XGBoost-Cox model effectively predicts mortality risk in elderly male patients with acute kidney injury: a retrospective cohort study integrating nonlinear feature engineering and interpretable machine learning

Song Qiuyue; Ao Qiangguo; Wu Yazhou

doi:10.16016/j.2097-0927.202607040

您当前的位置：

首页 >

文章列表页 >

XGBoost-Cox model effectively predicts mortality risk in elderly male patients with acute kidney injury: a retrospective cohort study integrating nonlinear feature engineering and interpretable machine learning

Intelligent Medicine and Prediction Model | 更新时间：2026-10-30

- XGBoost-Cox model effectively predicts mortality risk in elderly male patients with acute kidney injury: a retrospective cohort study integrating nonlinear feature engineering and interpretable machine learning
- Journal of Army Medical University Vol. 48, Issue 20, Pages: 2933-2948(2026)
- 作者机构：
  
  1.陆军军医大学（第三军医大学）军事预防医学系军队卫生统计学教研室，重庆
  2.中国人民解放军总医院第二医学中心肾脏病科，北京
- 作者简介：
  
  Wu Yazhou，E-mail：asiawu@tmmu.edu.cn
- 基金信息：
  
  the General Program of National Natural Science Foundation of China(82574207);the General Project of Natural Science Foundation of Chongqing(CSTB2025NSCQ-GPX0648)
- DOI：10.16016/j.2097-0927.202607040
  CLC： R195.1;R339.34;R692.07
- Received：09 July 2026，
  
  Revised：2026-09-10，
  
  Accepted：12 September 2026，
  
  Published：30 October 2026
- 稿件说明：
移动端阅览
宋秋月，敖强国，伍亚舟.XGBoost-Cox模型可有效预测老年男性急性肾损伤患者死亡风险：一项融合非线性特征工程与可解释机器学习的回顾性队列研究［J］.陆军军医大学学报，2026，48（20）：2933-2948.DOI：10.16016/j.2097-0927.202607040.

Song Q Y，Ao Q G，Wu Y Z.XGBoost-Cox model effectively predicts mortality risk in elderly male patients with acute kidney injury： a retrospective cohort study integrating nonlinear feature engineering and interpretable machine learning［J］.J Army Med Univ，2026，48（20）：2933-2948.DOI：10.16016/j.2097-0927.202607040.
宋秋月，敖强国，伍亚舟.XGBoost-Cox模型可有效预测老年男性急性肾损伤患者死亡风险：一项融合非线性特征工程与可解释机器学习的回顾性队列研究［J］.陆军军医大学学报，2026，48（20）：2933-2948.DOI：10.16016/j.2097-0927.202607040. DOI：

Song Q Y，Ao Q G，Wu Y Z.XGBoost-Cox model effectively predicts mortality risk in elderly male patients with acute kidney injury： a retrospective cohort study integrating nonlinear feature engineering and interpretable machine learning［J］.J Army Med Univ，2026，48（20）：2933-2948.DOI：10.16016/j.2097-0927.202607040. DOI：

摘要

目的

老年急性肾损伤（acute kidney injury， AKI）患者死亡风险高，且随AKI后肾功能变化而改变，而现有预测模型多基于基线变量、可解释性有限。本研究基于老年男性急性肾损伤患者真实世界随访数据，构建融合非线性特征工程和可解释机器学习生存模型的死亡风险预测框架，并评价其区分度、校准度、临床净获益和可解释性。

方法

本研究为单中心回顾性队列预测模型构建研究，纳入中国人民解放军总医院2008年1月至2025年12月期间发生急性肾损伤7 d后仍存活的老年男性住院患者，以全因死亡为终点结局。2008年1月至2018年12月的患者为模型开发队列，并按7∶3比例采用分层随机抽样方法分为训练集和内部测试集；2019年1月至2025年12月的患者为时间验证队列。收集人口学特征指标，心血管及血流动力学状态指标，肾功能指标，AKI严重程度及肾功能动态变化指标，血液学指标，生化、营养及代谢指标，电解质及矿物质指标，既往病史、合并症及治疗相关因素等指标。采用限制性立方样条（restricted cubic spline， RCS）分析关键连续变量与死亡风险之间的非线性关系；随后结合Boruta算法、随机生存森林（random survival forest， RSF）及临床可解释性进行特征筛选，构建3个递进式预测变量集；分别采用6种生存模型：Cox比例风险模型、样条扩展Cox模型、Elastic-net Cox、RSF、XGBoost-Cox和DeepSurv进行死亡风险预测。采用C-index和时间依赖AUC评价模型区分度，并通过校准曲线和决策曲线分析评价最佳模型的校准度和临床实用性。进一步采用SHAP方法对最佳模型进行全局和个体化解释。

结果

最终纳入老年男性AKI患者2 335例，其中随访期间死亡1 912例，存活423例。训练集、内部测试集和时间验证集分别为775例、332例和1 228例。B型利钠肽、血尿素氮及AKI后短期估算肾小球滤过率动态变化指标与死亡风险存在非线性关系（

0.01）。年龄、营养状态、心肾功能负荷、血液学指标、合并症负担、机械通气及AKI后肾功能变化情况是老年AKI患者死亡风险的重要预测因素。纳入AKI后动态肾功能指标的变量集整体预测性能优于仅包含基线变量或AKI发生时变量的模型。测试集中XGBoost-Cox模型获得最高C-index和3年AUC，分别为0.749（95%

：0.722~0.774）、0.846（95%

：0.802~0.891）；RSF模型在30 d和90 d预测中表现最好，分别为0.840（95%

：0.785~0.888）、0.829（95%

：0.785~0.868）。时间验证结果中，XGBoost-Cox模型在30 d、90 d、1年和3年死亡风险预测中的AUC分别为0.797（95%

：0.763~0.830）、0

.781（95%

：0.755~0.809）、0.802（95%

：0.778~0.824）和0.822（95%

：0.788~0.853），提示模型在时间验证队列中具有较为稳定的泛化性能。XGBoost-Cox模型区分能力较好、3年预测校准总体一致且具有临床净获益，肾功能负荷为其主导预测因素。

结论

融合非线性特征工程与可解释机器学习生存模型的预测框架能够有效识别老年男性AKI患者死亡风险。其中，XGBoost-Cox模型在内部测试与时间验证中区分能力较好，3年死亡风险预测校准总体一致、临床净获益优于全治与不治策略，模型预测依据具有可解释性。

Abstract

Objective

Elderly patients with acute kidney injury (AKI) face a high mortality risk that varies with renal function trajectories after AKI

whereas existing prediction models rely mainly on baseline variables and offer limited interpretability. Based on real-world follow-up data from elderly male patients with AKI

this study aimed to develop a mortality risk prediction framework that integrates nonlinear feature engineering with interpretable machine-learning survival models

and to evaluate its discrimination

calibration

clinical net benefit

and interpretability.

Methods

A single-center

retrospective cohort study was adopted to construct a prediction model based on older male inpatients who survived at least 7 days after AKI onset at Chinese PLA General Hospital between January 2008 and December 2025

with all-cause mortality as the endpoint outcome. Patients admitted between January 2008 and December 2018 constituted the model development cohort

and were randomly divided into a training set and an internal test set at a 7∶3 ratio; patients admitted between January 2019 and December 2025 constituted the temporal validation cohort. Variables collected included demographic characteristics

cardiovascular and hemodynamic status

renal function indicators

AKI severity and dynamic changes in renal function

hematological indicators

biochemical

nutritional and metabolic indices

electrolytes and minerals

medical history

comorbidities

and treatment-related factors. Restricted cubic splines (RCS) analysis was used to examine the nonlinear relationships between key continuous variables and mortality risk. Feature selection was performed by combining the Boruta algorithm

random survival forest (RSF)

and clinical interpretability to construct 3 progressively expanded predictor sets. Six survival models were applied for mortality risk prediction

namely the Cox proportional hazards model

spline-extended Cox model

elastic-net Cox model

RSF

XGBoost-Cox

and DeepSurv. Discrimination was evaluated by the C-index and time-dependent AUC

and the calibration and clinical utility of the best-performing model were assessed by calibration curves and decision curve analysis. SHapley Additive exPlanations (SHAP) were further applied to interpret the best model globally and at the individual level.

Results

A total of 2 335 elderly male patients with AKI were finally included

of whom 1 912 died during follow-up and 423 survived. The training

internal test

and temporal validation sets comprised 775

332

and 1 228 patients

respectively. BNP

BUN

and short-term post-AKI dynamic eGFR indicators showed nonlinear relationships with mortality risk (

0.01). Age

nutritional status

cardiorenal functional load

hematological indicators

comorbidity burden

mechanical ventilation

and post-AKI changes in renal function were important predictors of mortality risk in elderly patients with AKI. The predictor set incorporating post-AKI dynamic renal function indicators revealed better overall predictive performance than models containing only baseline variables or AKI-onset variables. In the test set

the XGBoost-Cox model achieved the highest C-index and 3-year AUC

which were 0.749 (95%

: 0.722 to 0.774) and 0.846 (95%

: 0.802 to 0.891)

respectively; RSF performed best in 30-day and 90-day predictions

with AUC values of 0.840 (95%

: 0.785 to 0.888) and 0.829 (95%

: 0.785 to 0.868)

respectively. In temporal validation

the AUCs

of the XGBoost-Cox model for 30-day

90-day

1-year

and 3-year mortality risk prediction were 0.797 (95%

: 0.763 to 0.830)

0.781 (95%

: 0.755 to 0.809)

0.802 (95%

: 0.778 to 0.824)

and 0.822 (95%

: 0.788 to 0.853)

respectively

indicating stable generalizability in the temporal validation cohort. The XGBoost-Cox model showed good discrimination

calibration generally consistent with observed risk at 3 years

and clinical net benefit

with renal functional load as its dominant predictor.

Conclusion

The prediction framework integrating nonlinear feature engineering with interpretable machine-learning survival models can effectively identify mortality risk in elderly male patients with AKI. The XGBoost-Cox model showed good discrimination in both internal testing and temporal validation

with generally consistent calibration and net benefit superior to treat-all and treat-none strategies for 3-year mortality risk prediction; SHAP analysis identified renal functional load as its dominant predictor

indicating that the model's predictions are interpretable.

关键词

Keywords

references

Ostermann M ， Lumlertgul N ， Jeong R ， et al . Acute kidney injury ［J］. Lancet ， 2025 ， 405 （ 10474 ）： 241 - 256 . DOI： 10.1016/S0140-6736（24）02385-7 http://dx.doi.org/10.1016/S0140-6736（24）02385-7 .

Ronco C ， Bellomo R ， Kellum J A . Acute kidney injury ［J］. Lancet ， 2019 ， 394 （ 10212 ）： 1949 - 1964 . DOI： 10.1016/S0140-6736（19）32563-2 http://dx.doi.org/10.1016/S0140-6736（19）32563-2 .

Kidney Disease： Improving Global Outcomes （KDIGO） Acute Kidney Injury Work Group. KDIGO clinical practice guideline for acute kidney injury ［R］. 2012 ： 1 - 138 .

Grant C H ， Dahiya A ， Palechuk T ， et al . Health inequalities and outcomes following acute kidney injury： a systematic review & meta-analyses of observational studies ［J］. BMC Nephrol ， 2025 ， 26 （ 1 ）： 494 . DOI： 10.1186/s12882-025-04391-x http://dx.doi.org/10.1186/s12882-025-04391-x .

Jr Harrell F E . Regression modeling strategies： With applications to linear models， logistic and ordinal regression， and survival analysis ［M］. Cham ： Springer International Publishing ， 2015 .

Wynants L ， van Smeden M ， McLernon D J ， et al . Three myths about risk thresholds for prediction models ［J］. BMC Med ， 2019 ， 17 （ 1 ）： 192 . DOI： 10.1186/s12916-019-1425-3 http://dx.doi.org/10.1186/s12916-019-1425-3 .

Collins G S ， Moons K G M ， Dhiman P ， et al . TRIPOD+AI statement： updated guidance for reporting clinical prediction models that use regression or machine learning methods ［J］. BMJ ， 2024 ， 385 ： e078378 . DOI： 10.1136/bmj-2023-078378 http://dx.doi.org/10.1136/bmj-2023-078378 .

Kursa M B ， Rudnicki W R . Feature selection with the boruta package ［J］. J Stat Softw ， 2010 ， 036 （ i11 ）： 1 - 13 . DOI： 10.18637/jss.v036.i11 http://dx.doi.org/10.18637/jss.v036.i11 .

伍亚舟，陈锡程，易东 . 人工智能在临床领域的研究进展及前景展望［J］. 第三军医大学学报， 2022 ， 44 （ 1 ）： 89 - 102 . DOI： 10.16016/j.1000-5404.202112092 http://dx.doi.org/10.16016/j.1000-5404.202112092 .

Wu Y Z ， Chen X C ， Yi D . Advances and perspective of artificial intelligence in clinical area ［J］. J Third Mil Med Univ ， 2022 ， 44 （ 1 ）： 89 - 102 . DOI： 10.16016/j.1000-5404.202112092 http://dx.doi.org/10.16016/j.1000-5404.202112092 .

Haug C J ， Drazen J M . Artificial intelligence and machine learning in clinical medicine， 2023 ［J］. N Engl J Med ， 2023 ， 388 （ 13 ）： 1201 - 1208 . DOI： 10.1056/NEJMra2302038 http://dx.doi.org/10.1056/NEJMra2302038 .

Rajpurkar P ， Chen E ， Banerjee O ， et al . AI in health and medicine ［J］. Nat Med ， 2022 ， 28 （ 1 ）： 31 - 38 . DOI： 10.1038/s41591-021-01614-0 http://dx.doi.org/10.1038/s41591-021-01614-0 .

Ishwaran H ， Kogalur U B ， Blackstone E H ， et al . Random survival forests ［J］. Ann Appl Stat ， 2008 ， 2 （ 3 ）： 841 - 860 . DOI： 10.1214/08-aoas169 http://dx.doi.org/10.1214/08-aoas169 .

Chen T Q ， Guestrin C . XGBoost： a scalable tree boosting system ［C］// Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining ， 2016 ： 785 - 794 . DOI： 10.1145/2939672.2939785 http://dx.doi.org/10.1145/2939672.2939785 .

Katzman J L ， Shaham U ， Cloninger A ， et al . DeepSurv： personalized treatment recommender system using a Cox proportional hazards deep neural network ［J］. BMC Med Res Methodol ， 2018 ， 18 （ 1 ）： 24 . DOI： 10.1186/s12874-018-0482-1 http://dx.doi.org/10.1186/s12874-018-0482-1 .

Van Calster B ， McLernon D J ， van Smeden M ， et al . Calibration： the Achilles heel of predictive analytics ［J］. BMC Med ， 2019 ， 17 （ 1 ）： 230 . DOI： 10.1186/s12916-019-1466-7 http://dx.doi.org/10.1186/s12916-019-1466-7 .

Vickers A J ， Elkin E B . Decision curve analysis： a novel method for evaluating prediction models ［J］. Med Decis Making ， 2006 ， 26 （ 6 ）： 565 - 574 . DOI： 10.1177/0272989X06295361 http://dx.doi.org/10.1177/0272989X06295361 .

Heagerty P J ， Lumley T ， Pepe M S . Time-dependent ROC curves for censored survival data and a diagnostic marker ［J］. Biometrics ， 2000 ， 56 （ 2 ）： 337 - 344 . DOI： 10.1111/j.0006-341x.2000.00337.x http://dx.doi.org/10.1111/j.0006-341x.2000.00337.x .

Jiao Y R ， Wu Z ， Zhang Y B ， et al . Explainable machine learning-based 28-day mortality prediction model for elderly patients with acute kidney injury ［J］. BMC Nephrol ， 2026 ， 27 （ 1 ）： 157 . DOI： 10.1186/s12882-026-04792-6 http://dx.doi.org/10.1186/s12882-026-04792-6 .

梁晶晶，肖回宇，屈壕，等 . LightGBM预测脓毒症相关急性肾损伤患者28天死亡风险性能最优：基于MIMIC-Ⅳ与eICU数据库构建可解释性机器学习模型及验证对比［J］. 陆军军医大学学报， 2026 ， 48 （ 15 ）： 2129 - 2138 . DOI： 10.16016/j.2097-0927.202605041 http://dx.doi.org/10.16016/j.2097-0927.202605041 .

Liang J J ， Xiao H Y ， Qu H ， et al . LightGBM demonstrates optimal performance in predicting 28-day mortality risk in patients with sepsis-associated acute kidney injury： development and validation of an interpretable machine learning model based on MIMIC-Ⅳ and eICU databases ［J］. J Army Med Univ ， 2026 ， 48 （ 15 ）： 2129 - 2138 . DOI： 10.16016/j.2097-0927.202605041 http://dx.doi.org/10.16016/j.2097-0927.202605041 .

Liu M Q ， Fan Z P ， Gao Y ， et al . A two-tier feature selection method for predicting mortality risk in ICU patients with acute kidney injury ［J］. Sci Rep ， 2024 ， 14 （ 1 ）： 16794 . DOI： 10.1038/s41598-024-63793-3 http://dx.doi.org/10.1038/s41598-024-63793-3 .

Lundberg S M ， Erion G ， Chen H ， et al . From local explanations to global understanding with explainable AI for trees ［J］. Nat Mach Intell ， 2020 ， 2 （ 1 ）： 56 - 67 . DOI： 10.1038/s42256-019-0138-9 http://dx.doi.org/10.1038/s42256-019-0138-9 .

Zarbock A ， Nadim M K ， Pickkers P ， et al . Sepsis-associated acute kidney injury： consensus report of the 28th Acute Disease Quality Initiative workgroup ［J］. Nat Rev Nephrol ， 2023 ， 19 （ 6 ）： 401 - 417 . DOI： 10.1038/s41581-023-00683-3 http://dx.doi.org/10.1038/s41581-023-00683-3 .

Chawla L S ， Bellomo R ， Bihorac A ， et al . Acute kidney disease and renal recovery： consensus report of the Acute Disease Quality Initiative （ADQI） 16 Workgroup ［J］. Nat Rev Nephrol ， 2017 ， 13 （ 4 ）： 241 - 257 . DOI： 10.1038/nrneph.2017.2 http://dx.doi.org/10.1038/nrneph.2017.2 .

Steyerberg E W ， Vergouwe Y . Towards better clinical prediction models： seven steps for development and an ABCD for validation ［J］. Eur Heart J ， 2014 ， 35 （ 29 ）： 1925 - 1931 . DOI： 10.1093/eurheartj/ehu207 http://dx.doi.org/10.1093/eurheartj/ehu207 .

Views

下载量

CSCD

Alert me when the article has been cited

提交

Tools

Publicity Resources

LightGBM demonstrates optimal performance in predicting 28-day mortality risk in patients with sepsis-associated acute kidney injury: development and validation of an interpretable machine learning model based on MIMIC-Ⅳ and eICU databases

Prediction of risk for acute kidney injury and its progression to mortality in obese patients admitted to ICU postoperatively

Prediction of postoperative sepsis mortality risk based on machine learning in patients undergoing abdominal surgery

Predicting grade≥2 chemotherapy-induced myelosuppression in older patients with colorectal cancer receiving first-line chemotherapy: a multi-algorithm consensus approach using 9 routine parameters to develop and validate an explainable boosting machine

Bioinformatics-based identification of ferroptosis biomarkers and a diagnostic model with in vivo validation in chronic obstructive pulmonary disease

Related Author

Liang Jingjing

Xiao Huiyu

Qu Hao

Chen Xin

Wu Yazhou

Song Qiuyue

LI Qiang

MU Guo

Related Institution

Department of Anesthesiology， Affiliated Hospital of Southwest Medical University

Department of Anesthesiology， Zigong Fourth People’s Hospital

Department of Anesthesiology， West China Hospital of Sichuan University

Department of Anesthesiology, First Affiliated Hospital, Army Medical University (Third Military Medical University)

Regiment Five, Basical Medicine College, Army Medical University (Third Military Medical University)

⁰