基于机器学习与SHAP可解释性住院老年2型糖尿病患者衰弱风险预测模型的构建及验证

吴佳琪; 方素真; 黎楷; 黄敏; 张艳群; 罗祥蓉; 贾彬; 李冰蕾; 谢翠华

doi:10.16460/j.issn2097-6569.2025.11.007

PDF(1138 KB)

护理学报 ›› 2025, Vol. 32 ›› Issue (11) : 7-12. DOI: 10.16460/j.issn2097-6569.2025.11.007

研究生园地

基于机器学习与SHAP可解释性住院老年2型糖尿病患者衰弱风险预测模型的构建及验证

吴佳琪^1a,2, 方素真^1b, 黎楷^1a,2, 黄敏^1a,2, 张艳群^1c,2, 罗祥蓉^1b, 贾彬^1a,2, 李冰蕾^1a,2, 谢翠华^1a

作者信息 +

Construction and validation of frailty risk prediction model for hospitalized elderly patients with type 2 diabetes based on machine learning and SHAP Explainabilility

WU Jiaqi^1a,2, FANG Suzhen^1b, LI Kai^1a,2, HUANG Min^1a,2, ZHANG Yanqun^1c,2, LUO Xiangrong^1b, JIA Bin^1a,2, LI Binglei^1a,2, XIE Cuihua^1a

Author information +

文章历史 +

摘要

目的构建住院老年2型糖尿病患者衰弱风险预测模型,并对模型进行解释。方法采用便利抽样法,选取2024年3-8月广州市2所三级甲等医院的509例住院老年2型糖尿病患者为研究对象。采用Python 3.8分析,按照7∶3的比例将数据集随机分为训练集和验证集。采用5种机器学习算法开发模型并对模型性能进行比较,确定最优预测模型并解释。结果 CatBoost在5个机器学习模型中表现最好,受试者工作特征曲线（Receiver Operating Characteristic, ROC）下面积（AUC）为0.755,精确率为0.660,召回率为0.433,特异度为0.897,准确率为0.752,F1分数为0.522。SHAP摘要图显示了前7个最重要的特征,依次为抑郁、年龄、老年营养风险指数、血红蛋白、糖尿病自我管理水平、睡眠、糖化血红蛋白。结论本研究开发评估并且解释了住院老年2型糖尿病患者衰弱的风险预测模型,为早期识别衰弱高风险的人群并精准制定干预措施提供参考。

Abstract

Objective To construct a frailty risk prediction model for elderly inpatients with type 2 diabetes mellitus (T2DM) and interpret the model. Methods Convenience sampling was used to select 509 elderly inpatients with T2DM from two tertiary grade-A hospitals in Guangzhou from March to August 2024. The data were analyzed using Python 3.8 statistical software and randomly divided into a training set and a validation set in a 7:3 ratio. Five machine learning algorithms were employed to develop the models, and their performance was compared to determine the optimal prediction model, which was then interpreted. Results Among the five machine learning models, the one developed by Catboost performed the best, with an area under the receiver operating characteristic (ROC) curve (AUC) of 0.755, precision of 0.660, recall of 0.433, specificity of 0.897, accuracy of 0.752, and F1 score of 0.522. The SHAP summary plot identified the top seven most important features influencing model prediction as follows: depression, age, Geriatric Nutritional Risk Index (GNRI), hemoglobin diabetes self-management level, sleep and HbA1c. Conclusion This study develops, evaluates, and explains the frailty risk prediction model for elderly inpatients with T2DM, providing a reference for early identification of high-risk individuals and the development of targeted interventions.

导出引用

吴佳琪, 方素真, 黎楷, 黄敏, 张艳群, 罗祥蓉, 贾彬, 李冰蕾, 谢翠华. 基于机器学习与SHAP可解释性住院老年2型糖尿病患者衰弱风险预测模型的构建及验证[J]. 护理学报. 2025, 32(11): 7-12 https://doi.org/10.16460/j.issn2097-6569.2025.11.007

WU Jiaqi, FANG Suzhen, LI Kai, HUANG Min, ZHANG Yanqun, LUO Xiangrong, JIA Bin, LI Binglei, XIE Cuihua. Construction and validation of frailty risk prediction model for hospitalized elderly patients with type 2 diabetes based on machine learning and SHAP Explainabilility[J]. Journal of Nursing. 2025, 32(11): 7-12 https://doi.org/10.16460/j.issn2097-6569.2025.11.007

中图分类号： R473.5

参考文献

[1] Sun H, Saeedi P, Karuranga S, et al.IDF Diabetes atlas: global, regional and country-level diabetes prevalence estimates for 2021 and projections for 2045[J]. Diabetes Res Clin Pract, 2022, 183:109119. DOI:10.1016/j.diabres.2021.109119.
[2] Frailty in older people[J].Frailty in older people[J]. Lancet, 2013, 381(9868): 752-762. DOI:10.1016/S0140-6736(12)62167-9.
[3] Li G, Prior J C, Leslie W D, et al.Frailty and risk of fractures in patients with type 2 diabetes[J]. Diabetes Care, 2019, 42(4): 507-513. DOI:10.2337/dc18-1965.
[4] Zhao Y, Li X, Li S, et al.Using machine learning techniques to develop risk prediction models for the risk of incident diabetic retinopathy among patients with type 2 diabetes mellitus: a cohort study[J]. Front Endocrinol,2022, 13: 876559. DOI:10.339/fendo.2022.876559.
[5] Zhang W, Liu X, Dong Z, et al.New diagnostic model for the differentiation of diabetic nephropathy from non-diabetic nephropathy in chinese patients[J]. Front Endocrinol, 2022, 13: 913021. DOI:10.3389/fendo.2022.913021.
[6] 王冰雪, 林婷, 吴静, 等. 糖尿病足溃疡患者住院时间延长风险预测模型构建及验证[J]. 四川大学学报(医学版), 2024, 55(4): 972-979.
[7] The Lancet Respiratory Medicine null.Opening the black box of machine learning[J]. Lancet Respir Med, 2018, 6(11): 801. DOI:10.1016/S2213-2600(18)30425-9.
[8] Alberti KG, Zimmet PZ.Definition, diagnosis and classification of diabetes mellitus and its complications. Part 1: diagnosis and classification of diabetes mellitus provisional report of a WHO consultation[J]. Diabet Med, 1998, 15(7): 539-553.DOI:10.1002/(SICI)1096-9136(199807).
[9] Ogundimu EO, Altman DG, Collins GS.Adequate sample size for developing prediction models is not simply related to events per variable[J]. J Clin Epidemiol, 2016, 76: 175-182. DOI:10.1016/j.jclinepi.2016.02.031.
[10] 李凤, 王建华, 宋开兰, 等. 老年糖尿病患者衰弱Meta分析[J]. 中国老年学杂志, 2021, 41(20): 4377-4382.
[11] Fried L P, Tangen C M, Walston J, et al.Frailty in older adults: evidence for a phenotype[J]. J Gerontol A Biol Sci Med Sci, 2001, 56(3): M146-156. DOI:10.1093/gerona/56.3.m146.
[12] 王燕.协同护理模式对老年THR术后衰弱的干预研究[D]. 延边:延边大学, 2021.[2024-11-28]. DOI:10.27439/d.cnki.gybdu.2020.000537.
[13] Buysse DJ, Reynolds CF, Monk TH, et al.The Pittsburgh Sleep Quality Index: a new instrument for psychiatric practice and research[J]. Psychiatry Res, 1989, 28(2): 193-213. DOI:10.1016/0165-1781(89)90047-4.
[14] 刘贤臣, 唐茂芹, 胡蕾, 等. 大学生睡眠质量与心理健康状况的相关性研究[J]. 山东精神医学, 1994(4): 4-9.
[15] Toobert D J, Hampson S E, Glasgow R E.The summary of diabetes self-care activities measure: results from 7 studies and a revised scale[J]. Diabetes Care, 2000, 23(7): 943-950. DOI:10.2337/diacare.23.7.943.
[16] 孙胜男, 赵维纲, 董颖越, 等. 糖尿病患者自我管理现状及影响因素分析[J]. 中华护理杂志, 2011, 46(3): 229-233.
[17] Charlson ME, Pompei P, Ales K L, et al.A new method of classifying prognostic comorbidity in longitudinal studies: development and validation[J]. J Chronic Diseases, 1987, 40(5): 373-383. DOI:10.1016/0021-9681(87)90171-8.
[18] Hoyl MT, Alessi CA, Harker JO, et al.Development and testing of a five-item version of the Geriatric Depression Scale[J]. J Am Geriatr Soc, 1999, 47(7): 873-878.DOI:10.1111/j.1532-5415.1999.tb03848.x.
[19] Friedman B, Heisel MJ, Delavan RL.Psychometric properties of the 15-item geriatric depression scale in functionally impaired, cognitively intact, community-dwelling elderly primary care patients[J]. J Am Geriatr Soc, 2005, 53(9): 1570-1576. DOI:10.1111/j.1532-5415.2005.53461.x.
[20] Bouillanne O, Morineau G, Dupont C, et al.Geriatric nutritional risk index: a new index for evaluating at-risk elderly medical patients2[J]. Am J Clin Nutr, 2005, 82(4): 777-783. DOI:10.1093/ajcn/82.4.777.
[21] 李小娟, 练玉颖, 杨宇. 营养风险筛查2002、微型营养评定简化版和老年营养风险指数在老年食管癌病人营养风险筛查中的应用[J]. 安徽医药, 2024, 28(12): 2443-2447.
[22] Wang L, Zhang D, Xu J.Association between the Geriatric Nutritional Risk Index, bone mineral density and osteoporosis in type 2 diabetes patients[J]. J Diabetes Investig, 2020, 11(4): 956-963. DOI:10.1111/jdi.13196.
[23] Hong J, Huang QQ, Liu WY, et al.Three nutritional indices are effective predictors of mortality in patients with type 2 diabetes and foot ulcers[J]. Frontiers in Nutrition, 2022, 9: 851274. DOI:10.3389/fnut.2022.851274.
[24] Cho Aj, Hong YS, Park HC, et al.Geriatric nutritional risk index is associated with retinopathy in patients with type 2 diabetes[J]. Sci Rep, 2022, 12(1): 11746.DOI:10.1038/s 41598-022-15463-5.
[25] Chung PC, Chan TC.Physical frailty and oral health as risk factors for low bone density in community-dwelling residents[J]. Sci Rep, 2024,14(1):18131.DOI:10.1038/s41598-024-68958-8.