慢性阻塞性肺疾病急性加重住院患者出院状态预测研究

李少凡; 李莉芳; 何航帜; 张垚烨; 原一玮; 赵卉; 张岩波

doi:10.16462/j.cnki.zhjbkz.2024.06.011

慢性阻塞性肺疾病急性加重住院患者出院状态预测研究

doi: 10.16462/j.cnki.zhjbkz.2024.06.011

李少凡¹,
李莉芳²,
何航帜¹,
张垚烨¹,
原一玮¹,
赵卉^2, ,,
张岩波^{1, 3, ,}

1.
山西医科大学公共卫生学院卫生统计学教研室，重大疾病风险评估山西省重点实验室，太原 030001
2.
山西医科大学第二医院呼吸与危重症医学科，太原 030001
3.
山西中医药大学，晋中 030619

基金项目:

国家自然科学基金 82173631

山西省科技合作交流专项 202204041101031

详细信息

通讯作者:
张岩波，E-mail: sxmuzyb@126.com

赵卉，E-mail: hui_zhao@sxmu.edu.cn

中图分类号: R195.1;R563
计量
- 文章访问数: 83
- HTML全文浏览量: 32
- PDF下载量: 9
- 被引次数: 0
出版历程
- 收稿日期: 2024-02-28
- 修回日期: 2024-04-02
- 网络出版日期: 2024-07-13
- 刊出日期: 2024-06-10

Predictive study on discharge status of hospitalized patients with acute exacerbation of chronic obstructive pulmonary disease

LI Shaofan¹,
LI Lifang²,
HE Hangzhi¹,
ZHANG Yaoye¹,
YUAN Yiwei¹,
ZHAO Hui^{2
, ,},
ZHANG Yanbo^{1, 3
, ,}

1.
Department of Biostatistics, School of Public Health, Shanxi Medical University, Shanxi Provincial Key Laboratory of Major Diseases Risk Assessment, Taiyuan 030001, China
2.
Department of Cardiology, Pulmonary and Critical Care Medicine, Second Hospital, Shanxi Medical University, Taiyuan 030001, China
3.
Shanxi University of Chinese Medicine, Jinzhong 030619, China

Funds:

National Natural Science Foundation of China 82173631

Special of Science and Technology Cooperation and Exchange of Shanxi Province 202204041101031

More Information

Corresponding author: ZHAO Hui, E-mail: hui_zhao@sxmu.edu.cn; ZHANG Yanbo, E-mail: sxmuzyb@126.com

摘要

摘要: 目的为解决肺功能检测不易获取、测量误差大等问题，结合出院状态和住院的时间，构建机器学习预后预测模型，实现慢性阻塞性肺疾病急性加重期(acute exacerbation of chronic obstructive pulmonary disease, AECOPD)患者预后的精准预测。方法选择2011年10月―2020年5月因AECOPD于山西医科大学第二医院呼吸科住院的患者3 035例。结局变量为中位住院时长内是否好转出院。通过构建5种机器学习模型[逻辑回归(logistic regression, LR)、支持向量机(support vector machine, SVM)、随机森林(random forest, RF)、Catboost(categorical boosting)、多层感知机(multilayer perceptron, MLP)]建立预测模型，比较受试者工作特征(receiver operating characteristic, ROC)的曲线下面积(area under curve, AUC)等评价指标，选出最优模型。最后使用最优模型进行决策曲线分析，验证其临床实用性。结果 RF相较于其他机器学习模型的综合预测性能最佳，AUC为0.780、准确率为69.69%、精确率为64.50%、召回率为75.18%、F1分数为69.44%、布里尔分数为18.77%，校准曲线基本与对角线一致，决策曲线分析有较好的临床收益。结论基于RF的预测模型可以在无法获得肺功能检测相关指标的情况下实现对AECOPD患者预后的准确预测，为临床医生在评估与治疗决策中提供一定的支持。
- 慢性阻塞性肺疾病 /
- 急性加重期 /
- 机器学习 /
- 预测模型
Abstract: Objective A machine learning prognosis prediction model was created by innovatively combining discharge status and length of hospital stay, to accurately predict the prognosis of patients with acute exacerbation of chronic obstructive pulmonary disease (AECOPD). This was done in order to address the issues of difficult to obtain pulmonary function tests and large measurement error. Methods A total of 3 035 inpatients with AECOPD were recruited from the second hospital of Shanxi Medical University between October 2011 and May 2020. The outcome variable is whether or not the patient recovered and was discharged within the median length of hospitalization.The prediction model is created using five distinct machine learning models: logistic regression, support vector machine, random forest, Catboost, and multi-layer perceptron. By contrasting evaluation metrics like area under the working characteristic curve (AUROC), the optimal model is determined. In order to verify the decision curve's clinical applicability, the best model was used to assess it. Results In comparison to other machine learning models, random forest has the greatest overall prediction performance, with AUC of 0.780, accuracy of 69.69%, precision of 64.50%, recall of 75.18%, F1 score of 69.44%, and Brier score of 18.77%. The decision curve analysis has high clinical value, and the calibration curve is largely compatible with the diagonal. Conclusions The prediction model based on random forest may reliably forecast the prognosis of patients with AECOPD and provide some aid to physicians in evaluation and treatment decision-making when the important indices of the lung function test cannot be acquired.
- Chronic obstructive pulmonary disease /
- Acute exacerbation period /
- Machine learning /
- Prediction model

HTML全文

图 1 各模型的预测效果与最佳模型的决策曲线

LR: 逻辑回归; SVM: 支持向量机; RF: 随机森林; MLP: 多层感知机; AUC: 曲线下面积; A: 各模型的评价指标; B: 各模型的受试者工作特征曲线; C: 各模型的校准曲线; D: RF的决策曲线。

Figure 1. The prediction effect of each model and the decision curve of the best model

LR: logistic regression; SVM: support vector machine; RF: random forest; MLP: multilayer perceptron; AUC: area under curve; A: evaluation metrics of each model; B: receiver operating characteristic curve of each model; C: calibration curve of each model; D: decision curve of RF.

下载: 全尺寸图片幻灯片

表 1 3 035例慢性阻塞性肺疾病急性加重期患者的一般情况

Table 1. General situation of 3 035 patients with acute exacerbation of chronic obstructive pulmonary disease

变量 Variable	分类 Class	合计 ^① Total ^① (n=3 035)	中位住院时长内好转出院 ^① Recovered and discharged within the median length of hospitalization ^① (n=1 388)	中位住院时长内未好转出院 ^① Not recovered and discharged within the median length of hospitalization ^① (n=1 647)	t/z/$\chi^2$ 值 value	P值 value
年龄/岁 Age/years		73.6±11.1	71.6±10.6	75.4±11.3	-9.365	0.001
性别 Gender	男性 Male	2 337(77.0)	1 066(76.8)	1 271(77.2)	0.058	0.810
	女性 Female	698(23.0)	322(23.2)	376(22.8)
BMI/(kg·m^-2)		22.0±3.2	22.0±3.1	22.0±3.3	0.149	0.881
CCI		2(1, 2)	2(1, 2)	2(1, 3)	-9.506	＜0.001
Alb/(g·L^-1)		35.9±5.2	36.9±4.9	35.1±5.3	9.854	＜0.001
HDL-C/(mmol·L^-1)		16.1(1.3, 30.3)	16.0(1.3, 29.1)	16.2(1.3, 31.2)	-1.305	0.192
LDL-C/(mmol·L^-1)		2.25(1.8, 2.7)	2.3(1.9, 2.7)	2.2(1.8, 2.6)	2.968	0.003
Eos/(10⁹·L^-1)		0.1(0.1, 0.2)	0.1(0.1, 0.2)	0.1(0.1, 0.2)	-0.113	0.910
Neu/(10⁹·L^-1)		4.9(3.5, 7.2)	4.6(3.4, 6.6)	5.2(3.7, 7.5)	-4.371	＜0.001
WBC/(10⁹·L^-1)		6.9(5.3, 9.0)	6.7(5.2, 8.6)	7.1(5.4, 9.4)	-3.807	＜0.001
Fib/(g·L^-1)		3.5(2.8, 3.7)	3.5(2.8, 3.7)	3.6(2.8, 3.8)	-2.061	0.039
咳痰 Expectoration	是 Yes	2 704(89.1)	1 233(88.8)	1 471(89.3)	0.179	0.672
	否 No	331(10.9)	155(11.2)	176(10.7)
呼吸困难 Dyspnea	是 Yes	325(10.8)	114(8.2)	211(12.8)	16.654	＜0.001
	否 No	2 710(89.2)	1 274(91.8)	1 436(87.2)
胸闷 Chest distress	是 Yes	306(10.0)	131(9.4)	175(10.6)	1.171	0.279
	否 No	2 729(90.0)	1 257(90.6)	1 472(89.4)
喘息 Gasping	是 Yes	986(32.5)	431(31.1)	555(33.7)	2.404	0.121
	否 No	2 049(67.5)	957(68.9)	1 092(66.3)
食欲不振 Low appetite	是 Yes	1 599(52.6)	646(46.5)	953(57.9)	38.726	＜0.001
	否 No	1 436(47.4)	742(53.5)	694(42.1)
气短 Shortness of breath	是 Yes	1 777(58.5)	818(58.9)	959(58.2)	0.155	0.694
	否 No	1 258(41.5)	570(41.1)	688(41.8)
发热 Fever	是 Yes	692(22.9)	276(19.9)	416(25.3)	12.355	＜0.001
	否 No	2 343(77.1)	1 112(80.1)	1 231(74.7)
失眠 Insomnia	是 Yes	1 331(43.9)	538(38.8)	793(48.1)	26.957	＜0.001
	否 No	1 704(56.1)	850(61.2)	854(51.9)
β肾上腺素受体激动剂 β-adrenergic receptor agonist	是 Yes	2 213(72.9)	978(70.5)	1 235(75.0)	7.805	0.005
	否 No	822(27.1)	410(29.5)	412(25.0)
抗胆碱能药物 Anticholinergic drugs	是 Yes	2 368(78.0)	1 053(75.9)	1 315(79.8)	6.950	0.008
	否 No	667(22.0)	335(24.1)	332(20.2)
茶碱类 Theophylline drugs	是 Yes	2 590(85.3)	1 146(82.6)	1 444(87.7)	15.717	＜0.001
	否 No	445(14.7)	242(17.4)	203(12.3)
抗细菌药 Antibacterial drugs	是 Yes	2 720(89.6)	1 202(86.6)	1 518(92.2)	25.106	＜0.001
	否 No	315(10.4)	186(13.4)	129(7.8)
抗真菌药 Antifungal drugs	是 Yes	454(15.0)	103(7.4)	351(21.3)	114.248	＜0.001
	否 No	2 581(85.0)	1 285(92.6)	1 296(78.7)
注：CCI，共病指数；ALb，白蛋白；HDL-C，高密度脂蛋白胆固醇；LDL-C，低密度脂蛋白胆固醇；Eos，嗜酸性粒细胞；Neu，中性粒细胞；WBC，白细胞；Fib，纤维蛋白原。 ①以(x±s)、M(P₂₅，P₇₅)或人数(占比/%)表示。 Note：CCI，charlson comorbidity index；ALb，albumin；HDL-C，high-density lipoprotein cholesterol；LDL-C，low-density lipoprotein cholesterol；Eos，eosinophil；Neu，neutrophil；WBC，white blood cell；Fib，fibrinogen. ① (x±s), M(P₂₅，P₇₅)or number of people(proportion/%).

下载: 导出CSV

表 2 预测模型各评价指标

Table 2. Evaluation metrics of prediction models

评价指标 Evaluation metrics	LR^①	SVM^①	RF^①	CB^①	MLP^①	P值 value
AUC	0.743(0.679~0.808)	0.778(0.713~0.843)	0.780(0.716~0.845)	0.780(0.715~0.845)	0.772(0.707~0.837)	＜0.001
准确率/% Accuracy/%	67.7(61.2~74.2)	70.0(63.5~76.5)	69.7(63.2~76.2)	69.4(62.9~75.8)	71.0(64.5~77.5)	＜0.001
精准率/% Precision/%	65.2(58.7~71.7)	64.5(58.0~70.9)	64.5(58.0~71.0)	63.8(57.3~70.2)	67.0(60.5~73.5)	＜0.001
召回率/% Recall/%	63.3(56.8~69.8)	77.0(70.5~83.4)	75.2(68.7~81.7)	76.6(70.1~83.1)	72.3(65.8~78.8)	＜0.001
F1分数/% F1 score/%	64.2(57.8~70.7)	70.2(63.7~76.6)	69.4(63.0~75.9)	69.6(63.1~76.1)	69.6(63.1~76.0)	＜0.001
布里尔分数/% Brier score/%	21.1(14.6~27.5)	18.9(12.4~25.4)	18.8(12.3~25.2)	18.9(12.4~25.4)	19.3(12.9~25.8)	＜0.001
注：AUC，曲线下面积；LR，逻辑回归；SVM，支持向量机；RF，随机森林；MLP，多层感知机。 ①以x(95% CI)表示; ②是对5个模型中的6个指标进行单因素方差分析的结果。 Notes: AUC, area under curve; LR, logistic regression; SVM, support vector machine; RF, random forest; MLP, multilayer perceptron; CB, catboost. ① x(95% CI); ② is the result of one-way ANOVA for six metrics of the five models.

下载: 导出CSV

参考文献(16)

[1]	Adeloye D, Song PG, Zhu YJ, et al. Global, regional, and national prevalence of, and risk factors for, chronic obstructive pulmonary disease (COPD) in 2019: a systematic review and modelling analysis[J]. Lancet Respir Med, 2022, 10(5): 447-458. DOI: 10.1016/s2213-2600(21)00511-7.
[2]	López-Campos JL, Tan W, Soriano JB. Global burden of COPD[J]. Respirology, 2016, 21(1): 14-23. DOI: 10.1111/resp.12660.
[3]	Yin P, Wu JY, Wang LJ, et al. The burden of COPD in China and its provinces: findings from the global burden of disease study 2019[J]. Front Public Health, 2022, 10: 859499. DOI: 10.3389/fpubh.2022.859499.
[4]	中华医学会呼吸病学分会慢性阻塞性肺疾病学组, 中国医师协会呼吸医师分会慢性阻塞性肺疾病工作委员会. 慢性阻塞性肺疾病诊治指南(2021年修订版)[J]. 中华结核和呼吸杂志, 2021, 44(3): 170-205. DOI: 10.3760/cma.j.cn112147-20210109-00031. Chronic Obstructive Pulmonary Disease Committee, Respiratory Society, Chinese Medical Association. Guidelines for the diagnosis and management of chronic obstructive pulmonary disease(revised version 2021)[J]. Chin J Tuberc Respir Dis, 2021, 44(3): 170-205. DOI: 10.3760/cma.j.cn112147-20210109-00031.
[5]	Global Initiative for Chronic Obstructive Lung Disease (GOLD). Global strategy for pre-vention, diagnosis and management of COPD: 2023 report[EB/OL]. (2023-02-17)[2023-06-12]. https://goldcopd.org/2023-gold-report-2/.
[6]	中华医学会, 中华医学会杂志社, 中华医学会全科医学分会, 等. 常规肺功能检查基层指南(2018年)[J]. 中华全科医师杂志, 2019, 18(6): 511-518. DOI: 10.3760/cma.j.issn.1671-7368.2019.06.003. Chinese Medical Association, Chinese Medical Association Publishing House, Chinese Society of General Practice, et al. Guideline for pulmonary function testing in primary care (2018)[J]. Chin J Gen Pract, 2019, 18(6): 511-518. DOI: 10.3760/cma.j.issn.1671-7368.2019.06.003.
[7]	Ushida K, Shimizu A, Hori S, et al. Hospital frailty risk score predicts outcomes in chronic obstructive pulmonary disease exacerbations[J]. Arch Gerontol Geriatr, 2022, 100: 104658. DOI: 10.1016/j.archger.2022.104658.
[8]	Quintana JM, Unzurrunzaga A, Garcia-Gutierrez S, et al. Predictors of hospital length of stay in patients with exacerbations of COPD: a cohort study[J]. J Gen Intern Med, 2015, 30(6): 824-831. DOI: 10.1007/s11606-014-3129-x.
[9]	杨弘, 田晶, 王可, 等. 混合型缺失数据填补方法比较与应用[J]. 中国卫生统计, 2020, 37(3): 395-399. https://www.cnki.com.cn/Article/CJFDTOTAL-ZGWT202003016.htm Yang H, Tian J, Wang K, et al. Comparison and application of mixed missing data filling methods[J]. Chinese Journal of Health Statistics, 2020, 37(3): 395-399. https://www.cnki.com.cn/Article/CJFDTOTAL-ZGWT202003016.htm
[10]	Using grid search to optimize hyperparameters[EB/OL]. (2021-03-18)[2023-06-12]. https://www.section.io/engineering-education/grid-search/.
[11]	Rufibach K. Use of Brier score to assess binary predictions[J]. J Clin Epidemiol, 2010, 63(8): 938-939;authorreply939. DOI: 10.1016/j.jclinepi.2009.11.009.
[12]	Stolz D, Meyer A, Rakic J, et al. Mortality risk prediction in COPD by a prognostic biomarker panel[J]. Eur Respir J, 2014, 44(6): 1557-1570. DOI: 10.1183/09031936.00043814.
[13]	Waschki B, Kirsten A, Holz O, et al. Physical activity is the strongest predictor of all-cause mortality in patients with COPD: a prospective cohort study[J]. Chest, 2011, 140(2): 331-342. DOI: 10.1378/chest.10-2521.
[14]	Marin JM, Alfageme I, Almagro P, et al. Multicomponent indices to predict survival in COPD: the COCOMICS study[J]. Eur Respir J, 2013, 42(2): 323-332. DOI: 10.1183/09031936.00121012.
[15]	Esteban C, Quintana JM, Moraza J, et al. BODE-index vs HADO-score in chronic obstructive pulmonary disease: which one to use in general practice?[J]. BMC Med, 2010, 8: 28. DOI: 10.1186/1741-7015-8-28.
[16]	Ou CY, Chen CZ, Yu CH, et al. Discriminative and predictive properties of multidimensional prognostic indices of chronic obstructive pulmonary disease: a validation study in Taiwanese patients[J]. Respirology, 2014, 19(5): 694-699. DOI: 10.1111/resp.12313.

施引文献

资源附件(0)

访问统计

点击查看大图

图(1) / 表(2)

计量

文章访问数: 83
HTML全文浏览量: 32
PDF下载量: 9
被引次数: 0

姓名
邮箱
手机号码
标题
留言内容
验证码

留言板

慢性阻塞性肺疾病急性加重住院患者出院状态预测研究

doi: 10.16462/j.cnki.zhjbkz.2024.06.011

通讯作者:
张岩波，E-mail: sxmuzyb@126.com

赵卉，E-mail: hui_zhao@sxmu.edu.cn

计量

Predictive study on discharge status of hospitalized patients with acute exacerbation of chronic obstructive pulmonary disease

Corresponding author: ZHAO Hui, E-mail: hui_zhao@sxmu.edu.cn; ZHANG Yanbo, E-mail: sxmuzyb@126.com

计量

目录

留言板

慢性阻塞性肺疾病急性加重住院患者出院状态预测研究

doi: 10.16462/j.cnki.zhjbkz.2024.06.011

通讯作者: 张岩波，E-mail: sxmuzyb@126.com 赵卉，E-mail: hui_zhao@sxmu.edu.cn

计量

出版历程

Predictive study on discharge status of hospitalized patients with acute exacerbation of chronic obstructive pulmonary disease

Corresponding author: ZHAO Hui, E-mail: hui_zhao@sxmu.edu.cn; ZHANG Yanbo, E-mail: sxmuzyb@126.com

计量

出版历程

目录

通讯作者:
张岩波，E-mail: sxmuzyb@126.com

赵卉，E-mail: hui_zhao@sxmu.edu.cn