1(Department of Biomedical Engineering South China University of Technology,Guangzhou 510006, China) 2(Division of Nephrology Guangdong Provincial People′s Hospital, Guangdong Academy of Medical Sciences, Guangzhou 510080, China)
Abstract：Acute kidney injury (AKI) is one of the most common critical diseases in inpatients, which has high morbidity and mortality. This study aimed to build model to predict AKI and analyze the risk factors affecting AKI occurrence, leading to timely intervene for patients with potential risk of AKI. A total of 90780 subjects (7983AKI and 82797 non-AKI, dividing training sets and independent test sets by 4:1) who stayed from 2 to 14 days were selected, and their biochemical indicators, drug use, demographic information and other clinical information were collected. Three models including logical regression, random forest and LightGBM were adopted to fulfill AKI prediction, and the performance of different models was evaluated and compared using sensitivity, specificity, the area under the characteristic working curve (AUC) and F1 value. LighGBM obtained the best performance to predict occurrence of AKI in 24 hours, the optimal F1, AUC, sensitivity, specificity was 0.800, 0.871, 0.755 and 0.987 respectively. The important factors affecting the occurrence of AKI include: Clinical features: age, the length of stay, admission department; Laboratory test: Creatinine value in the first examination, levels of sodium, potassium, chlorine, uric acid and glycosylated hemoglobin in blood; Medicine: Anti-infective drugs, non-steroidal anti-inflammatory drugs, diuretics or dehydrants, adrenergic receptor agonists, contrast agents, ACEI/ARB antihypertensive drugs as well as the types of drug used, days of drug treatment; Comorbidities: Moderate and severe nephropathy. Using the clinical information of inpatients, machine learning model can effectively predict the risk of AKI in 24 hours, reveal the important factors affecting the occurrence of AKI, and provide important basis for rational and effective treatment of inpatients and timely intervention for patients with AKI risk.
[1] Mizuno T, Sato W, Ishikawa K, et al. KDIGO (Kidney Disease: Improving Global Outcomes) criteria could be a useful outcome predictor of cisplatin-induced acute kidney injury [J]. Oncology, 2012, 82(6): 354-359. [2] Fortescue EB, Bates DW, Chertow GM. Predicting acute renal failure after coronary bypass surgery: Cross-validation of two risk-stratification algorithms [J]. Kidney International, 2000, 57(6): 2594-2602. [3] Wijeysundera DN, Karkouti K, Dupuis JY, et al. Derivation and validation of a simplified predictive index for renal replacement therapy after cardiac surgery [J]. Jama-Journal of the American Medical Association, 2007, 297(16): 1801-1809. [4] Devarajan P, Mishra J, Supavekin S, et al. Gene expression in early ischemic renal injury: clues towards pathogenesis, biomarker discovery, and novel therapeutics [J]. Molecular Genetics and Metabolism, 2003, 80(4): 365-376. [5] Han WK, Bailly V, Abichandani R, et al. Kidney injury molecule-1 (KIM-1): A novel biomarker for human renal proximal tubule injury [J]. Kidney International, 2002, 62(1): 237-244. [6] Cheruvanky A, Zhou H, Pisitkun T, et al. Rapid isolation of urinary exosomal biomarkers using a nanomembrane ultrafiltration concentrator [J]. American Journal of Physiology-Renal Physiology, 2007, 292(5): F1657-F1661. [7] Laszczyska O, Severo M, Azevedo A. Electronic medical record-based predictive model for acute kidney injury in an acute care hospital [J]. Exploring Complexity in Health: An Interdisciplinary Systems Approach, 2016, 228: 810-812. [8] Matheny ME, Miller RA, Ikizler TA, et al. Development of inpatient risk stratification models of acute kidney injury for use in electronic health records [J]. Medical Decision Making, 2010, 30(6): 639-650. [9] Cheng P, Waitman LR, Hu Y, et al. Predicting Inpatient Acute Kidney Injury over Different Time Horizons: How Early and Accurate? [J]. AMIA Annu Symp Proc, 2017, 2017: 565-574. [10] 宋林, 杨大利, 崔丽艳. 急性肾损伤早期预测方法 [J]. 北京信息科技大学学报(自然科学版), 2015, 30(5): 67-71. [11] Koyner JL, Carey KA, Edelson DP, et al. The development of a machine learning inpatient acute kidney injury prediction model [J]. Critical Care Medicine, 2018, 46(7): 1070-1077. [12] Mohamadlou H, Lynn-Palevsky A, Barton C, et al. Prediction of acute kidney injury with a machine learning algorithm using electronic health record data [J]. Canadian Journal of Kidney Health and Disease, 2018, 5: 2054358118776326. [13] Kate RJ, Perez RM, Mazumdar D, et al. Prediction and detection models for acute kidney injury in hospitalized older adults [J]. Bmc Medical Informatics & Decision Making, 2016, 16(1): 1-11. [14] Chen T, Guestrin C. XGBoost: A scalable tree boosting system [C]//Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. San Francisco: 2016: 785-794. [15] Wu YH, Chen YH, Li SW, et al. Value of electronic alerts for acute kidney injury in high-risk wards: A pilot randomized controlled trial [J]. International Urology and Nephrology, 2018, 50(8): 1483-1488. [16] 王彦娥, 丁伶清, 宋洪涛,等. 非ICU患者肾毒性药物与急性肾损伤关系的病例对照研究 [J]. 中国医院药学杂志, 2018, 38(8): 869-873. [17] Cessie SL, Houwelingen JCV. Ridge estimators in logistic regression [J]. Journal of the Royal Statistical Society, 1992, 41(1): 191-201. [18] Breiman L. Random forests [J]. Machine Learning, 2001, 45(1): 5-32. [19] Ke GL, Meng Q, Finley T, et al. LightGBM: A highly efficient gradient boosting decision tree [C] //Proceedings of the 31st International Conference on Neural Information Processing Systems. Long Beach; Curran Associates Inc. 2017: 3149-3157. [20] 马晓君, 沙靖岚, 牛雪琪. 基于LightGBM算法的P2P项目信用评级模型的设计及应用 [J]. 数量经济技术经济研究, 2018, 35(5): 144-160. [21] 王妍, 鄂静, 李博,等. 高尿酸血症与急性肾损伤的相关性Meta分析 [J]. 宁夏医学杂志, 2016, 38(3): 215-217. [22] Dhodi D, Bhagat S, Pathak D, et al. Drug-induced nephrotoxicity [J]. International Journal of Basic & Clinical Pharmacology, 2014, 3(4): 681-682. [23] 黄英姿, 杨毅. 重症患者急性肾损伤诊治的热点与思考 [J]. 中国医师进修杂志, 2017, 40(1): 9-12. [24] Wolpert DH, Macready WG. An efficient method to estimate Bagging′s generalization error [J]. Machine Learning, 1999, 35(1): 41-55.
