This single-center, retrospective cohort study was conducted in patients with SCAP, aged ≥ 18 years who were admitted to the ICU of a tertiary academic hospital from January 2017 to December 2019. Patients were excluded if they (1) were suspected of having hospital-acquired pneumonia (14); (2) permanently resided in a nursing home; or if (3) the initial diagnosis of SCAP was not confirmed during their ICU stay.

This study was approved by the institutional review board of the China-Japan Friendship Hospital (2019-80-K52) and was performed in accordance with the Helsinki Declaration. The requirement for informed consent was waived because this non-interventional study collected data from previous electronic medical records and did not involve personal privacy or commercial interests.

SCAP was defined as meeting either one major criterion or at least three minor criteria of the Infectious Diseases Society of America/American Thoracic Society criteria (11). Immunosuppression was defined based on consensus, determined by meeting one of the following criteria (15): primary immune deficiency disease; active malignancy; receiving cancer chemotherapy; HIV infection with CD4 T-lymphocyte count < 200 cells/μL or percentage < 14%; solid organ transplantation; hematopoietic stem cell transplantation; receiving corticosteroid therapy with a prednisone dose of 20 mg or equivalent daily for ≥14 days or a cumulative dose >700 mg; receiving biologic immune modulators; or receiving disease-modifying anti-rheumatic or other immunosuppressive drugs.

Microbiological tests were performed within 48 h of ICU admission in all patients, using bronchoalveolar lavage fluid (BALF) or endotracheal aspirates. Microbiological tests included bacterial and fungal smear and culture, acid-fast stain, and real-time polymerase chain reaction for cytomegalovirus (CMV), Pneumocystis jirovecii (PJ), influenza virus, respiratory syncytial virus, adenovirus, Legionella pneumophila, Chlamydia pneumoniae, and Mycoplasma pneumoniae. Metagenomic next-generation sequencing (mNGS) was performed, if necessary, at the clinician's discretion (more details see Additional file 1).

Pathogens were identified by clinicians based on microbiological tests, clinical manifestations, and chest radiology findings. Atypical pathogens included Legionella, Mycoplasma, and Chlamydia (16). Polymicrobial infection was defined as having more than one type of pathogen diagnosed by clinicians within 48 h of ICU admission.

Clinical data, including demographics, comorbidities, season of onset, vital signs, fraction of inspired oxygen (FiO₂), respiratory support, arterial blood gas analyses, Acute Physiology and Chronic Health Evaluation (APACHE) II scores, Sequential Organ Failure Assessment (SOFA) scores, laboratory results on ICU admission, pathogens detected within 48 h of ICU admission, and clinical outcomes were extracted from the electronic medical record system. The clinical outcomes included the requirement for invasive mechanical ventilation (IMV), IMV duration, length of ICU stay, mortality within 7 days, and ICU mortality.

Continuous variables were expressed as means and standard deviations or medians and interquartile ranges. Categorical variables were presented as frequencies and percentages. Differences between characteristics, testing results, pathogens detected, treatments, and outcomes of immunocompetent and immunocompromised patients were tested for statistical significance using t-tests for relatively symmetrically distributed continuous variables, Wilcoxon rank-sum test for asymmetrically distributed continuous variables, and Pearson's chi-square test for categorical variables. Two-tailed P < 0.05 were considered statistically significant. All statistical analyses were performed using IBM SPSS software, version 22 (IBM Corp., Armonk, NY, United States).

To assess for potential confounders, clinically important variables such as age, lymphocyte count, and CD4 cell count that demonstrated possible statistical associations in the univariate analysis (P < 0.1) were transformed from continuous variables into categorical variables according to clinical reference values and standards.

To identify the risk factors of ICU mortality, variables with P < 0.1 in the univariate analyses were entered into a multivariable logistic regression model, using stepwise backward selection based on the likelihood ratio. Variable entry was set at P < 0.05, and variable removal at P > 0.1.

Of the 457 patients with SCAP who were screened, 64 were excluded based on the exclusion criteria, leaving a total of 393 patients in the analysis, including 274 (69.7%) immunocompetent and 119 (30.3%) immunocompromised patients. The three most common immunocompromising conditions were corticosteroid therapy (61/119, 51.2%), immunosuppressive therapy (28/119, 23.5%), and active malignancy (16/119, 13.4%) (Supplementary Table S1).

Compared to immunocompetent patients, immunocompromised patients included significantly more females (45.4 vs. 30.3%, P = 0.004) and patients with chronic lung disease (31.9 vs. 20.8%, P = 0.018) (Table 1). The season of disease onset differed significantly (P = 0.009) according to immunocompromised status and was mainly in the spring and autumn (49.6%) among immunocompromised patients and the winter among immunocompetent patients (42.3%). The white blood cell (9.1 vs. 10.2, P = 0.030), neutrophil (7.9 vs. 8.9, P = 0.005), lymphocyte (0.5 vs. 0.7, P = 0.044), platelet (150.0 vs. 175.0, P = 0.038), and CD4 T cell (201.5 vs. 286.0, P = 0.001) counts; and hemoglobin (104.0 ± 21.5 vs. 115.6 ± 25.0, P < 0.001), albumin (29.0 vs. 31.0, P < 0.001), and procalcitonin (0.8 vs. 1.2, P = 0.020) levels on ICU admission were significantly lower in immunocompromised patients than in immunocompetent patients (Supplementary Table S2). Other baseline laboratory results, SOFA scores, and APACHE II scores were similar between patients in the two groups (Table 1). The mean time from SCAP onset to ICU admission in immunocompetent and immunocompromised patients was 8.0 and 7.0 days, respectively (P = 0.617).

The use of IMV during ICU stay was less frequent in immunocompromised patients (54/119, 45.4%) than in immunocompetent patients (164/274, 60.1%) (P = 0.003); however, the duration of IMV (5.0 vs. 8.0 days; P = 0.176) and ICU stay (8.0 vs. 9.0 days; P = 0.182) were similar between the two groups (Table 1). The rates of mortality within 7 days (26.1 vs. 13.1%, P = 0.002) and ICU mortality (49.6 vs. 37.6%, P = 0.027) were higher in immunocompromised patients (Table 1).

Pathogens were identified in 83.2% (99/119) of immunocompromised patients and 69.0% (189/274) of immunocompetent patients (P = 0.003) (Table 1). Immunocompromised patients had fewer atypical pathogen infections than immunocompetent patients (4.0 vs. 15.3%, P = 0.004) (Table 2). Influenza virus infection was more prevalent in immunocompetent patients than in immunocompromised patients (51.3 vs. 15.1%, P < 0.001). However, immunocompromised patients had a higher prevalence of infection caused by viruses other than influenza virus (49.5 vs. 12.2%, P < 0.001) and cytomegalovirus (47.5 vs. 3.2%, P < 0.001) than immunocompetent patients (Table 2). Fungal infections were more prevalent in immunocompromised patients (70.7 vs. 23.3%, P < 0.001) (Table 2). Immunocompromised patients had a higher frequency of polymicrobial infections (56.6 vs. 27.5%, P < 0.001), especially mixed viral-fungal infections (62.5 vs. 42.3%, P = 0.036) (Table 2).

The details of the pathogens identified are shown in Figure 1 and Supplementary Table S3. The five most frequent pathogens identified in immunocompromised patients were PJ (55/99, 56%), CMV (47/99, 47%), Aspergillus (23/99, 23%), Staphylococcus aureus (9/99, 9%), and influenza A virus (9/99, 9%) (Figure 1). Influenza A virus was the most common pathogen (82/189, 43%) in immunocompetent patients. PJ, nontuberculous mycobacteria (2/99, 2%), and Nocardia spp. (2/99, 2%) were detected only in immunocompromised patients. Adenovirus (10/189) and Streptococcus pneumoniae (9/189) were detected only in immunocompetent patients (Supplementary Table S3).

Among immunocompromised patients, the univariate analysis showed that 11 variables were significantly associated with ICU mortality, including age, SOFA score, lymphocyte count, platelet count, D-dimer level, fibrinogen level, PaO₂/FiO₂, FiO₂ group, lactate level, CD4 cell count, and chronic kidney disease (Table 3, Supplementary Table S4). In addition to these variables, the APACHE II score, neutrophil count, and IMV on ICU admission were included in the multivariable logistic regression model (Supplementary Table S4). The final multivariable logistic regression model found that age ≥65 years, SOFA score, lymphocyte count < 0.8 × 10⁹/L, D-dimer level, 0.5 ≤ FiO₂ < 0.7, FiO₂ > 0.7, and serum lactate level (were independent risk factors for ICU mortality in immunocompromised patients (Table 4).

The univariate analysis of ICU mortality in immunocompetent patients revealed 13 variables associated with an increased risk of ICU mortality (Supplementary Table S5). In the multivariable analysis, only 3 factors were independent risk factors for ICU mortality in immunocompetent patients (Supplementary Table S6): days from illness onset to ICU admission (OR: 1.101, 95% CI: 1.039–1.166, P = 0.001), platelet count < 100 × 10⁹/L (OR: 2.729, 95% CI: 1.191–6.256, P = 0.018), and FiO₂ > 0.7 (OR: 3.372, 95% CI: 1.433–7.932; P = 0.005). Only FiO₂ > 0.7 was also a risk factor for ICU mortality in immunocompromised patients (OR: 3.372 vs. 10.228 in immunocompetent and immunocompromised patients, respectively).

In the multivariable analysis including all patients, immunocompromised status (OR: 2.043, 95% CI: 1.114–3.748, P = 0.021) age ≥65 years, days from illness onset to ICU admission, neutrophil count, platelet count < 100 × 10⁹/L, CD4 T cell count < 200 cells/μL, FiO₂ > 0.7, and lactate level were independent risk factors for ICU mortality (Table 5). Among them, three risk factors were the same as those for immunocompromised patients: age ≥ 65 years (OR: 4.267 vs. 9.098), FiO₂ > 0.7 (OR: 4.110 vs. 10.228), and serum lactate level (OR: 1.324 vs. 4.849).