Children possess immature immune systems that render them vulnerable to infectious illnesses (1, 2). In low- and middle-income nations, the mortality rate associated with Pediatric Sepsis can reach 25%, posing a significant hazard to children's well-being and survival (3).

Sepsis is characterized as a life-threatening dysfunction of organs resulting from an imbalanced host response triggered by an infection (4). The pathogenesis of sepsis is complex, and until the year 2020, clear diagnostic criteria for early pediatric sepsis were lacking. The publication of the International Guidelines of the Campaign to Save Sepsis in 2020 provided a framework for diagnosing sepsis in children (5), stipulating that a Pediatric Sequential Organ Failure Assessment (PSOFA) score of ≥2 in patients with infections or suspected infections indicates sepsis. However, this guideline is subject to limitations, as the PSOFA score is derived from the Sequential Organ Failure Assessment (SOFA) score utilized in adults. Children cannot be considered as miniature versions of adults, hence a straightforward application of adult criteria to assess organ function in children may not be precise (2). The onset and progression of sepsis entail specific temporal dynamics and alterations. Children with sepsis may not exhibit typical septic presentations initially (6), thereby complicating early diagnosis.

Inadequate understanding of sepsis among some pediatric practitioners and the prevalent issue of delayed diagnoses underscore the ongoing challenge of enhancing early sepsis detection through routine laboratory assessments. There is a pressing need for a predictive model capable of accurately discerning individuals at risk of sepsis progression, enabling the prompt administration of standardized treatments to mitigate mortality rates. In this study, we conducted a retrospective analysis of laboratory findings from children with and without sepsis, and developed and validated a composite model utilizing Least Absolute Shrinkage and Selection Operator (LASSO) regression and multifactorial Logistic regression, which achieved favorable results.

In this study, we developed and validated a new, simple and effective predictive model for the risk of sepsis onset up to 48 h in advance, updated continuously as new data is received. Utilizing fundamental vital signs and key laboratory parameters including respiratory rate, body temperature, percentage of naïve granulocytes, platelet count, procalcitonin, fibrinogen and lactate, column-line diagrams were constructed. The predictive model underwent internal validation, demonstrating strong predictive capabilities, and can serve as a valuable tool for pediatricians in evaluating children at risk of sepsis development.

Given the diverse clinical symptomatology in the early phases of sepsis, the limited specificity of laboratory markers, and the absence of definitive biomarkers, numerous researchers have advocated for combining multiple biomarkers to enhance early sepsis diagnosis through scoring systems. Sakyi (10) et al. conducted a case-control study involving 60 children aged 0–6 years with sepsis, and 30 non-septic children as controls. They identified CRP, PCT, and preprotease (sCD14-ST) as markers with substantial predictive value and as independent predictors of sepsis in children. While the predictive capacity of PCT aligns with our study's findings, we did not identify CRP as an independent predictor of sepsis, potentially attributed to the limited case numbers in their investigation. Several studies have indicated an association between preprotease (sCD14-ST) and sepsis development (11–13). However, the limited availability of presepsin testing, particularly in primary care settings, significantly hampers the practical application of the model. Our study's strengths lie in the ample sample size, well-defined variables, and the accessibility of their corresponding data. For instance, platelet count and immature granulocyte percentage can be obtained from CBC analyses, while lactate levels can be assessed through arterial blood gas (ABG) analyses. These routine laboratory tests are widely accessible in healthcare facilities across various settings. Florian's (14) team also developed a pediatric sepsis prediction model incorporating laboratory measures of IL-6, platelet count, PCT, CRP, and four clinical parameters including PICU length of stay, presence of central venous catheter, body temperature, and cumulative number of sepsis and SIRS episodes prior to diagnosis. The complexity of the variables in this model may hinder early diagnosis, for instance, PICU length of stay could impede sepsis identification in non-ICU settings. Moreover, the cumulative number of sepsis and SIRS episodes may not be easily interpreted by individuals with low literacy levels. In terms of early and accurate diagnosis, our predictive model surpasses the former. Other researchers have utilized sepsis-related genes to design diagnostic prediction models (15–18), but these models are primarily applicable in large, well-equipped hospitals. In less-developed regions and primary care settings, delayed sepsis diagnosis and elevated mortality rates are more prevalent.

Three of the seven variables screened in this study, including calcitoninogen, platelet count and lactate, are widely acknowledged for their strong association with sepsis diagnosis and severity, as supported by existing research. Platelets are crucial in diagnostic criteria such as the SOFA score and the newly proposed Phoenix score (9). Lactate holds significance in the Phoenix score. The lactate categorization in this study adhered to the latest Phoenix score criteria to ensure broad recognition. PCT exhibits rapid responsiveness post-bacterial infection, typically rising 2–4 h after sepsis onset (19). With its high sensitivity and specificity, PCT serves as a valuable biomarker for sepsis diagnosis and prognostication of disease severity (20). In the non-infected state of the organism, PCT concentrations <0.05 ng/ml are observed, while concentrations exceeding 10 ng/ml suggest severe sepsis or septic shock (8, 21, 22). In a randomized controlled trial by Ali et al. (23), PCT was found to exhibit higher accuracy, specificity, and sensitivity compared to CRP in sepsis diagnosis (80.79% vs. 69.45%, 36% vs. 28.7%, 87.6% vs. 72.4%). In a study by Doerflinger et al. (24) focusing on an early prediction model for pediatric sepsis diagnosis using procalcitonin and interleukin-10, the combination of PCT (≥0.425 ng/ml) and IL-10 (≥4.37 pg/ml) demonstrated a sensitivity of 100% (95% CI: 68.8%–100%) and a specificity of 89% (95% CI 80.0–95.0%). This study reported an 11-fold increase in PCT levels in the sepsis group compared to the non-sepsis group, aligning with the findings of the current investigation.

The present study also identified respiratory rate and body temperature as predictors of sepsis in children. The concept of SIRS was introduced as early as 1991 in sepsis 1.0 (25), indicating that a clinical response to non-specific injury (temperature > 38°C or <36°C; heart rate > 90 bpm; respiratory rate > 20 bpm; total white blood cell count > 12*10⁻⁹/L or <4*10⁻⁹/L, or proportion of immature (rod-shaped nuclei) neutrophils > 10%) meeting two criteria plus a suspected infection is adequate for sepsis diagnosis. Similarly, the diagnostic criteria of sepsis 2.0 in 2001 included assessment of respiratory rate and temperature (26), although these criteria were later deemed to lack specificity and are no longer in use. Nevertheless, they can still reflect the body's inflammatory response status and criticality level to some extent. Temperature changes can be influenced by various factors, including feeding, environmental temperature, and the level of immune stimulation, as demonstrated in animal studies (27, 28). Severe infections tend to induce more significant and varied temperature fluctuations compared to other causes. A large prospective study showed (29) a bimodal distribution of body temperatures in septic patients, with the development of sepsis linked to hyperthermia and increased mortality associated with hypothermia. This implies that hyperthermia and hypothermia may indicate a more severe and an excessive immune response. Consistent with the findings of this study, a body temperature <36°C or >38.5°C also resulted in a single high score of 20.9. Respiratory dysfunction can cause hypoxemia and dyspnea, with respiratory rate serving as an indicator of respiratory distress (30). Following the onset of sepsis, the respiratory system is commonly affected, and studies have indicated a significant association between abnormal respiratory rate and sepsis diagnosis (31).

Monitoring both temperature and respiratory rate is routine and easily accessible. In this study, alongside assessing sepsis likelihood using respiratory rate and body temperature status, we also combined PCT, PLT, and lactate with other laboratory indicators that are closely related to the development of sepsis to make a comprehensive determination, which is a highly effective guide for the identification of sepsis in children.

Sepsis and septic shock can lead to sepsis-induced coagulation dysfunction by activating inflammatory mediators and causing damage to endothelial cells (32). These fibrinolytic changes induced may lead to microthrombosis and microcirculatory disturbances, ultimately leading to organ failure and increased mortality (33). Fibrinogen plays a key role in haemostasis and antimicrobial host defence (34). In patients with sepsis, fibrinogen levels are often decreased and are associated with a poor prognosis (32). Signoff et al. (35) observed that children with sepsis combined with hypofibrinogenemia may have a more complicated disease course (73.7% vs. 29.2%; P < 0.001) and a higher 28-day mortality rate (26.3% vs. 7. 1%, P = 0.002). A retrospective study demonstrated a significant association between fibrinogen levels below 2 g/L upon PICU admission and an elevated risk of mortality in pediatric sepsis patients (36). The immature granulocyte percentage, often underappreciated in clinical settings, may rise in infected individuals, particularly in the presence of bacterial infections, as immature naive granulocytes are released prematurely into the peripheral blood. In a study on naive cells as the earliest biomarkers of sepsis, it was mentioned that absolute and immature granulocyte percentage serve as the earliest discriminators of sepsis (37). These markers exhibit an area under the curve of 0.81 and 0.82, respectively, and can be elevated up to 24 h prior to sepsis diagnosis. Therefore, absolute and immature granulocyte percentage represent readily accessible and cost-effective markers with high specificity and sensitivity for distinguishing septic from non-septic patients. A prospective study by Nierhaus et al. (38) similarly found that IG% could differentiate between infected and non-infected patients (P < 0.0001) with a sensitivity of 89.2% and a specificity of 76.4%, especially within 48 h following sepsis onset.

In this study, a set of seven variables closely associated with sepsis development and readily accessible were chosen to construct a column chart and an online calculator. To guarantee clinical reliability and practical applicability, the web-based interactive calculator is currently undergoing final stages of optimization and validation testing. Due to the complexity of consciousness impairment in children with severe sepsis, assessment cannot rely solely on the Glasgow Coma Scale (GCS) and delirium screening tools (39). Moreover, clinical evaluation of consciousness status demonstrates high variability and subjectivity, particularly considering age-related differences (40), Therefore, consciousness assessment was excluded from these variables. The column chart created in this study, based on seven quantifiable indicators, enhances clinical applicability and holds significant potential for reducing child mortality and disability, particularly in resource-limited settings. It should be emphasized that this model is primarily designed for sepsis screening in children upon hospital admission or during clinical deterioration. This approach may have limited clinical utility in children definitively diagnosed with sepsis and concomitant multiple organ dysfunction syndrome (MODS) in the PICU. While the model demonstrates strengths in operational utility and broad applicability, it is essential to acknowledge its limitations, for instance, the single-center retrospective design and lack of external validation may introduce bias in the study. Future endeavors should include external validation or larger prospective studies to enhance the accuracy and generalizability of the prediction model.