Introduction
Sepsis is a lethal syndrome of physiologic, pathologic, and biochemical abnormalities induced by infection, one of the major global public health concerns [
1]. Although extensive research has recently demonstrated the mechanism and treatment of sepsis, sepsis is still the principal cause of death in intensive care patients worldwide [
2]. For the evaluation of organ dysfunction or failure, there is currently the Sequential Organ Failure Assessment (SOFA) or Quick Sequential Organ Failure Assessment (qSOFA), which does not include pulmonary function other than with a respiratory rate [
3]. Patients might be impacted with acute respiratory failure (ARF) if they fulfilled either diagnostic criteria. Notably, early diagnosis and treatment can help reduce mortality.
Therefore, acute respiratory failure was the common sepsis-associated organ injury resulting in critical mortality [
4‐
6]. ARF is due to severe dysfunction of pulmonary ventilation. Moreover, a decrease in the arterial partial pressure of oxygen (PaO2) is a sign of pulmonary dysfunction in patients with sepsis [
7]. According to accumulating studies, acute respiratory failure had a 60% mortality rate in the ICU, an average hospital stay is 7.1 days, and related medical costs are up to US$54 billion annually in the USA [
8,
9]. Clinical risk factors, pathobiology, response to treatment, and elements of pulmonary recovery have been extensively studied, which has improved the prevention, detection, and treatment of acute respiratory with sepsis [
10,
11].
Nevertheless, the pathogenesis of sepsis-related acute respiratory failure is still in an ongoing phase of exploration, and the different sources of risk factors make it a significant clinical challenge for early detection [
12]. In recent studies, the disease characteristics of SA-ARF patients have been used to identify increased risks. However, most of them have not combined these with clinical prediction models [
13,
14]. Nevertheless, few studies have focused on patients with sepsis complicated by ARF. Thus, the purpose of this study by an extensive clinical database is to evaluate the impact of SA-ARF on the 30-day mortality and to develop a predictive nomogram for predicting the probability of 30-day mortality in patients with SA-ARF.
Discussion
This retrospective analysis study extracted clinical data from an extensive database named MIMIC IV, which has more than 500,000 medical records from 2008 to 2019. Using Cox regression models, we conducted the risk factors related to 30-day mortality of ARF with sepsis, which included age, DBP, lactate, bilirubin, Pao2, SAPS II, CHF, and the use of mechanical ventilation. We established a prognostic nomogram for these patients in the ICU. The nomogram can be used to predict and diagnose. As far as we know, this is the first study to evaluate the risk factors associated with 30-day mortality about ARF with sepsis in the ICU and establish a nomogram.
An inflammatory response, immune suppression, and oxidative stress led to sepsis. Moreover, the effects of an inappropriate response to this infection resulted in impaired cellular function, mitochondrial dysfunction, and ultimately acute respiratory failure [
18]. Although the mechanisms by which sepsis leads to respiratory failure are still unclear, the development of these mechanisms can be balanced with the variables in our model [
19,
20]. Therefore, this model may have important implications for the development of acute respiratory failure in sepsis.
Acute respiratory failure is a common complication in patients with sepsis. Early diagnosis of ARF is usually confirmed by clinical manifestations, radiographs, CT, and pulmonary function, which are cumbersome operational steps [
21]. In recent years ultrasound technology has also played an essential role in diagnosing ARF patients in the ICU [
22]. What is more, sepsis is usually diagnosed by SOFA score or qsofa score [
23]. However, few studies have mentioned association between ARF and short-term mortality in sepsis. Based on this, it is vital for clinicians to conduct a thorough evaluation of the risk of death from sepsis and to objectively estimate the risks and benefits of medical interventions so that patients and families can make medical decisions with a careful assessment of the impact of potential treatment options. This will not only prevent medical disputes but also reduce certain medical costs. Thus, prediction nomograms are crucial for improving the risk stratification process of sepsis and can be used by clinicians to provide clear, accurate information to families of patients with SA-ARF. Therefore our nomogram used six factors that were easily accessible and could be collected on the first day of admission. We hope that this chart will enable us to identify sepsis with acute respiratory failure quickly.
Our study's initial vital signs include diastolic blood pressured and PaO2, which were considered an independent risk factor for patients with acute respiratory failure in SA-ARFs, the same result as the previous study [
24]. Age was widely recognized as the most powerful risk factor for organ failure in ICU. Increasing age was a major determinant of organ failure and overall mortality [
25,
26]. A large cohort study proved that patients who died with sepsis also tended to be older adults [
27]. In summary, age is strongly associated with poor prognosis in patients with acute respiratory failure in sepsis. Lactate was more dominant in the nomogram, which was the most significant predictor of mortality at 30 days in patients with sepsis-associated acute respiratory failure. Lactate levels can indicate the severity of the underlying disease. Moreover, it has been shown that elevated lactate levels may predict death in critically ill patients, while reduced lactate levels have been reported to be associated with improved clinical outcomes [
28,
29]. Jean-Louis Vincent et al. indicated that lactate levels are associated with prognosis in sepsis and acute respiratory failure patients [
30,
31]. Consistently, our study found that lactate is an Independent predictor for 30-day mortality of SA-ARF. In one sense, sepsis can be seen as a race to the death between the pathogens and the host immune system. The exaggerated response can lead to multi-organ failure (MOF), especially respiratory failure. As we all know, the interaction of platelets with immune cells and endothelial cells is an anti-infective response.
Nevertheless, during sepsis, the mechanisms become dysregulated and contribute to organ damage [
32]. One experiment proved that thrombocytopenia exacerbated the inflammatory response to sepsis and increased mortality [
33]. These experimental results are consistent with the results of our study. Besides, during sepsis, most patients require ventilation therapy. According to a randomized clinical trial, mechanical ventilation and conservative oxygen therapy are associated with poor outcomes in patients with sepsis [
34]. In short, these six factors are all predictors of mortality in respiratory failure. They are all clinically accessible nomograms that have been widely used in cancer research, and are increasingly being utilized in the prediction of diseases [
35,
36]. The nomogram was developed to survive acute respiratory failure in the elderly at 28 days, 60 days, and one year [
37]. For the mortality of SA-ARF, our study is the first to propose a simple nomogram to predict its mortality. The six included predictors were readily available on the first day of ICU admission. Furthermore, following calibration by internal validation to the actual 30-day mortality, we found high agreement between the training set and the validation set. In this cohort, the C-index was more than 0.7, and the calibration analysis performed in two cohorts revealed that the predicted 30-day mortality was similar. In addition, all decision curves in the cohort have net benefit rates above 50 percent.
Furthermore, there are two points to consider when using the nomogram. In our study, vital signs are calculated based on the average of each ICU patient's first 24 h in the ICU. Accordingly, the nomogram is not relevant to patients who die within 24 h of admission or who leave the intensive care unit within 24 h. In addition, laboratory tests included in the nomogram are the first results to be obtained from the ICU; consequently, all the laboratory tests included in the nomogram should be finished within 24 h of admission to the ICU.
The paper has some limitations despite these inspiring findings. First, this retrospective study returned the necessary recall bias, so more prospective studies need to be done to confirm this further. Second, since the values for vital signs and laboratory tests are for the first 24 h of admission to the ICU, the model does not apply to patients who die within 24 h of admission to the ICU. Third, with regard to data processing, we discarded variables with missing values greater than 40%; the retained variables were then supplemented by means of interpolation, multiple interpolation, and model predictions. In this way, the results of the analysis may have been biased, and the C-index may have been less accurate. Finally, we have only performed internal validation, so a larger sample size is needed to demonstrate its feasibility.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.