Prognostic factors for severe community-acquired pneumonia in high altitude areas

Abstract

High altitude and hypoxia may impact the prognosis of critically ill patients. Currently, research on the outcomes of patients with severe community-acquired pneumonia (SCAP) in high-altitude regions remains limited. This study was conducted to determine the 30-day mortality rate and identify associated risk factors in SCAP patients on the Qinghai-Tibet Plateau. We prospectively enrolled 183 adults with SCAP admitted to Lhasa People’s Hospital between 1 January 2023 and 31 December 2024. Patients were divided into survival group and non-survival group based on the survival status within 30 days. Demographic data, clinical features, laboratory indicators, and treatment measures were compared between these two groups. Multivariate Cox regression analysis was used to identify risk factors for 30-day mortality and survival curves were drew in patients with SCAP. Receiver operating characteristic (ROC) curve was used to evaluate the predictive value of relevant risk factors for 30-day mortality, and compare the differences of predictive value among different factors. For model development and validation, the entire cohort was randomly split into a training set (70% of n) and a validation set (30% of n) using a computer-generated random sequence. Among the 183 SCAP patients enrolled, 56 died within 30 days, with a 30-day mortality rate of 30.6%. Multivariate Cox regression analysis showed that age ≥ 65 years (Hazard Ratio [HR] = 1.849, 95% Confidence Interval (CI): 1.012 ~ 3.379, P = 0.046), septic shock (HR = 4.340, 95% CI 1.845 ~ 10.208, P = 0.001), arterial partial pressure of oxygen/fraction of inspired oxygen ratio(P/F ratio) < 150 mmHg (HR = 3.333, 95% CI 1.866 ~ 5.952, P < 0.001), D-dimer > 3.0 mg/L (HR = 1.965, 95% CI 1.044 ~ 3.699, P = 0.036), and unknown etiology (HR = 2.391, 95% CI 1.319 ~ 4.335, P = 0.004) were independent risk factors for 30-day mortality. In the training set (n = 126, 70%), the prediction model integrating these five indicators had a sensitivity of 78.6% and specificity of 82.7% for predicting 30-day mortality (Receiver Operating Characteristic-Area Under the Curve [ROC-AUC] = 0.884, 95% CI 0.829 ~ 0.927, P < 0.001), superior to CURB-65 score (AUC = 0.673), Pneumonia Severity Index (PSI) score (AUC = 0.695), Acute Physiology and Chronic Health Evaluation (APACHE) II score (AUC = 0.783) and Sequential Organ Failure Assessment (SOFA) score (AUC = 0.725). In the validation set (n = 57, 30%), the prediction model had a sensitivity of 80.0% and a specificity of 85.7% for predicting 30-day mortality (AUC = 0.862, 95% CI 0.760 ~ 0.964, P < 0.001), superior to CURB-65 score (AUC = 0.637), PSI score (AUC = 0.571), APACHE II score (AUC = 0.769), and SOFA score (AUC = 0.723). Advanced age (≥ 65 years), septic shock, decreased P/F ratio, elevated plasma D-dimer levels, and unknown etiology are independent risk factors for 30-day mortality in patients with SCAP. The combination of these five risk factors can effectively predict the occurrence of death within 30 days, with a predictive value better than CURB-65 score, PSI score, APACHE II score and SOFA score.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-36609-9.

Introduction

Community-acquired pneumonia (CAP), defined as an infectious inflammation of the lung parenchyma occurring outside hospital settings, represents a prevalent respiratory infection. Although in-hospital mortality of CAP has shown a declining trend over the past decade¹, the 30-day mortality rate for severe community-acquired pneumonia (SCAP) remains alarmingly high, ranging from 23 to 50%^2–4. High-altitude regions, typically defined as areas above 2500 m, present unique environmental challenges including hypobaric hypoxia, low oxygen saturation, extreme cold, and arid conditions. These factors collectively impair respiratory defense mechanisms, leading to increased susceptibility to infections and consequently higher pneumonia incidence rates. The Qinghai-Tibet Plateau, recognized as the "Roof of the World" and the “Third Pole” with an average elevation exceeding 4000 m, exemplifies such extreme high-altitude conditions. Previous research by Meilang et al.⁵ investigated the clinical characteristics and etiological distribution of CAP patients in Tibet, highlighting the ongoing therapeutic challenges in managing CAP at high altitudes. More recent studies from Qinghai province have demonstrated that Chronic Obstructive Pulmonary Disease (COPD) patients developing CAP in high-altitude regions experience more severe disease manifestations and prolonged clinical courses. Despite these findings, there remains a notable paucity of global data specifically addressing SCAP in high-altitude populations. Current literature has predominantly focused on analyzing severe cases associated with coronavirus disease (COVID-19)^6–10, with no reported studies examining non-COVID-19 related SCAP in these regions. Furthermore, comprehensive data on SCAP in high-altitude areas is currently lack in China. Therefore, the primary objective of this prospective study was to identify the independent risk factors associated with 30-day mortality in patients with SCAP admitted to a tertiary hospital on the Qinghai-Tibet Plateau. As a secondary, exploratory aim, we sought to develop a preliminary prediction model based on these identified factors and compare its performance with conventional severity scores.

Objects and methods

Research subjects and data collection

We prospectively enrolled consecutive adult patients with SCAP who were admitted to the Lhasa People’s Hospital between January 2023 and December 2024. Based on their 30-day survival status, these patients were categorized into survival and non-survival groups. Inclusion criteria included: (1) permanent residents of the Qinghai-Tibet Plateau aged ≥ 18 years; (2) availability of complete clinical data. Exclusion criteria comprised: (1) pregnancy; (2) concurrent active malignancy; (3) history of acute myocardial infarction, stroke, major trauma, or major surgery within the preceding month.

Patient demographics and clinical data (including medical history, exposure history, symptoms, signs, treatment interventions, complications, and 30-day survival outcomes) were extracted from the electronic medical record system, along with laboratory data collected within the first 24 h after admission. For patients discharged with clinical improvement within 30 days, 30-day survival status was confirmed via telephone follow-up. The P/F ratio was calculated using the formula: arterial partial pressure of oxygen (PaO₂)/fraction of inspired oxygen (FiO₂). For high-altitude regions, the P/F ratio was adjusted by the formula: PaO₂/FiO₂ × (760/atmospheric pressure)¹¹. Body mass index was computed as weight (kg)/height (m)². The detailed selection process is visually represented in Fig. 1 through a comprehensive flowchart.

Fig. 1.

Flow diagram of SCAP patient’s enrollment.

This study was approved by the Ethics Committee of Lhasa People’s Hospital (Approval No. SYLL2124057) and complied with the ethical principles of the Declaration of Helsinki (1964) and its subsequent amendments or comparable ethical standards. Written informed consent was obtained from all participating patients and/or their legal guardians.

Diagnostic criteria and definitions

The diagnosis of CAP was made based on the 2019 American Thoracic Society (ATS)/Infectious Diseases Society of America (IDSA) clinical practice guidelines¹². The diagnostic criteria included: (1) community onset; (2) clinical features consistent with pneumonia (including at least one of the following: (a) new or worsening cough, with or without sputum production, chest pain, dyspnea, or hemoptysis; (b) fever; (c) signs of consolidation and/or audible crackles on auscultation; (d) leukocytosis (> 10.0 × 10⁹/L) or leukopenia (< 4.0 × 10⁹/L), with or without a left shift); and (3) new radiographic infiltrates on chest imaging (e.g., patchy opacities, lobar or segmental consolidation, ground-glass opacities, or interstitial changes), with or without pleural effusion. A diagnosis of CAP was made when the patient fulfilled criteria (1) and (3), along with at least one item from criterion (2), while excluding other conditions such as pulmonary tuberculosis, lung tumors, non-infectious interstitial lung disease, and pulmonary edema. The diagnostic criteria for SCAP were also based on the 2019 ATS/IDSA clinical practice guidelines¹², requiring either one major criterion or ≥ 3 minor criteria. Major criteria included: septic shock requiring vasopressor therapy or respiratory failure necessitating mechanical ventilation. Minor criteria comprised: respiratory rate ≥ 30 breaths/min, P/F ratio ≤ 250 mmHg, multilobar infiltrates, confusion/disorientation, azotemia (blood urea nitrogen [BUN] > 7 mmol/L), leukopenia (white blood cell count < 4.0 × 10⁹/L due to infection), thrombocytopenia (platelet count < 100 × 10⁹/L due to infection), hypothermia (core temperature < 36.0 °C), and hypotension requiring adequate fluid resuscitation. Septic shock was defined according to the Third International Consensus Definitions for Sepsis and Septic Shock (Sepsis-3)¹³. Disease severity was assessed using the CURB-65 score (incorporating confusion, BUN > 7 mmol/L, respiratory rate ≥ 30 breaths/min, systolic blood pressure < 90 mmHg, and age ≥ 65 years), Pneumonia Severity Index (PSI) score, Acute Physiology and Chronic Health Evaluation (APACHE) II score, and Sequential Organ Failure Assessment (SOFA) score^14,15. All enrolled patients underwent comprehensive etiological testing of respiratory specimens (throat swabs/sputum/bronchoalveolar lavage fluid) and/or hematological pathogen identification (including conventional culture and/or multiple PCR and/or next-generation sequencing) prior to antibiotic administration. Antibacterial agents were strictly selected based on pathogen culture results and antimicrobial susceptibility testing. For critically ill patients without identified pathogens, empiric antimicrobial therapy was guided by experienced respiratory specialists in accordance with established clinical guidelines. All diagnostic evaluations, complication assessments, comorbidity determinations, and severity scoring were performed by attending physicians.

Statistical analysis

Data analysis was performed using SPSS 23.0 (Statistical Package for the Social Sciences, Chicago, IL, USA) and R software version 4.3.3. The normality of continuous variables was assessed using the Shapiro–Wilk test. Normally distributed data are presented as mean ± standard deviation (x̅ ± s) and were compared between groups using the t-test. Non-normally distributed data are presented as median M (Q1, Q3) and were compared using the Mann–Whitney U test. Categorical data are expressed as number (percentage) and were compared using the χ² test or Fisher’s exact test. Variables showing significant differences between survival and non-survival groups in univariate analyses, along with factors that may be related to dependent variables from the perspective of professional knowledge, were included in a multivariate Cox proportional hazards regression model to identify risk factors associated with prognosis in patients with SCAP. Results were expressed as hazard ratios (HR) with 95% confidence intervals (CI). The sensitivity and specificity of risk factors for predicting prognosis were evaluated using receiver operating characteristic (ROC) analysis. The areas under the ROC curves (AUC) were compared using the method described by DeLong et al.¹⁶ to assess the diagnostic performance of different models. To enhance the clinical utility of the prediction model, a nomogram was constructed based on the identified predictors. Calibration curves were generated using the bootstrap method with 1000 resamples to evaluate the agreement between predicted probabilities and observed outcomes; a well-calibrated model would display points aligning along the 45-degree diagonal line. For model development and validation, the entire cohort was randomly split into a training set (70% of n) and a validation set (30% of n) using a computer-generated random sequence. The multivariate Cox regression model was developed in the training set to identify independent prognostic factors and to build the nomogram. Prediction Model Construction and Nomogram: The prediction model was derived from a multivariable Cox proportional hazards regression analysis. The nomogram was constructed to represent this model, with point assignments for each variable determined by a linear transformation of their corresponding regression coefficients (β values). The model’s performance in predicting 30-day mortality in SCAP was then assessed in the validation set by plotting calibration curves and ROC curves. Survival curves were plotted using the Kaplan–Meier method and compared using the log-rank test.

Risk stratification based on the nomogram was performed exclusively within the training set to avoid overfitting and model-dependent bias. After calculating the total nomogram score for each patient in the training set, we used tertiles (i.e., the 33rd and 66th percentiles) as cut points to categorize patients into low-, intermediate-, and high-risk groups. This tertile-based division was intended to create groups with approximately equal numbers of patients, enabling balanced comparisons of survival outcomes.

Results

Demographic and clinical characteristics of SCAP patients in survival and non-survival groups

The study enrolled 183 SCAP patients (109 males and 74 females) with a median age of 73 years (IQR 61–80). The cohort comprised 176 Tibetan patients (96.2%), 5 Han Chinese patients, and 2 patients of other ethnicities (Fig. 1). According to diagnostic criteria, 90 patients (49.2%) met major SCAP criteria while 93 patients (50.8%) fulfilled minor criteria. Continuous Renal Replacement Therapy (CRRT) was administered to 7 patients (3.8%). Fifty-six patients died during the follow-up period, with a 30-day mortality rate of 30.6% (56/183).

Comparative analysis revealed that non-survivors demonstrated significantly higher proportions of septic shock and invasive mechanical ventilation (IMV) requirement, significantly less etiology-evidence, along with significantly lower P/F ratio, adjusted P/F ratio, and serum albumin levels compared to survivors. The death group exhibited significantly prolonged prothrombin time and elevated fasting blood glucose, BUN, serum creatinine, plasma D-dimer levels, CURB-65 scores, PSI scores, APACHE II scores, and SOFA scores (all P < 0.05). No statistically significant differences were observed in age, gender distribution, ethnic composition, altitudinal gradient, body mass index, smoking prevalence, comorbidities (chronic respiratory diseases, hypertension, coronary artery disease, cerebrovascular diseases, diabetes mellitus), proportions of COVID-19 or influenza pneumonia, incidence of fungal co-infections, white blood cell counts, platelet counts, hemoglobin levels, C-reactive protein, N-terminal pro-brain natriuretic peptide (NT-proBNP), left ventricular ejection fraction, and utilization rates of corticosteroids and immunoglobulins, as detailed in Table 1 (all P > 0.05).

Table 1.

Demographic and clinical characteristics of surviving versus deceased SCAP patients.

Characteristic	Total (N = 183)	Survival group (n = 127)	Non-survival group (n = 56)	P
Age (years)	73 (61, 80)	73 (62, 80)	70 (60, 78)	0.547
BMI (kg/m2)	21.8 (19.6, 24.4)	22.0 (19.8, 24.7)	21.1(18.9, 24.0)	0.125
Male, n(%)	109 (59.6)	77 (60.6)	32 (57.1)	0.658
Tibetan nationality, n(%)	176 (96.2)	122 (96.1)	54 (96.4)	0.905
Altitude	3650(3650, 4200)	3650(3650, 4200)	3650(3650, 4200)	0.231
Altitudinal gradient*, n (%)				0.387
2500 ~ 3650 m [n(%)]	16 (8.7)	11 (8.7)	5 (8.9)
3650 m [n(%)]	90 (49.2)	57 (44.9)	33 (58.9)
3650 ~ 4500 m [n(%)]	65 (35.5)	51 (40.2)	14 (25.0)
> 4500 m [n(%)]	12 (6.6)	8 (6.3)	4 (7.1)
Current smoker, n (%)	21 (11.5)	14 (11.0)	7 (12.5)	0.773
Symptom-to-admission time (days)	7 (3, 10)	7 (3, 14)	6 (4, 10)	0.744
COVID-19, n (%)	35 (19.1)	27 (21.3)	8 (14.3)	0.269
Influenza pneumonia, n (%)	14 (7.7)	9 (7.1)	5 (8.9)	0.896
Concurrent Fungal Infection, n (%)	35 (19.1)	27 (21.3)	8 (14.3)	0.269
Fulfills ATS/IDSA major criteria for SCAP, n(%)	90 (49.2)	44 (34.6)	46 (82.1)	< 0.001
Comorbidities
CRD n (%)	20 (10.9)	13 (10.2)	7 (12.5)	0.651
Hypertension	71 (38.8)	50 (39.4)	21 (37.5)	0.811
Coronary artery disease, n (%)	5 (2.7)	2 (1.6)	3 (5.4)	0.340
Cerebrovascular disease, n (%)	10 (5.5)	6 (4.7)	4 (7.1)	0.756
Diabetes mellitus, n (%)	14 (7.7)	6 (4.7)	8 (14.3)	0.052
Disorder of consciousness, n (%)	26 (14.2)	15 (11.8)	11 (19.6)	0.162
Septic shock, n (%)	70 (38.3)	28 (22.0)	42 (75.0)	< 0.001
Laboratory data
P/F ratio (mmHg)	167 (122, 217)	179 (142, 225)	124 (90, 183)	< 0.001
White blood cells (× 109/L)	8.1 (5.3, 12.0)	8.2 (5.5, 11.3)	8.0 (4.3, 14.9)	0.830
Platelets (× 109/L)	159 (96, 226)	167 (106, 234)	144 (82, 212)	0.075
Hemoglobin (g/L)	144 (126, 164)	146 (128, 167)	140 (124, 156)	0.109
C-reactive protein (mg/L)	74 (22, 154)	72 (19, 133)	85 (27, 181)	0.168
Fasting blood glucose (mmol/L)	6.7 (5.4, 9.4)	6.2 (5.3, 7.6)	8.3 (5.6, 11.6)	0.004
Blood urea nitrogen (mmol/L)	7.2 (5.1, 10.6)	6.7 (4.9, 9.8)	9.4 (6.0, 14.9)	0.002
Serum creatinine (μmol/L)	75.8 (58.4, 106.9)	70.4 (57.0, 92.4)	92.5 (61.0,187.3)	0.003
Serum albumin (g/L)	31.1 ± 5.6	31.8 ± 5.4	29.5 ± 5.9	0.010
D-dimer (mg/L)	3.0 (1.7, 6.6)	2.4 (1.4, 5.2)	4.3 (2.8, 12.2)	< 0.001
Prothrombin Time (s)	13.9 (12.6, 15.8)	13.5 (12.6, 15.6)	14.2 (13.0, 15.9)	0.046
N-terminal pro-brain natriuretic peptide (pg/mL)	1280 (430, 4230)	1021 (398, 3980)	2028 (597,4726)	0.190
LVEF (%)	62.0 (59.0, 67.0)	62.5 (60.0, 67.0)	61.0 (57.3, 67.5)	0.154
IMV, n (%)	61 (33.3)	24 (18.9)	37 (66.1)	< 0.001
Glucocorticoid therapy, n (%)	65 (35.5)	40 (31.5)	25 (44.6)	0.087
IVIG, n (%)	10 (5.5)	5 (3.9)	5 (8.9)	0.310
Etiology-evidence (%)	87 (47.5)	67 (52.8)	20 (35.7)	0.033
Scores associated with disease
CURB-65 score	2 (1, 2)	1 (1, 2)	2 (1, 3)	< 0.001
PSI score	106 (90, 128)	100 (86, 120)	122 (99, 141)	< 0.001
APACHE II score	12 (9, 18)	10 (8, 13)	19 (12, 25)	< 0.001
SOFA score	4 (3, 7)	4 (2, 6)	7 (4, 10)	< 0.001

*Altitudinal gradient was defined as the residential altitude and categorized as follows: 2500–3650 m, 3650 m, 3650 ~ 4500 m, and > 4500 m.

SCAP severe community-acquired pneumonia, BMI body mass index, COVID-19 coronavirus disease 2019, CRD chronic respiratory disease, LVEF left ventricular ejection fraction, IMV invasive mechanical ventilation, IVIG intravenous immunoglobulin, P/F ratio arterial partial pressure of oxygen/fraction of inspired oxygen ratio, CURB-65 confusion, urea nitrogen, respiratory rate, blood pressure, and age ≥ 65, PSI pneumonia severity index, APACHE II acute physiology and chronic health evaluation II, SOFA sequential organ failure assessment.

Multivariable cox regression analysis of 30-day mortality in the training cohort

To develop the prediction model, univariable and multivariable Cox regression analyses were performed on the training set (n = 126). In the final multivariable model, five independent predictors of 30-day mortality were identified: age ≥ 65 years (HR = 1.97, 95% CI 1.03 ~ 4.16, P = 0.045), presence of septic shock (HR = 3.69, 95% CI 1.48 ~ 9.20, P = 0.005), P/F ratio < 150 mmHg (HR = 3.17, 95% CI 1.55 ~ 6.51, P = 0.002), D-dimer level > 3.0 mg/L (HR = 2.51, 95% CI 1.11 ~ 5.66, P = 0.027), and pneumonia of unknown etiology (HR = 2.91, 95% CI 1.39 ~ 6.11, P = 0.005). The results of the complete regression analysis for the training set are presented in Table 2.

Table 2.

Univariable and Multivariable Cox regression analysis of 30-Day Mortality in the Training Set of SCAP patients.

Factors	Univariable HR (95% CI)	P value	Multivariable HR (95% CI)	P value
Age ≥ 65 years	1.155 (0.589, 2.263)	0.675	1.969 (1.033, 4.155)	0.045
Septic Shock	5.693 (2.843, 11.399)	< 0.001	3.691 (1.480, 9.203)	0.005
P/F ratio < 150 mmHg	4.366 (2.223, 8.576)	< 0.001	3.174 (1.548, 6.509)	0.002
Fasting blood glucose ≥ 7 mmol/L	2.007 (1.078, 3.738)	0.028	1.540 (0.792, 2.994)	0.203
Blood urea nitrogen ≥ 7 mmol/L	2.537 (1.329, 4.846)	0.005	1.371 (0.604, 3.113)	0.451
Serum creatinine ≥ 97 μmol/L	2.887 (1.562, 5.336)	0.001	1.348 (0.588, 3.087)	0.481
Serum albumin ≥ 30 g/L	1.926 (1.042, 3.559)	0.037	1.193 (0.582, 2.442)	0.630
D-dimer > 3.0 μg/mL	3.737 (1.829, 7.634)	< 0.001	2.507 (1.109, 5.664)	0.027
Prothrombin time ≥ 14.0 s	2.355 (1.246, 4.451)	0.008	1.465 (0.731, 2.936)	0.282
IMV	4.195 (2.215, 7.946)	< 0.001	1.407 (0.595, 3.326)	0.437
Unknown Etiology	2.035 (1.067, 3.883)	0.031	2.909 (1.385, 6.110)	0.005

SCAP severe community-acquired pneumonia; IMV invasive mechanical ventilation. P/F ratio arterial partial pressure of oxygen/fraction of inspired oxygen ratio.

Prediction model development and validation

The entire cohort of 183 patients was randomly split into a training set (n = 126, 70%) for model development and an internal validation set (n = 57, 30%). The baseline characteristics, including demographics, comorbidities, laboratory parameters, disease severity scores, and clinical interventions, were well-balanced between the training and validation sets, with no statistically significant differences observed (all P > 0.05; Supplementary Table 1). The final multivariable Cox regression model, incorporating the five independent predictors identified from the training set, was used to construct a prognostic nomogram for estimating 30-day survival probability.

Prognostic nomogram development and its performance in the training set

Based on the results of the multivariate Cox regression analysis from the training set, we constructed a nomogram for the individualized prediction of 30-day mortality risk (Fig. 2). The nomogram (Fig. 2) visually demonstrates the association between five independent predictors (age ≥ 65 years, septic shock, P/F ratio < 150 mmHg, D-dimer > 3.0 mg/L, and unknown etiology) and 30-day survival status in SCAP patients. For clinical application, vertical lines are drawn upward from each variable’s value to determine corresponding points on the point scale. The sum of these five variables’ points yields a total score, from which a vertical line drawn downward predicts the individual’s 30-day survival probability.

Fig. 2.

Prognostic Nomogram of 30-Day Mortality for SCAP patients in the Training Set. This nomogram was developed based on multivariate Cox regression analysis from the training set, incorporating five independent predictors: Age ≥ 65 years, Septic shock, P/F ratio < 150 mmHg, D-dimer > 3.0 µg/mL, and Unknown Etiology. SCAP severe community-acquired pneumonia; P/F ratio arterial partial pressure of oxygen/fraction of inspired oxygen ratio.

The prediction model demonstrated excellent discriminative ability in the training set, with an AUC of 0.884 (95% CI 0.829 ~ 0.927) (Fig. 3), a sensitivity of 78.6%, and a specificity of 82.7%. Its predictive performance was superior to that of conventional scoring systems: CURB-65 score (AUC = 0.673, 95% CI 0.600 ~ 0.741; P < 0.001 for curve comparison), PSI score (AUC = 0.695, 95% CI 0.623 ~ 0.761; P < 0.001 for curve comparison), APACHE II score (AUC = 0.783, 95% CI 0.717 ~ 0.841; P = 0.016 for curve comparison), and SOFA score (AUC = 0.725, 95% CI 0.654 ~ 0.788; P < 0.001 for curve comparison). The detailed predictive performance of each individual variable in the model is presented in Table 3.

Fig. 3.

ROCs for SCAP patients in the Training Set. (A) ROC curves of individual predictors (Age ≥ 65 years, Septic shock, P/F ratio < 150 mmHg, D-dimer > 3.0 µg/mL, Unknown etiology) and the combined prediction model. (B) Comparative ROC analysis between the prediction model and conventional severity scores (CURB-65, PSI, APACHE II, SOFA). The predicted model incorporates five variables listed in panel A. APACHE II acute physiology and chronic health evaluation II, AUC area under the curve, PSI pneumonia severity index, ROC receiver operating characteristic, SCAP severe community-acquired pneumonia, SOFA sequential organ failure assessment, P/F ratio arterial partial pressure of oxygen/fraction of inspired oxygen ratio, CURB-65 confusion, urea nitrogen, respiratory rate, blood pressure, age ≥ 65.

Table 3.

Predictive performance of variables and scores for 30-day mortality in the training set.

Variable	Cut-off	Sensitivity (%)	Specificity (%)	AUC	95%CI	P
Model Variables
Age	≥ 65 years†	62.5	52.0	0.572	0.492 ~ 0.668	0.086
Septic Shock	Yes†	75.0	78.0	0.765	0.697 ~ 0.824	< 0.001
P/F ratio	< 150 mmHg†	64.3	70.9	0.676	0.603 ~ 0.743	< 0.001
D-dimer	> 3.0 mg/L†	73.2	58.3	0.657	0.584 ~ 0.726	< 0.001
Unknown Etiology	Yes†	62.5	52.0	0.572	0.497 ~ 0.645	0.066
Prediction model	125 points*	78.6	82.7	0.884	0.829 ~ 0.927	< 0.001
Conventional Scores
CURB-65 score	1*	75.0	52.0	0.673	0.600 ~ 0.741	< 0.001
PSI score	117*	57.1	73.2	0.695	0.623 ~ 0.761	< 0.001
APACH II score	13*	69.6	78.0	0.783	0.717 ~ 0.841	< 0.001
SOFA score	6*	53.6	79.5	0.725	0.654 ~ 0.788	< 0.001

APACHE II acute physiology and chronic health evaluation II, AUC area under the receiver operating characteristic curve, CURB-65 confusion, urea nitrogen, respiratory rate, blood pressure, age ≥ 65, PSI pneumonia severity index, SCAP severe community-acquired pneumonia, SOFA sequential organ failure assessment, P/F ratio arterial partial pressure of oxygen/fraction of inspired oxygen ratio.

^†Cut-off was pre-defined as the variable’s inclusion criterion in the regression model.

*Optimal cut-off value was determined by maximizing Youden’s index.

Predictive performance of individual variables and the nomogram

The discriminatory performance of each predictor included in the final model, along with conventional severity scores, was quantitatively assessed in the training set and is summarized in Table 3. The sensitivity and specificity for the continuous severity scores (CURB-65, PSI, APACHE II, and SOFA) were calculated based on their optimal cut-off values determined by maximizing the Youden’s index. Our newly developed prediction model achieved a significantly higher AUC of 0.884 (95% CI 0.829 ~ 0.927) compared to all individual variables and conventional scores.

Predictive performance of the nomogram in the validation set

To evaluate the model’s generalizability, we applied this nomogram to an independent validation set comprising 57 SCAP patients. The results showed that the model maintained good predictive accuracy, with an AUC of 0.862 (95% CI 0.760 ~ 0.964), a sensitivity of 80.0%, and a specificity of 85.7%. Its predictive performance was significantly superior to that of conventional scoring systems: CURB-65 score (AUC = 0.637, 95% CI 0.498 ~ 0.760; P = 0.007 for curve comparison), PSI score (AUC = 0.571, 95% CI 0.433 ~ 0.702; P = 0.001), APACHE II score (AUC = 0.769, 95% CI 0.660 ~ 0.826; P = 0.046), and SOFA score (AUC = 0.723, 95% CI 0.600 ~ 0.802; P = 0.023). Furthermore, the calibration curve in the validation set (Fig. 4) demonstrated good agreement between the model-predicted probabilities and the actual observed outcomes.

Fig. 4.

Calibration Plot for the Predictive Nomogram Model in the Validation Set. Calibration curve for the 30-day mortality prediction model in the validation set. The plot assesses the nomogram based on the five-variable prediction model. SCAP severe community-acquired pneumonia, P/F ratio arterial partial pressure of oxygen/fraction of inspired oxygen ratio.

Stratified survival analysis based on points of nomogram

Based on the total points derived from the nomogram, patients in the training cohort were categorized into three distinct risk groups using tertiles as the cut-off values: low-risk (total points < 66, n = 68), intermediate-risk (total points 66 ~ 141, n = 54), and high-risk (total points > 141, n = 61).

Kaplan–Meier survival analysis demonstrated a significant and graded difference in 30-day survival among these three groups (Log-rank test, P < 0.001). The observed mortality rates increased sharply across the risk strata, confirming the effective risk stratification capability of the nomogram. The corresponding survival curves are presented in Fig. 5.

Fig. 5.

Survival Curves of SCAP patients (Kaplan–Meier method; Log-rank test). Survival curves for patients with severe community-acquired pneumonia (SCAP), stratified into low-, medium-, and high-risk groups based on the nomogram total points. (P < 0.001). Risk stratification was performed by tertiles of the nomogram score. SCAP severe community-acquired pneumonia.

Discussion

In this study conducted at high altitude, we primarily investigated the risk factors for mortality in SCAP patients. This study revealed a 30-day mortality rate of 30.6% among SCAP patients in high-altitude regions. Age ≥ 65 years, septic shock, P/F ratio < 150 mmHg, D-dimer > 3.0 mg/L and unknown etiology were independent risk factors for mortality. A combined prediction model incorporating these five markers demonstrated superior predictive accuracy compared to conventional severity scores: CURB-65 score, PSI score, APACHE II score, and SOFA score, supporting its clinical utility for risk stratification in this population.

The high-altitude environment, characterized by low atmospheric pressure and chronic hypoxic, predisposes individuals to severe hypoxemia, which can exacerbate respiratory failure and increase the risk of organ dysfunction risks. Hypoxia-induced inflammatory responses further aggravate lung injury through systemic inflammation¹⁷. Additionally, the arid climate and significant diurnal temperature variations in high-altitude areas may impair respiratory mucosal defenses, increasing susceptibility to severe respiratory infections. These mechanisms collectively influence both the incidence and clinical progression of SCAP. Rogelio et al.¹⁸ first reported increased mortality related to H1N1 influenza pneumonia at high altitudes in Mexico. Although studies from the Qinghai-Tibet Plateau had described the etiology and clinical characteristics of CAP^5,19, few had specifically addressed outcomes for SCAP patients in high-altitude. Recent literature has focused on COVID-19, with conflicting results regarding the impact of altitude on prognosis^6–8. Santiago et al. reported increased COVID-19 mortality in Colombia’s high-altitude regions compared to lowland areas⁸, while others report better outcomes, possibly due to adaptive physiological responses^6,7. The 30-day mortality among post-COVID-19 SCAP patients in high-altitude regions remains poorly defined. Our observed mortality rate of 30.6% is consistent with reports from lowland regions^2–4.

Advanced age (≥ 65 years) is a recognized risk factor for SCAP and is significantly associated with poor clinical outcomes²⁰, a finding consistent with the results of this study. Elderly patients often present with declined immune function, multiple comorbidities, and reduced physiological reserve, which collectively contribute to diminished tolerance to severe infections and a more complicated clinical course²¹. Furthermore, in high-altitude hypoxic environments, the cardiopulmonary burden in elderly patients is further exacerbated, potentially worsening hypoxemia and multi-organ dysfunction, thereby increasing mortality risk²². These results align with previous studies identifying age as a core predictor of adverse outcomes in CAP, which has been incorporated into widely used clinical scores such as PSI and CURB-65²³. In high-altitude regions, the interaction between environmental factors and age-related physiological decline may further elevate risk. Therefore, special attention should be given to elderly populations in the assessment of SCAP, warranting more intensive monitoring and proactive intervention strategies.

Septic shock, a critical complication and key diagnostic criterion for SCAP, is associated with high mortality rates^4,12. Ambient hypoxia in high-altitude regions markedly reduces arterial oxygen content in humans. Whether arising from long-term residency, rapid ascent, chronic or acute hypoxia, or conditions such as septic shock, the microcirculation consistently demonstrates characteristic dysregulation. This pathological state is characterized by a severe systemic inflammatory response and hemodynamic collapse, in which diminished tissue perfusion pressure induces cellular hypoxia and multi-organ dysfunction¹³. In high-altitude environments, chronic hypoxemia may amplify infection-induced inflammatory cascades¹⁷, establishing a vicious cycle of progressive organ failure and consequently higher incidence and mortality of septic shock²⁴. Notably, even with comprehensive management including appropriate antibiotics, source control, and hemodynamic support, Tibetan patients with septic shock demonstrate persistently elevated mortality (65.7%)²⁴. Our study confirms septic shock as an independent risk factor for 30-day mortality in high-altitude SCAP (HR = 4.340), with significantly reduced cumulative survival rates observed in affected patients.

The P/F ratio reflects the severity of impaired gas exchange. In high-altitude environments, pre-existing hypoxemia compounds oxygenation deficits in SCAP, leading to markedly lower P/F ratio. Values P/F ratio ≤ 150 mmHg are associated with systemic inflammatory, mitochondrial dysfunction^17,25, and multi-organ failure²⁶. Compared to inhabitants of low-altitude regions, populations residing at high altitudes demonstrate elevated systemic levels of hypoxia-inducible factor-1α (HIF-1α) and heat shock protein 70 (HSP70). The inhibition of hydroxylase activity under hypoxic conditions modulates inflammatory responses through alterations in the interleukin-1β (IL-1β) signaling pathway, leading to an upregulation of pro-inflammatory cytokines²⁷. Reduced P/F ratio has previously been correlated with worse outcomes in high-altitude COVID-19 patients^28,29. Similarly, we found that P/F ratio < 150 mmHg was as an independent predictor of 30-day mortality (HR = 3.333), underscoring the need for aggressive respiratory support in this population.

D-dimer, a fibrin degradation product, indicates activated coagulation and fibrinolysis. Elevated levels of D-dimer are common across all CAP etiologies^30–32, and are particularly pronounced in SCAP, correlating with systemic inflammation, coagulopathy, thrombosis and poor outcomes^30–32. High-altitude residents exhibit higher baseline D-dimer levels and are have increased risk of thromboembolic events³³. Previous rodent studies under simulated high-altitude conditions also revealed a prothrombotic phenotype with platelet hyperreactivity and elevated P-selectin³⁴. This suggests that the combined environmental factors at high altitude—including hypoxia, low temperatures, dehydration, and low atmospheric pressure—contribute to raised D-dimer levels, heightened thrombotic risk and other poor outcomes¹⁷. Hypoxia exacerbates inflammatory and prothrombotic pathways^35–37, leading to disproportionately elevated D-dimer levels. In our cohort, median D-dimer was 3.0 mg/L, higher than typically reported in lowland SCAP populations^30–32, and levels of D-dimer > 3.0 mg/L were an independent risk factor for 30-day mortality (HR = 1.965), highlighting its role as both a biomarker and potential mediator of adverse outcomes.

The findings of this study also highlight the critical role of etiological identification in determining the prognosis of patients with SCAP. Multivariate Cox regression analysis demonstrated that unknown etiology was an independent risk factor of 30-day mortality (HR = 2.391). This finding indicates that the failure to identify a specific pathogen is associated with adverse clinical outcomes, and is consistent with previous studies indicating that accurate pathogen identification is significantly associated with improved survival³⁸. Etiological diagnosis provides considerable value, contributing not only to early risk assessment, but more importantly, enabling targeted antimicrobial therapy. Multiple studies have demonstrated that pathogen-directed treatment significantly reduces mortality in SCAP patients^12,39. From a clinical perspective, rapid and accurate etiological diagnosis facilitates individualized antibiotic therapy, thereby mitigating the drawbacks of empirical treatment. As emphasized by Mandell et al.⁴⁰, the unnecessary use of broad-spectrum antibiotics may exacerbate antimicrobial resistance. The current results further support the importance of a pathogen-oriented treatment strategy in reducing mortality, which is consistent with the recent IDSA/ATS guidelines⁴¹.

In summary, we identified five independent clinical predictors of 30-day mortality in high-altitude SCAP patients^17,35. A nomogram integrating these factors demonstrated superior discriminative ability compared to conventional scores and effectively stratified patients into distinct risk groups with significantly different survival outcomes. This enhanced performance likely stems from the model’s direct incorporation of key high-altitude pathophysiology, such as profound hypoxemia and coagulopathy, which are not fully captured by existing generic severity scores.

However, this single-center study has inherent limitations. First, as a single-center observational study with a limited sample size (n = 183), the findings may be influenced by potential biases. Second, although standardized management protocols were implemented—including early goal-directed therapy for septic shock, appropriate antimicrobial therapy, nurse-driven protocols for mechanical ventilation and sedation, and individualized thromboprophylaxis—the application and timing of advanced organ support varied. For example, continuous renal replacement therapy (CRRT) was used in only 3.8% of patients (n = 7), and extracorporeal membrane oxygenation (ECMO) was unavailable due to resource constraints. These discrepancies in advanced support may have confounded individual outcomes, a limitation inherent to the local context. Third, as a single-center study conducted in a specific high-altitude region, the generalizability of our findings to other high-altitude settings with different medical resources, population characteristics, or microbial ecology may be limited. Multicenter studies involving diverse high-altitude populations are needed to validate and extend these results. Finally, although our institution is a tertiary referral center, it remains challenging to disentangle the effects of hypobaric hypoxia from other regional factors—such as health-seeking behaviors, prehospital delays, or variability in early antibiotic treatment. The interplay between altitude-related physiological adaptations and health-system factors warrants further investigation. Future studies should assess the applicability of these predictors across various high-altitude environments and explore their integration with established scoring systems to develop more accurate prognostic models, thereby improving clinical decision-making and outcomes in SCAP patients residing at high altitude.

In conclusion, the distinct high-altitude environment may limit the applicability of conventional prognostic scoring systems in SCAP patients. In contrast, the combination of advanced age (≥ 65 years), septic shock, decreased P/F ratio(< 150 mmHg), elevated D-dimer levels (> 3.0 mg/L), and etiology-evidence demonstrates superior predictive accuracy, likely by better reflecting region-specific pathophysiology.

Author contributions

YYZ designed the study, collected clinical data, analyzed the data, and wrote the manuscript. BX, LY, XS and YTZ designed the study, and wrote the manuscript. NC helped manage the research, performed the statistical analyses, and revised the paper.

Funding

This study supported by the Natural Science Foundation of Tibet Autonomous Region (XZ2024ZR-ZY033[Z] to Dr Na Cui).

Data availability

All data analyzed during the study are presented in the main manuscript and Supplementary file. In addition, the anonymous dataset is available from the corresponding author upon reasonable request.

Declarations

Competing interests

The authors declare no competing interests.

Ethics approval and consent to participate

The study was approved by the Ethics Committee of Lhasa People’s Hospital (Approval No. SYLL2124057) and complied with the ethical principles of the Declaration of Helsinki (1964) and its subsequent amendments or comparable ethical standards. Written informed consent was obtained from all participating patients and/or their legal guardians.