A prediction model for hospital mortality in patients with severe community-acquired pneumonia and chronic obstructive pulmonary disease

To the best of our knowledge, this is the first practical prediction model to identify patients at risk of death in those with SCAP and COPD specifically. Our model integrates various basic clinical characteristics, including age, comorbidities, vital signs and laboratory examinations, indicating that the comprehensive evaluation based on these predictors is essential. The C indices and time-dependent AUCs of the nomogram when applied to the training and testing cohorts were similar and both approximately 0.8, which demonstrated the performance was relatively ideal. The score can be calculated by hand according to nomogram with routine parameters tested in the laboratory. Hence, it is rapid, cost-effective and can be easily implemented in clinical practice.

The independent risk factors of death in CAP patients with COPD varied widely in the existing literature. For example, Bonnesen et al. included 243 CAP patients with COPD and found that the factors related to mortality were age, premorbid condition, CURB-65 score, pleural effusion and multi-lobular infiltrate [19]. In another research, aspiration (OR 5.203; 95% CI 1.443, 18.757), D-dimer > 2.0 µg/mL (OR 5.026; 95% CI 1.395, 18.108) and CURB-65 ≥ 3 (OR 23.299; 95% CI 6.246,  86.903) were risk factors of in-hospital mortality in 230 CAP patients comorbid with COPD [20]. Multilobar pneumonia (OR 2.883; 95% CI 1.299–6.399), Pseudomonas aeruginosa pneumonia (OR 19.091; 95% CI 4.326–84.256) and high-risk PSI classes (OR 10.316; 95% CI 1.691–62.946) were also found to be independent risk factors for case-fatality rate in a prospective cohort of CAP patients with COPD [21]. Moreover, Shin et al. found the serum hemoglobin concentration (HR 0.759; 95% CI 0.616, 0.936) and albumin level (HR 0.429; 95% CI 0.185, 0.995) were significantly associated with 180-day mortality in 134 acute exacerbation of COPD (AECOPD) patients with CAP [22]. The inconsistency regarding diverging results across prior studies could be attributable to a combination of factors such as study design, population, severity of CAP and treatments. The study from Cilli et al. only included CAP patients in the ICU [9]. However, researchers did not assess the prediction performances of risk factors. Besides, few prior studies focused on the weight of each risk factors for outcomes. Therefore, it is likely that this study has several advantages or more important clinical implications compared with previous studies. First, we had a larger population with only SCAP patients in ICU enrolled, which is representative of the real-world pneumonia patient cohort that has the highest mortality. Then, it has been suggested that the biomarkers are a cornerstone in the management of SCAP to decrease treatment failure [23]. Considering that the combination of biomarkers would be of greater use than individual predictor, we developed a prediction model. Afterwards, we also evaluated and validated the model with several statistical methods. Third, there is no consensus on the optimal cut-off values of these predictors in SCAP patients. Therefore, they were included in the model as continuous variables. Moreover, we carefully investigated the prognostic accuracies and clinical utilities of them via correlation analysis and restricted cubic splines.

Patients with advanced age, chronic renal diseases, decreased systolic blood pressure, elevated BUN are also classified as high-risk population when conventional score calculations are applied in SCAP patients, such as CURB-65, pneumonia severity index (PSI), Sequential Organ Failure Assessment (SOFA) score and Acute Physiology and Chronic Health Evaluation (APACHE) II tool. As observed clinically and previously reported, our nomograms show that diabetes, a common comorbidity of COPD, is associated with worse prognosis. Several factors might be responsible for the mechanisms. Previous evidence suggested that both COPD and impaired lung function, especially restricted ventilation dysfunction, could increase the risk of diabetes as a consequence of systemic inflammatory processes [24]. In addition, treatment with corticosteroids in COPD could possibly lead to a variety of side effects, such as worsening hyperglycemia and deterioration of diabetes control [25]. And reversely, diabetes can worsen the prognosis of COPD due to the direct effects of hyperglycemia on lung physiology, inflammation and susceptibility to bacterial infection [26]. Moreover, diabetes is potentially associated with a wide spectrum of complications which negatively affect the prognosis of COPD, such as pulmonary hypertension [27]. Therefore, careful evaluation and management should be conducted in SCAP patients with COPD and diabetes due to the possible poor prognosis. IL-6 is involved in various hematopoietic, immune, and inflammatory responses. Therefore, it has been widely used as an early sensitive prognostic biomarker and a predictor of treatment failure and mortality in CAP [28]. He et al. found that IL-6 (hazard ratio [HR] 1.001; p = 0.001) could serve as independent predictors of 30-day mortality for CAP after adjusting for clinical data, including age, bilateral lung infection, procalcitonin, CURB-65, PSI, etc. [29]. Similarly, as an inflammatory marker and coagulation factor which is synthesized by hepatocytes and circulating in the bloodstream, the concentrations of fibrinogen are rapidly elevated in tissue injury, infection, inflammation, etc. It could also be used in the CAP severity evaluation [30]. Their prognostic values have also been investigated in COPD. In a meta-analysis with 61 studies in COPD, increased levels of IL-6 were associated with hospitalization (standardized mean difference [SMD] 0.12, 95%CI 0.04–0.20) and higher levels of fibrinogen were also associated with exacerbation (SMD 0.23 g/dL, 95%CI 0.14–0.33) and mortality (HR 3.13 per twofold increase, 95%CI 2.14–4.57) [31]. Zhou et al. conducted another meta-analysis with 45 studies and found a graded, concentration-dependent, significant relation between higher circulating fibrinogen and more severity of COPD [32]. Hence, it is plausible that elevated admission IL-6 and fibrinogen both are associated with hospital mortality in SCAP patients with COPD.

Some factors, such as increased creatinine and Troponin T, were associated with the mortality in univariate analysis. Nonetheless, the associations disappeared when adjusting for other risk factors. However, we should be cautious when explaining this conclusion because results from the existing literature on patients with SCAP or COPD are inconsistent with regard to whether they are associated with survival [33,34,35,36]. Future studies should address whether they could improve the evaluation and prediction of outcomes in SCAP patients with COPD.

The existing reports believed that risk stratification and early identification might contribute to optimizing the management of SCAP, with potential reduction of mortality [37]. Early assessment via prediction model could be instrumental to quantify in advance an individual patient’s risk of death when planning the therapies. On the other hand, the identification of patients at highest risk is pivotal to implement early measures and improve prognosis. The nomograms could be utilized as a complementary tool for decision making in clinical practice, or for SCAP-COPD patient selection in future studies on the basis of their risk stratification using the risk grouping. However, we acknowledge that some issues remain to be addressed. First, in our study, the diagnosis of COPD may lack strictness. It was difficult to determine the severity of COPD patients or to stratify them according to exacerbation histories, lung functions and symptoms from the data available. Thus, the identified independent risk factors need to be confirmed in COPD patients with different clinical characteristics. In addition, the nomogram might also have decreased predicting value in some specific subgroups of COPD patients. Then, the patients in present study are a little older (median age: 77 years old) compared to those SCAP patients in previous observational studies [19,20,21,22]. One leading cause is that we only strictly included confirmed SCAP-COPD patients because COPD is considered as an age-related disease. However, it is worth noting that different baseline characteristics existing among studies could result in diverse conclusions. Thus, large-scale, multicenter, prospective studies are desirable to validate, recalibrate, improve discriminative capacity and increase the generalizability of our prediction model. Besides, further information is needed to shed light on deeper understanding of pathophysiological mechanisms of SCAP patients with COPD. For instance, more efforts could be dedicated to investigate the impacts of various coexisting medical conditions, such as chronic cardiovascular diseases and diabetes, on the mortality of SCAP patients with COPD. Future researches should also consider the prognostic effects of more pre-admission individual features, including smoking status, vaccination history, prior antibiotic treatment and corticosteroid use, etc. Meanwhile, it is still controversial whether identified pathogens or imaging findings are related to the severity or mortality in those patients.

The main limitation of the current study is the single-center, retrospective design with selection bias. Then, the missing data might have reduced the effective sample size, caused inevitable bias and threatened the validity of the study. Third, although a number of potential risk factors have been analyzed, we cannot exclude that some unadjusted confounders could have affected the results or some untested variables would further improve the model. The model can be updated when more multicentric data become available.

留言 (0)

沒有登入
gif