index values of 0.785–0.791 for cause-specific mortality and 0.769–0.777 for all-cause mortality (Tables S14 and S17). E-value analyses were performed to assess the potential impact of unmeasured confounding on the associations between Enterobacterales (E. coli) and mortality (Tables S18 and S19). Cox regression analyses for cause-specific and all-cause mortality adjusted for treatment-related variables are shown in Tables S20–S22.