Article Text

Original research
Predictive value of laboratory indicators for in-hospital death in children with community-onset sepsis: a prospective observational study of 266 patients
  1. jing Liu1,2,
  2. Qin Hui1,
  3. Xiuxiu Lu3,
  4. Wei Li3,
  5. Ning Li3,
  6. Yuanmei Chen1,
  7. Qi Zhang1,2
  1. 1 Department of Pediatrics, China-Japan Friendship Hospital, Beijing, China
  2. 2 Peking Union Medical College & Chinese Academy of Medical Sciences, Beijing, China
  3. 3 Department of Intensive Care Unit, Capital Institute of Pediatrics, Beijing, China
  1. Correspondence to Dr Qi Zhang; zhangqikeyan{at}163.com

Abstract

Background and objectives This study aimed to identify predictors of sepsis-associated in-hospital mortality from readily available laboratory biomarkers at onset of illness that include haematological, coagulation, liver and kidney function, blood lipid, cardiac enzymes and arterial blood gas.

Methods Children with sepsis were enrolled consecutively in a prospective observational study involving paediatric intensive care units (PICUs) of two hospitals in Beijing, between November 2016 and January 2020. The data on demographics, laboratory examinations during the first 24 hours after PICU admission, complications and outcomes were collected. We screened baseline laboratory indicators using the Least Absolute Shrinkage and Selection Operator (LASSO) analysis, then we constructed a mortality risk model using Cox proportional hazards regression analysis. The ability of risk factors to predict in-hospital mortality was evaluated by receiver operating characteristic (ROC) curves.

Results A total of 266 subjects were enrolled including 44 (16.5%) deaths and 222 (83.5%) survivors. Those who died showed a shorter length of hospitalisation, and a higher proportion of mechanical ventilation, complications and organ failure (p<0.05). LASSO analysis identified 13 clinical parameters related to prognosis, which were included in the final Cox model. An elevated triglyceride (TG) remained the most significant risk factor of death (HR=1.469, 95% CI: 1.010 to 2.136, p=0.044), followed by base excess (BE) (HR=1.131, 95% CI: 1.046 to 1.223, p=0.002) and pH (HR=0.95, 95% CI: 0.93 to 0.97, p<0.001). The results of the ROC curve showed that combined diagnosis of the three indicators—TG+BE+pH—has the best area under the curve (AUC) (AUC=0.77, 95% CI: 0.69 to 0.85, p<0.001), with a 68% sensitivity and 80% specificity.

Conclusion Laboratory factors of TG, BE and pH during the first 24 hours after intensive care unit admission are associated with in-hospital mortality in PICU patients with sepsis. The combination of the three indices has high diagnostic value.

  • Mortality
  • Statistics
  • Data Collection

Data availability statement

Data are available upon reasonable request. The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

http://creativecommons.org/licenses/by-nc/4.0/

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.

Statistics from Altmetric.com

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

WHAT IS ALREADY KNOWN ON THIS TOPIC

  • Prior studies have laid a foundation for understanding paediatric sepsis but lacked clarity on prognostic laboratory indicators in previously healthy children with community-onset sepsis.

WHAT THIS STUDY ADDS

  • Our study identifies triglycerides, base excess and pH as robust predictors of in-hospital mortality in paediatric sepsis, offering novel insights into risk stratification and early intervention strategies.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

  • These findings advocate for the integration of early laboratory assessment, including triglycerides, base excess and pH, into clinical protocols, potentially shaping future research and guiding clinical management to improve outcomes in paediatric sepsis.

Introduction

Sepsis, a syndrome of physiological, pathological and biochemical abnormalities induced by infection, is a major public health concern.1 Sepsis incidence and the percentage of all global deaths (from any cause) which were related to sepsis peaked in early childhood.2–4 The most recent estimates of the global burden of sepsis indicate that there were 20.3 million incident sepsis cases resulting in more than 2.9 million deaths among children younger than 5 years, and 4.9 million sepsis cases caused 45 400 deaths among children and adolescents aged 5–19 years globally in 2017.3 Sepsis remains a leading cause of death in the intensive care unit (ICU) of hospitals, with the case fatality rate (CFR) from paediatric sepsis of 25%.5–7 However, the prospective differentiation of patients with sepsis at low risk of death remains a major clinical challenge for improving overall survival. Thus, predictors to identify patients who are at higher risk of sepsis-specific mortality are of great clinical relevance.

A traditional sepsis outcome prediction approach, the Sepsis-related Organ Failure Assessment (SOFA), has a good correlation with mortality.8 Furthermore, there is a body of literature exploring the application of SOFA to critically ill paediatric patients in both the paediatric ICU (PICU) and emergency department.9 10 Other scoring systems for assessing prognosis in paediatric patients with clinically critical condition, such as Paediatric Risk of Mortality III score,11 Paediatric Index of Mortality 312 and the Paediatric Early Warning Score are recognised for their ability to generate accurate probabilistic survival estimates.13 However, it is essential to acknowledge that these scoring systems, while providing valuable prognostic insights, may exhibit a trade-off between sensitivity and specificity. In particular, there may be instances where optimising sensitivity takes precedence over specificity, considering the critical nature of predicting adverse outcomes in paediatric patients. More recently, more attention has been directed towards identification of metabolite biomarkers such as metabolites or proteins.14 15 Novel biomarkers such as interleukin 8 and the neutrophil CD64 index have been reported to be stratification tools for diagnosing sepsis and predicting outcomes,16 17 but they are not yet readily available at the bedside. Search for biomarkers that could provide a reliable and early estimate of the risk stratification and mortality risk estimate still continues, and currently, there are few useful predictive markers in clinical practice to identify which patients are at the highest risk of death.

The aim of this paper is to assess the association between laboratory biomarkers readily available for clinical use and the in-hospital death of patients with sepsis.

Methods

Study design and setting

This is a prospective observational study implemented on participants admitted to the PICUs of two tertiary care hospitals between November 2016 and January 2020 in Beijing.

Study population

Patients were included if they met the following criteria: (1) community-acquired sepsis; (2) age <18 years; (3) admission to the PICU within 24 hours after sepsis diagnosis; (4) laboratory results were obtained within 24 hours of PICU arrival. Exclusion criteria were set as the following: (1) other chronic diseases: such as cancer; cardiovascular disease; genetic disorder; endocrine, nutritional and metabolic diseases; respiratory diseases; neurological diseases; autoimmune diseases or immunodeficiency; (2) severe congenital malformations and trauma; (3) congenital disease; (4) long-term medication history.

Data collection and laboratory assays

We collected data based on three aspects: (1) demographic characteristics: age and sex; (2) laboratory test results including: source of infection, haematological, infection markers, coagulation, liver and kidney function, blood lipid, cardiac enzymes, lymphocyte subsets, humoral immunity and arterial blood gas—these tests were performed in the clinical laboratory departments of the two hospitals according to standard protocols; (3) the primary study outcome was the in-hospital mortality (death during hospitalisation and follow-up ceased at discharge), and the secondary outcomes were the length of PICU stay, complications and organ failure. For those who had more than one result for a given laboratory test within 24 hours of PICU arrival, only the first value was included.

Definitions

In our study, the diagnosis of sepsis was based on the Pediatric Sepsis Consensus in 2015 and Surviving Sepsis Campaign International Guidelines in 2012: (1) temperature >38.5°C or <35°C; (2) tachycardia (a heart rate exceeding 2 SDs above normal age-related values, which may be absent in the hypothermic patient); (3) at least one of the following indications of altered organ function: altered mental status, hypoxaemia/(arterial oxygen pressure/fractional inspired oxygen <300 mm Hg), increased serum lactate level (>1 mmol/L) or bounding pulses.18 19 Other definitions of complications, such as acute respiratory distress syndrome (ARDS), multiple organ dysfunction syndrome (MODS), disseminated intravascular coagulation (DIC), septic shock and organ failure, are included in online supplemental table 1.

Supplemental material

Statistical analyses

Data were first subjected to a normality test using the Shapiro-Wilk test. Most continuous variables had a non-normal distribution and were expressed as median (Q1, Q3). Continuous variables with a normal distribution were expressed as means±SD; categorical variables as frequencies and percentages.

To compare the differences in characteristics between sepsis non-survivors and sepsis survivors, non-parametric test (Mann-Whitney U test), parametric t-test (Student’s t-test) or Χ2 test was used. A total of 39 laboratory indicators collected within 24 hours after admission were subjected to Least Absolute Shrinkage and Selection Operator (LASSO) analysis to identify potential predictors. Subsequently, the selected factors were further analysed using multifactor Cox regression to determine independent risk factors. Finally, we created a receiver operating characteristic (ROC) curve for variables in the final multivariate Cox model to evaluate the diagnostic value of these biomarkers. For each ROC curve, we calculated the area under the curve (AUC) with 95% CIs, Youden’s index and cut-off values.

A p value of <0.05 (two sided) was considered statistically significant. Data management and statistical analyses were performed using STATA SE V.17.0 (StataCorp, College Station, Texas, USA), and the ROC curves were performed by MedCalc software (V.20.022).

Results

Clinical characteristics of patients with sepsis

There were a total of 569 children with sepsis admitted to the PICU from November 2016 and January 2020. 290 children were excluded based on our predetermined inclusion criteria. Of the 279 children initially enrolled, 266 (95%) were included in this study. The flow chart of the study is shown in figure 1. The median age of the patients was 13.5 (4.0, 39.0) months, and 181 (68.5%) were male.

Figure 1

The flow chart of study screening selection process. PICU, paediatric intensive care unit.

The patients were divided into two groups: sepsis non-survivors (n=44, 16.5%) and sepsis survivors (n=222, 83.5%). There was no statistical difference in age and gender between groups (table 1). Length of PICU stay was significantly lower in the sepsis non-survivors group compared with the sepsis survivors group (median values, 11 and 16 days, respectively; p=0.012). Compared with the survivors group, the non-survivors group exhibited higher rates of mechanical ventilation (93.18% vs 54.50%, p<0.001), bloodstream infections (68.18% vs 40.54%, p<0.001) and gastrointestinal infection (43.18% vs 16.67%, p<0.001).

Table 1

Clinical characteristics of patients in the sepsis group who were survivors and non-survivors

There were significant differences between groups in the frequency of individual complications, including ARDS, MODS, DIC and septic shock, and in the organ failure aspect, including respiratory, heart, nervous system, liver and renal failure (all p<0.001).

Comparison of baseline characteristics of the sepsis survivors and non-survivors

More details of laboratory findings can be seen in online supplemental table 2. Platelets were significantly lower in non-survivors than in survivors (139.5 vs 315, p<0.001), without any change in white cell count (17.48 vs 15.36, p=0.146), red cell count (3.69 vs 4.00, p=0.185), haemoglobin (98.5 vs 110, p=0.389), neutrophil count (11.79 vs 9.34, p=0.13) and lymphocyte count (3.4 vs 4.0, p=0.05). C reactive protein (CRP) (79 vs 47, p=0.18) did not show a significant increasing trend with more severe condition of the disease, but the opposite in procalcitonin. At baseline, the survivors had a significantly higher prothrombin time (PT), activated partial thromboplastin time (APTT), international normalised ratio (INR), alanine aminotransferase (ALT), aspartate transaminase (AST), lactate dehydrogenase (LDH), hydroxybutyrate (HBDB), creatine kinase (CK), CK-MB, triglyceride (TG), creatinine (Cr), blood urea nitrogen (BUN), base excess (BE) and lactic acid; and lower albumin (ALB), CD4/CD8 ratio, IgM, high-density lipoprotein (HDL), low-density lipoprotein (LDL) cholesterol, pH, oxygen saturation and Ca2+ (all p<0.05).

Risk factors of in-hospital death in patients with sepsis

LASSO regression was employed to identify characteristic variables associated with in-hospital death from sepsis. Figure 2A illustrates the distribution of LASSO coefficients for clinical features among 39 clinical parameters, while figure 2B displays the tuning parameter (λ) selection in the LASSO model using 10-fold cross-validation based on the minimum criterion. Ultimately, 13 parameters were found to be related to prognosis: PT, APTT, AST, platelet, CRP, BE, Ca2+, HBDB, TG, HDL, BUN, fibrinogen degradation products(FDP) and pH.

Figure 2

Feature selection using the Least Absolute Shrinkage and Selection Operator regression model. (A) the distribution of LASSO coefficients. (B) LASSO model with the parameter (λ) adjusted by 10-fold cross-validation.

In the final multivariate Cox regression model (figure 3), we found that elevated TG remained the most significant risk factor of death (HR=1.469, 95% CI 1.010 to 2.136, p=0.044), followed by BE (HR=1.131, 95% CI 1.046 to 1.223, p=0.002). For every 0.01 increase in pH, we observed a decrease in the risk of death by about 5% (HR=0.95, 95% CI 0.93 to 0.97, p<0.001) in the final model.

Figure 3

Cox regression hazard analysis of laboratory indicators screened by Least Absolute Shrinkage and Selection Operator analysis. aPer 0.1 mmol/L increase; bPer 0.1 mmol/L increase; cPer 0.01 increase. APTT, activated partial thromboplastin time; AST, aspartate transaminase; BE, base excess; BUN, blood urea nitrogen; CRP, C reactive protein; FDP, fibrinogen degradation products;HBDB, hydroxybutyrate; HDL, high-density lipoprotein; PLT, platelet; PT, prothrombin time; TG, triglyceride. Statistically significant variables have been bolded.

The predictive value of factors for differentiating non-survivors and survivors in patients with sepsis

In order to estimate the relative accuracy and generalisation properties of the multivariate Cox model, a ROC curve analysis was performed (figure 4), obtaining an AUC of pH (0.71, 0.62 to 0.81, p<0.001), TG (0.61, 0.51 to 0.70, p=0.013) and BE (0.64, 0.54 to 0.74, p=0.002). The combined diagnosis of the three indicators of TG+BE+pH has the best AUC (AUC=0.77, 95% CI 0.69 to 0.85, p<0.001), with a 68% sensitivity and 80% specificity (table 2).

Figure 4

Receiver operating characteristic curve of factors in predicting in-hospital mortality of children with sepsis. BE, base excess; TG, triglyceride.

Table 2

Accuracy of multivariate Cox model in differentiating non-survivors and survivors among patients with sepsis

Discussion

Despite all efforts to improve sepsis mortality, determining reliable mortality risk is still challenging and a very difficult process early in the clinical setting. To the best of our knowledge, this is one of the few prospective multicentre studies to address this clinical dilemma, describing clinical characteristics associated with 266 patients with sepsis.

Our data suggest that the CFR from paediatric sepsis was 16.5% (44 of 266). This finding is consistent with a meta-analysis conducted among 94 studies and 7561 children that showed that pooled CFRs were 31.7% and 19.3% in developing and developed countries, respectively.20 In addition, a study revealed that the mortality rate of paediatric patients with severe sepsis managed in Japanese PICUs was 19%.21

Our results showed that most of the sepsis cases were children younger than 1.5 years, with a higher proportion of males. About one-sixth of our patients with sepsis died in hospital and half developed severe complications such as MODS, ARDS, DIC and shock. On the other hand, the most common complications were MODS (90.9%) and shock (86.4%) in the non-survivors group. Septic shock, the most severe form of sepsis, is commonly associated with MODS.22 Respiratory failure (86.4%) was the most common organ failure whether in the total population or in the non-survivors group. Risks of respiratory infections (72.8%) and mechanical ventilation (93.2%) were highest for the non-survivors group. Unlike in adult studies, only 40% of patients with severe sepsis received mechanical ventilation, and the association between case volume and mortality was more modest among severe sepsis cases requiring mechanical ventilation.23

In this study, we show that platelets were markedly reduced in the non-survivors group. Although not significant in the multivariate analysis, low-level platelets were a significant risk factor in the univariate model. The thrombocytopenia that occurs in sepsis is believed to reflect accelerated platelet consumption due to increased blood coagulation.24 Alterations in platelet number and function are common in both clinical and experimental sepsis. Previous research has implicated that both thrombocytopenia and platelet hyper-reactivity have been associated with increased sepsis-related morbidity and mortality.25 Also, a recent study has revealed sepsis alters the transcriptional and translational landscape of human and murine platelets, and sepsis-induced upregulation of ITGA2B predicts mortality.26

The mechanism of poor outcomes in patients with sepsis may be multiple and cannot be ascertained by our data. An excessive inflammatory activation response by pathogens, massive activation of coagulation and various types of cell death seem to be the main mechanisms of the high mortality from sepsis.27 This is consistent with the increase in inflammatory markers, such as CRP, procalcitonin and coagulation indices including PT, APTT and INR in our study. Prolonged APTT and PT indicate a consumption or inhibition of clotting factors. Coagulation activation may have a prominent role in the patients with sepsis, and this may explain their poorer clinical course. Although our studies have observed a possible relationship between sepsis mortality and coagulation, the mechanism is unclear. It is possible that blood clotting is the major cause of host death following inflammasome activation,28 and inflammasomes, such as NLRP3, NALP1 and caspase-11, bridge inflammation with thrombosis.29 30

Organ dysfunction occurs frequently in and represents a significant comorbidity of sepsis, but the mechanistic basis for sepsis-induced organ dysfunction is controversial. Our data suggest that sepsis altered biochemical markers, (eg, LDH, HBDB, CK, CK-MB, ALT, AST, ALB, Cr, BUN) in the heart, liver and kidney. Sepsis causes profound myocardial depression. There is also evidence that bedside echocardiography frequently reveals severe biventricular dysfunction.31 We did not find that CK-MB was significantly associated with sepsis mortality in multivariate model. Other studies have shown that no significant association was found between left ventricular ejection fraction and mortality.32 More research is needed to explore the other possible predictions driving the relationship between sepsis-related cardiac function and death. Sepsis-associated acute kidney injury is also a common, life-threatening complication, increasing in-hospital mortality sixfold to eightfold, up to a quarter of whom will require renal replacement therapy.33 In our study, we observed that half of patients had renal failure, and BUN and Cr were significantly increased in the non-survivor group indicating kidney dysfunction as the probable cause of death. BUN was a risk factor of in-hospital death in univariate analysis. Other studies have shown similar findings. A higher BUN level is independently linked with the presence and severity of neonatal sepsis.34 A retrospective cohort study presented that BUN with 41.1 mg/dL as a cut-off level was associated with 28-day mortality in adult patients with sepsis.35

We found that ALT and AST were significantly elevated, and ALB decreased slightly in the non-survivors group. Basic research shows liver dysfunction is an early and commonplace event in the rat model of sepsis and detected changes significantly precede conventional markers and are reflected by early alterations in plasma bile acids.27 A deeper understanding of sepsis-induced changes at the organ level through multiomics analysis of several tissues (including liver, heart, kidney, etc) would provide an essential guide for clinical practice.

Alterations in lipid metabolism and the activation of lipid signalling pathways have important physiological and pathological meanings in the progression of sepsis. In the non-survivors group, we demonstrate elevated plasma levels of TG and decreased levels of HDL and LDL. We found that elevated TGs were the most significant risk factor of sepsis in-hospital death. In addition, similar adult results have been reported that baseline lipid values, particularly TG concentrations, were also associated with hospital mortality in patients with sepsis.36 Barber et al also found that low HDL is associated with poor outcome in sepsis.37 We recommend the utility of plasma TG as a prognostic marker. Lipid mediators play an important role in the proinflammatory and counter-regulatory anti-inflammatory changes in the microvasculature in sepsis. Future research focusing on production and metabolism of various lipid species in sepsis has the potential to identify novel biomarkers and therapeutic targets.

Our results revealed that pH and BE were important variables for in-hospital death in patients with sepsis. pH was different and was maintained within the normal range between non-survivors and survivors. It seems that tissue hypoperfusion and metabolic disturbances may not be evident in the early stages. However, very small pH and BE changes were very important in predicting sepsis outcomes. Notably, meta-analyses provide pooled estimates of pH (7.21 vs 7.31) and base deficit (4.6 vs 2.7) between non-survivors and survivors.38 Although lactate for sepsis diagnosis has been well identified, it was not found to be a determinant of sepsis mortality in our multivariate model. But some investigations pointed out that lactate normalisation (<2 mmol/L) within 4 hours was associated with decreased persistent organ dysfunction.39 Calcium levels were significantly decreased, while potassium, sodium and chloride ions exhibited no difference in non-survivors on admission. Previous research has shown that movement of calcium and phosphorus is part of the process in which vitamin D, parathyroid hormone, fibroblast growth factor and klotho interact with sepsis defence mechanisms.40

Despite our promising findings, our study has limitations. First, data on patients’ sepsis status serially during hospitalisation were not available for patients included in the present study. Second, we included only patients presenting with community-acquired sepsis; those with complex underlying medical conditions were excluded. Thus, patients included in our study may be slightly healthier than the general population. The limited cohort size may impact the external validity of our findings, particularly in terms of generalisability to broader sepsis populations. Lastly, our study only explored objective laboratory tests as predictors of sepsis mortality. Future studies may benefit from incorporating scoring systems or other clinical parameters to improve the identification of death from sepsis.

Conclusion

In conclusion, we have shown that the mortality rate of community-acquired sepsis in Chinese PICUs was 16.5%. Laboratory factors of TG, BE and pH during the first 24 hours after ICU admission were associated with in-hospital mortality in PICU patients with sepsis. The combination of the three indices had high diagnostic value. These results indicate that these readily available laboratory biomarkers have the potential to effectively identify previously well paediatric patients with community-acquired sepsis at high risk of in-hospital mortality. Potential benefits from timely risk stratification could allow for closer monitoring and more timely interventions.

Data availability statement

Data are available upon reasonable request. The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Ethics statements

Patient consent for publication

Ethics approval

The study was conducted in accordance with the Declaration of Helsinki (as revised in 2013). This study was approved by the Ethics Committee of the Capital Institute of Pediatrics (approval number: SHERLL 2013075), and written informed consent was obtained from legal guardians.

Acknowledgments

We thank all the children and their families for participating in this study.

References

Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

Footnotes

  • jL and QH contributed equally.

  • Contributors Conception and design—JL and QZ. Administrative support—QZ. Provision of study materials or patients—XL, WL, YC and QZ. Collection and assembly of data—XL, WL, NL and YC. Data analysis and interpretation—JL and QH. Manuscript writing—JL and QZ. QZ is responsible for the overall content as guarantor. Final approval of the manuscript—all authors. The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

  • Funding This work was financially supported by the Scientific Research Foundation of China-Japan Friendship Hospital (2019-RC-3).

  • Competing interests None declared.

  • Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.