目的 基于感染、凝血、组织灌注及宿主储备等多维度的常规实验室指标，构建儿童脓毒症联合预测模型，并据此开发简便易行的早期预警评分量表（SEWS）。 方法 回顾性选取2021年1月至2023年12月深圳市儿童医院收治的193例感染患儿，根据病情转归分为脓毒症组（n = 92 ）和普通感染组（n = 101）。比较两组实验室指标差异，通过受试者工作特征（ROC）曲线进行初步筛选，采用多因素logistic回归分析确定独立危险因素，构建联合预测模型。基于回归系数制定评分量表，并收集 2024年1—4月 的 40 例住院感染患儿进行外部验证。 结果 多因素 logistic 回归分析显示，PCT、INR、肌红蛋白、白蛋白、Hb 及 MPV 是儿童脓毒症的独立危险因素（P < 0.05）。联合预测模型的ROC曲线下面积（AUC）高达 0.983（95%CI： 0.969 ~ 0.997），显著优于单项指标。进一步构建的简易评分量表在截断值为 6 分时，诊断脓毒症的灵敏度为 94.6%，特异度为 92.1%，AUC高达 0.981（95%CI： 0.967 ~ 0.995）。外部验证结果显示，该量表的灵敏度为 100%，特异度为 84.0%。 结论 基于多维度常规实验室指标构建的SEWS量表具有显著的诊断价值。当评分≥ 6分时，提示脓毒症发生风险极高，该量表可作为临床医生早期识别脓毒症并实施预警干预的可靠工具。

Objective To develop and externally validate a multidimensional Early Warning Scoring Scale （SEWS） for pediatric sepsis， grounded exclusively in routinely available laboratory parameters reflecting four pathophysiological domains： infection， coagulation dysfunction， impaired tissue perfusion， and depletion of host physiological reserves. Methods We conducted a retrospective cohort study involving 193 children with confirmed infection admitted to Shenzhen Children′s Hospital between January 2021 and December 2023. Participants were stratified into two clinically defined groups： the sepsis group （n = 92） and the non-sepsis infection group （n = 101）， based on consensus diagnostic criteria （e.g.， International Pediatric Sepsis Consensus Conference definitions）. Comparative analyses of baseline laboratory parameters were performed between groups. Candidate variables were initially selected using univariate analysis and area under the receiver operating characteristic curve （AUC-ROC） evaluation； subsequently， independent predictors of pediatric sepsis were identified via multivariable logistic regression. A weighted scoring scale—the SEWS—was derived directly from regression coefficients. Internal model performance was rigorously assessed， and external validation was carried out using an independent cohort of 40 consecutive cases collected prospectively from January to April 2024. Results Multivariable logistic regression identified six independent predictors of pediatric sepsis： Procalcitonin （PCT）， international normalized ratio （INR）， myoglobin， albumin， hemoglobin （Hb）， and mean platelet volume （MPV） （all P < 0.05）. The composite prediction model achieved an AUC of 0.983 （95%CI： 0.969 - 0.997）， significantly outperforming any single biomarker. The SEWS demonstrated excellent discriminative ability in the derivation cohort： sensitivity = 94.6%， specificity = 92.1%， and AUC = 0.981 （95%CI： 0.967 - 0.995） at an optimal cutoff score of ≥ 6 points. In the external validation cohort （n = 40）， SEWS maintained high sensitivity （100%） and strong specificity （84.0%）. Conclusions The SEWS is a pragmatic， multidimensional clinical tool built solely on routinely measured laboratory parameters. It exhibits robust diagnostic accuracy for early identification of pediatric sepsis， with a score of ≥ 6 conferring high positive predictive value. This evidence supports its integration into clinical workflows to facilitate timely risk stratification and prompt therapeutic intervention.