Abstract: Background Lung cancer remains the foremost cause of cancer-related morbidity and mortality worldwide. Short-term unplanned readmissions（UPR） place a significant burden on healthcare systems，compromising care quality，nursing efficiency，and patient outcomes. The systemic immune-inflammation index（SII） serves as a composite marker of systemic inflammation and immune status and has demonstrated prognostic potential in lung cancer progression and therapeutic response. Elucidating its association with UPR in patients treated with immune checkpoint inhibitors（ICIs） may assist in refining clinical risk stratification strategies. Objective This study aimed to evaluate whether SII levels are associated with the likelihood of UPR within 31 days among patients with advanced-stage lung cancer receiving ICIs therapy. Methods This retrospective study included 247 patients with stage Ⅲ - Ⅳ lung cancer who received immune checkpoint inhibitors（ICIs） between January 2023 and May 2024. Patients were categorized into the unplanned readmission（UPR） group（n=164） and the non-UPR group （n=83）. Demographic data，clinical treatment information，routine biochemical parameters，and tumor-related markers were collected. Independent risk factors for UPR were identified using multivariate logistic regression and subgroup analyses. Receiver operating characteristic（ROC） curves were generated with the pROC package to evaluate the predictive accuracy，sensitivity，and specificity of both the regression model and SII. Restricted cubic spline（RCS） regression combined with Shapley additive explanations（SHAP） was employed to examine the dose-response relationship between S II and UPR risk，to determine the SII threshold，and to perform subgroup analyses based on this threshold. Results The 31-day incidence of UPR was 33.6% （83/247）. Multivariate logistic regression identified length of hospital stay（OR=1.073，95%CI=1.015-1.134，P=0.013），NRS2002 score ≥ 3（OR=4.457，95%CI=1.774-11.198，P<0.001），CA50 >25 U/mL（OR=2.667，95%CI=1.044-6.816，P=0.040），AFP>7 μg/L（OR=5.355，95%CI=1.845-15.539，P<0.001），and SII（OR=3.204，95%CI=1.079-9.512， P=0.036） as independent predictors of UPR. After adjusting for hospital stay，NRS 2002，CA50，and AFP，patients in the highest SIIquartile（Q3：SII ≥ 1,018.26） had a significantly increased risk of UPR compared with those in the lowest quartile（Q1：SII ≤ 406.42）（OR=2.262，95%CI=1.026-4.987，P=0.042）. ROC curve analysis showed that the regression model had strong discriminatory ability，with an AUC of 0.826（95%CI=0.769-0.882），accuracy of 0.802，sensitivity of 0.700，and specificity of 0.854. By contrast，SII alone yielded an AUC of 0.660（95%CI=0.585-0.735），accuracy of 0.668，sensitivity of 0.590，and specificity of 0.707. RCS analysis revealed a nonlinear positive association between SII and UPR risk，with an inflection point at 1,363.78. When SII was ≥ 1,363.78，the risk of UPR increased 6.37-fold（95%CI=1.93-14.90）. Subgroup analyses demonstrated that among patients with NRS2002<3，AFP<7 μg/L，and CA50<25 U/mL，those with SII ≥ 1,363.78 had 2.55，3.23，and 3.67-fold higher risks of UPR，respectively，compared with patients with lower SII values. Stratified analyses by hospital stay showed no statistically significant differences（P>0.05）. Conclusion An elevated SII（ ≥ 1 363.78） independently predicts 31 d UPR in patients with advanced lung cancer undergoing ICIs treatment. Given the robust discriminative capacity of the predictive model，incorporation of regular SII monitoring into clinical practice is recommended for early identification of high-risk individuals. Preventive interventions should be considered when SII sUPRasses the identified threshold to reduce the likelihood of readmission.

Key words: Lung cancer, Unplanned readmission, Immune checkpoint inhibitors, Systemic immuno-inflammatory

摘要： 背景 肺癌是全球发病率和死亡率最高的恶性肿瘤，其短期非计划再入院（UPR）情况对医疗质量、护理效率及患者预后具有影响。系统免疫炎症指数（SII）可反映机体炎症 - 免疫状态，在肺癌进展与治疗中具有预测价值。探讨 SII 与免疫检查点抑制剂（ICIs）治疗肺癌患者短期 UPR 的相关性，可为临床风险分层提供参考依据。目的 探讨 SII 水平与接受 ICIs 治疗的中晚期肺癌患者短期内 UPR 风险的相关性。方法 回顾性纳入 2023 年 1 月—2024 年 5 月期间接受 ICIs 治疗的 247 名Ⅲ ~ Ⅳ期肺癌患者为研究对象。其中，UPR 组 164 例，非 UPR 组 83 例，收集其人口学特征及临床治疗指标、常规生化指标、肿瘤相关指标。采用多因素 Logistic 回归及亚组分析筛选肺癌ICIs 治疗患者 UPR 的独立危险因素。使用 pROC 包绘制受试者工作特征（ROC）曲线，评估回归模型及 SII 单独预测肺癌患者非计划再入院的准确度、灵敏度及特异度。采用限制性立方样条回归（RCS）联合沙普利可解释性分析（SHAP）探索 SII 与患者非计划再入院风险之间的剂量效应关系，确定 SII 阈值，并依据 SII 阈值对病例进行亚组分析。结果 纳入患者 31 d 内 UPR 发生率为 33.60%（83/247），多因素 Logistic 回归分析显示住院天数（OR=1.073，95%CI=1.015~1.134，P=0.013）、NRS2002 ≥ 3 分（OR=4.457，95%CI=1.774~11.198，P<0.001）、CA50>25 U/mL（OR=2.667，95%CI=1.044~6.816，P=0.040）、AFP >7 μg/L（OR=5.355，95%CI=1.845~15.539，P<0.001）及 SII（OR=3.204，95%CI=1.079~9.512，P=0.036）是肺癌患者发生 UPR 的独立危险因素。在控制住院天数、NRS2002、CA50、AFP 等因素后，进一步多因素 Logistic 回归分析，与 Q1 组（SII ≤ 406.42）相比，Q3 组（SII ≥ 1 018.26）发生 UPR 的风险增加（OR=2.262，95%CI=1.026~4.987，P=0.042）。ROC 曲线分析结果显示：回归模型预测肺癌 UPR 的 AUC 为 0.826（95%CI=0.769~0.882），准确度为 0.802，灵敏度为 0.700，特异度为 0.854，SII 单独预测肺癌 UPR 的 AUC为 0.660（95%CI=0.585~0.735），准确度为 0.668，灵敏度为 0.590，特异度为 0.707。限制性立方样条分析结果显示：SII 与患者非计划再入院之间存在非线性正相关，曲线拐点为 1 363.78；且当 SII ≥ 1 363.78 时，患者 UPR 风险增加 6.37倍（95%CI=1.93~14.90）。进一步亚组分析结果显示：当 NRS 2002 评分 <3 分、AFP<7 μg/L、CA50<25 U/mL 时，SII ≥ 1 363.78 亚组患者发生 UPR 风险分别是 SII<1 363.78 亚组的 2.55、3.23、3.67 倍。住院天数分层分析中，两组比较无统计学意义（P>0.05）。结论 研究证实 SII ≥ 1363.78 是肺癌 ICIs 治疗患者短期内非计划再入院的独立预测因子，基于模型良好的判别效能，建议临床将 SII 动态监测纳入该类患者再入院风险管理，当 SII 超过临界值时，可考虑实施预防性干预措施，以降低再入院风险。

关键词: 肺癌, 非计划再入院, 免疫检查点抑制剂, 系统免疫炎症指数