基于列线图的原发性头痛辅助决策模型的构建

doi:10.3969/j.issn.1674-8115.2021.10.009

摘要/Abstract

摘要：

目的·建立基于列线图的原发性头痛辅助决策模型。方法·回顾性收集2019年10月—2020年12月就诊于上海交通大学医学院附属第九人民医院的偏头痛或紧张型头痛患者210例；其中2019年10月—2020年8月就诊的152例患者作为建模组，2020年9月—2020年12月就诊的患者58例作为验证组。应用单因素及多因素Logistic回归分析筛选出区分偏头痛和紧张型头痛的独立预测因素。基于自变量的回归系数，应用R软件建立偏头痛和紧张型头痛个体化列线图决策模型。通过Bootstrap进行模型的内部验证，使用验证组数据进行模型的外部验证。分别采用受试者操作特征曲线（receiver operator characteristic curve，ROC曲线）、曲线下面积（area under the curve，AUC）及校准曲线评价模型的区分度及校准度。结果·建模组中偏头痛患者80例，紧张型头痛患者72例；验证组中偏头痛患者35例，紧张型头痛患者23例。建模组和验证组患者在一般人口学资料和头痛特征上差异无统计学意义。根据单因素Logistic分析结果提取9个特征变量纳入多因素分析。多因素Logistic回归分析得出病程、头痛是否位于后枕部、头痛的严重程度、是否伴有恶心/呕吐、是否伴有畏光/畏声、活动后头痛的变化是区分偏头痛和紧张型头痛的独立预测指标。以此结果构建列线图决策模型。对模型进行内部和外部验证发现，建模组和验证组的AUC值分别为0.896［95%置信区间（confidence interval，CI）0.842~0.950］和0.884（95%CI 0.793~0.976），说明模型具有良好的区分度；建模组和验证组的校准曲线与标准曲线均极为接近，具有良好的校准度，说明该模型在2组中较为一致。结论·研究构建了基于列线图的偏头痛和紧张型头痛决策模型，模型具有较好的区分度和校准度，有利于提高临床医师对偏头痛和紧张型头痛的早期识别和诊断能力。

关键词: 原发性头痛, 偏头痛, 紧张型头痛, 决策模型, 模型验证, 列线图

Abstract:

Objective·To establish a decision-making model for primary headache based on Nomogram.

Methods·Two hundred and ten patients with migraine or tension-type headache who visited the Department of Neurology, Shanghai Ninth People's Hospital, Shanghai Jiao Tong University School of Medicine between October 2019 to December 2020 were studied retrospectively. Among them, 152 patients who visited the hospital from October 2019 to August 2020 were served as the modeling group. Fifty eight patients who visited the hospital from September 2020 to December 2020 were served as the validation group. Univariate and multivariate Logistic regression were used to analyze the independent predictive factors to distinguish migraine and tension-type headache. According to the regression coefficient of independent variables, R software was used to construct a nomogram model of migraine or tension-type headache. The internal verification of the model was carried out through Bootstrap, and the external verification was carried out according to the data of the validation group. Receiver operator characteristic curve (ROC curve), area under the curve (AUC) and calibration curve were used respectively to estimate the discrimination and calibration of the prediction model.

Results·There were 80 patients with migraines and 72 patients with tension-type headaches in the modeling group. There were 35 patients with migraines and 23 patients with tension-type headaches in the validation group. There was no statistical difference in characteristics between the modeling group and the validation group. According to the results of univariate Logistic analysis, 9 characteristic variables were extracted and included in the multivariate analysis. Multivariate Logistic regression analysis showed the independent predictive factors that were used to distinguish migraine and tension-type headache, including course, whether the headache was located in the occipital, severity intensity of the headache, whether the headache was accompanied by nausea/vomiting, whether the headache was accompanied by photophobia/phonophobia and the change of headache after activities. The decision Nomogram model was constructed based on this result. The internal and external verification of the model found that AUC of the modeling group and the validation group were 0.896 [95% confidence interval (CI) 0.842?0.950] and 0.884 (95%CI 0.793?0.976) respectively, suggesting that the prediction model has a good discrimination capacity.The calibration curve of the modeling group and the validation group was very close to the standard curve, and had a good calibration degree, which showed that the model was consistent in the two groups.

Conclusion·This study has constructed a decision-making model for distinguishing migraine and tension-type headache based on Nomogram. The model has a good discrimination and calibration in the modeling group and the validation group, which is beneficial to improve clinicians' early identification and diagnosis capabilities for migraine and tension-type headache.

Key words: primary headache, migraine, tension-type headache, decision-making model, model validation, Nomogram

中图分类号:

R741.041

刘芳芳, 包关水, 闫梦侠. 基于列线图的原发性头痛辅助决策模型的构建[J]. 上海交通大学学报(医学版), 2021, 41(10): 1323-1329.

Fang-fang LIU, Guan-shui BAO, Meng-xia YAN. Construction of a decision-making model for primary headache based on Nomogram[J]. JOURNAL OF SHANGHAI JIAOTONG UNIVERSITY (MEDICAL SCIENCE), 2021, 41(10): 1323-1329.

图/表 6

参考文献 16

1	Hagen K, Åsberg AN, Uhlig BL, et al. The epidemiology of headache disorders: a face-to-face interview of participants in HUNT4[J]. J Headache Pain, 2018, 19(1): 25.
2	Yu SY, Liu RZ, Zhao G, et al. The prevalence and burden of primary headaches in China: a population-based door-to-door survey[J]. Headache, 2012, 52(4): 582-591.
3	Yao CY, Wang Y, Wang LJ, et al. Burden of headache disorders in China, 1990—2017: findings from the Global Burden of Disease Study 2017[J]. J Headache Pain, 2019, 20(1): 102.
4	Saylor D, Steiner TJ. The global burden of headache[J]. Semin Neurol, 2018, 38(2): 182-190.
5	Takeshima T, Wan Q, Zhang Y, et al. Prevalence, burden, and clinical management of migraine in China, Japan, and South Korea: a comprehensive review of the literature[J]. J Headache Pain, 2019, 20(1): 111.
6	Leonardi M, Raggi A. A narrative review on the burden of migraine: when the burden is the impact on people's life[J]. J Headache Pain, 2019, 20(1): 41.
7	International Headache Society. Headache classification committee of the international headache society (IHS) the international classification of headache disorders, 3rd edition[J]. Cephalalgia, 2018, 38(1): 1-211.
8	尹梓名, 董钊, 孔祥勇. 基于国际头痛诊断标准的原发性头痛辅助决策系统[J]. 计算机应用研究, 2019, 36(02): 461-465.
9	Krawczyk B, Simić D, Simić S, et al. Automatic diagnosis of primary headaches by machine learning methods[J]. Central Eur J Med, 2013, 8(2): 157-165.
10	Khayamnia M, Yazdchi M, Heidari A, et al. Diagnosis of common deadaches using hybrid expert-based systems[J]. J Med Signals Sens, 2019, 9(3): 174-180.
11	Simić S, Banković Z, Villar JR, et al. A hybrid fuzzy clustering approach for diagnosing primary headache disorder[J]. Log J IGPL, 2021, 29(2): 220-235.
12	Jun FEH. Regression modeling strategies with applications to linear models, logistic regression, and survival analysis[M]. New York: Springer Verlag, 2001.
13	GBD 2017 Disease and Injury Incidence and Prevalence Collaborators. Global, regional, and national incidence, prevalence, and years lived with disability for 354 diseases and injuries for 195 countries and territories, 1990—2017: a systematic analysis for the Global Burden of Disease Study 2017[J]. Lancet, 2018, 392(10159): 1789-1858.
14	GBD 2016 Neurology Collaborators. Global, regional, and national burden of neurological disorders, 1990—2016: a systematic analysis for the Global Burden of Disease Study 2016[J]. Lancet Neurol, 2019, 18(5): 459-480.
15	Martin VT, Penzien DB, Houle TT, et al. The predictive value of abbreviated migraine diagnostic criteria[J]. Headache, 2005, 45(9): 1102-1112.
16	Lipton RB, Dodick D, Sadovsky R, et al. A self-administered screener for migraine in primary care: the ID migraine validation study[J]. Neurology, 2003, 61(3): 375-382.

Variable	Modeling group （n=152）	Validation group （n=58）	χ² value	P value
Gender / n（%）			0.006	0.939
Male	48 （31.6）	18 （31.0）
Female	104 （68.4）	40 （69.0）
Course / n（%）			0.004	0.947
Within half a year	40 （26.3）	15 （25.9）
More than half a year	112 （73.7）	43 （74.1）
Throbbing / n（%）			3.194	0.074
Yes	19 （12.5）	13 （22.4）
No	133 （87.5）	45 （77.6）
Occipital / n（%）			2.298	0.130
Yes	51 （33.6）	26 （44.8）
No	101 （66.4）	32 （55.2）
Tempus / n（%）			0.900	0.343
Yes	74 （48.7）	24 （41.4）
No	78 （51.3）	34 （58.6）
Severe intensity / n（%）			4.280	0.118
Light	31 （21.7）	17 （29.3）
Medium	85 （55.9）	34 （58.6）
Heavy	36 （22.4）	7 （12.1）
Visual aura symptom / n（%）			0.286	0.593
Yes	13 （8.6）	3 （5.2）
No	139 （91.4）	55 （94.8）
Nausea/vomiting/ n（%）			2.056	0.152
Yes	58 （38.2）	16 （27.6）
No	94 （61.8）	42 （72.4）
Photophobia/phonophobia/ n（%）			2.921	0.087
Yes	31 （20.4）	6 （10.3）
No	121 （79.6）	52 （89.7）
Change of headache after activities/ n（%）			4.553	0.124
Aggravate	55 （36.2）	22 （38.0）
Unchanged	83 （54.6）	35 （60.3）
Relieve	14 （9.2）	1 （1.7）
Related to menstruation/ n（%）			0.111	0.739
Yes	21 （13.8）	7 （12.1）
No	131 （86.2）	51 （87.9）
Alleviative way/ n（%）			6.955	0.064
Persistence	23 （15.1）	4 （6.9）
Rest	53 （34.9）	29 （50.0）
Drug	69 （45.4）	25 （43.1）
Else	7 （4.6）	0
Diagnose / n（%）			1.008	0.315
Migraine	80 （52.6）	35 （61.1）
Tension-type headache	72 （47.4）	23 （39.7）

Variable	Modeling group （n=152）	Validation group （n=58）	χ² value	P value
Gender / n（%）			0.006	0.939
Male	48 （31.6）	18 （31.0）
Female	104 （68.4）	40 （69.0）
Course / n（%）			0.004	0.947
Within half a year	40 （26.3）	15 （25.9）
More than half a year	112 （73.7）	43 （74.1）
Throbbing / n（%）			3.194	0.074
Yes	19 （12.5）	13 （22.4）
No	133 （87.5）	45 （77.6）
Occipital / n（%）			2.298	0.130
Yes	51 （33.6）	26 （44.8）
No	101 （66.4）	32 （55.2）
Tempus / n（%）			0.900	0.343
Yes	74 （48.7）	24 （41.4）
No	78 （51.3）	34 （58.6）
Severe intensity / n（%）			4.280	0.118
Light	31 （21.7）	17 （29.3）
Medium	85 （55.9）	34 （58.6）
Heavy	36 （22.4）	7 （12.1）
Visual aura symptom / n（%）			0.286	0.593
Yes	13 （8.6）	3 （5.2）
No	139 （91.4）	55 （94.8）
Nausea/vomiting/ n（%）			2.056	0.152
Yes	58 （38.2）	16 （27.6）
No	94 （61.8）	42 （72.4）
Photophobia/phonophobia/ n（%）			2.921	0.087
Yes	31 （20.4）	6 （10.3）
No	121 （79.6）	52 （89.7）
Change of headache after activities/ n（%）			4.553	0.124
Aggravate	55 （36.2）	22 （38.0）
Unchanged	83 （54.6）	35 （60.3）
Relieve	14 （9.2）	1 （1.7）
Related to menstruation/ n（%）			0.111	0.739
Yes	21 （13.8）	7 （12.1）
No	131 （86.2）	51 （87.9）
Alleviative way/ n（%）			6.955	0.064
Persistence	23 （15.1）	4 （6.9）
Rest	53 （34.9）	29 （50.0）
Drug	69 （45.4）	25 （43.1）
Else	7 （4.6）	0
Diagnose / n（%）			1.008	0.315
Migraine	80 （52.6）	35 （61.1）
Tension-type headache	72 （47.4）	23 （39.7）

Variable	Univariate Logistic regression			Multivariate Logistic regression
Variable	OR	95%CI	P value	OR	95%CI	P value
Gender	2.047	1.011‒4.146	0.047	NA	NA	NA
Course	3.720	1.660‒8.336	0.001	3.364	1.145‒9.888	0.027
Throbbing	0.616	0.233‒1.629	0.329	‒	‒	‒
Occipital	2.106	1.051‒4.220	0.036	3.271	1.166‒9.181	0.024
Tempus	0.658	0.347‒1.249	0.200	‒	‒	‒
Severe intensity	2.566	1.504‒4.379	0.001	2.036	1.010‒4.104	0.047
Visual Aura symptoms	0.368	0.108‒1.253	0.110	‒	‒	‒
Nausea/vomiting	0.100	0.046‒0.218	0.000	0.094	0.034‒0.260	0.000
Photophobia/phonophobia	0.061	0.018‒0.213	0.000	0.081	0.017‒0.385	0.002
Change of headache after activities	0.396	0.225‒0.696	0.001	0.414	0.199‒0.863	0.019
Related to menstruation	0.233	0.081‒0.675	0.007	NA	NA	NA
Alleviative way	1.504	0.995‒2.274	0.053	NA	NA	NA

Variable	Univariate Logistic regression			Multivariate Logistic regression
Variable	OR	95%CI	P value	OR	95%CI	P value
Gender	2.047	1.011‒4.146	0.047	NA	NA	NA
Course	3.720	1.660‒8.336	0.001	3.364	1.145‒9.888	0.027
Throbbing	0.616	0.233‒1.629	0.329	‒	‒	‒
Occipital	2.106	1.051‒4.220	0.036	3.271	1.166‒9.181	0.024
Tempus	0.658	0.347‒1.249	0.200	‒	‒	‒
Severe intensity	2.566	1.504‒4.379	0.001	2.036	1.010‒4.104	0.047
Visual Aura symptoms	0.368	0.108‒1.253	0.110	‒	‒	‒
Nausea/vomiting	0.100	0.046‒0.218	0.000	0.094	0.034‒0.260	0.000
Photophobia/phonophobia	0.061	0.018‒0.213	0.000	0.081	0.017‒0.385	0.002
Change of headache after activities	0.396	0.225‒0.696	0.001	0.414	0.199‒0.863	0.019
Related to menstruation	0.233	0.081‒0.675	0.007	NA	NA	NA
Alleviative way	1.504	0.995‒2.274	0.053	NA	NA	NA

Variable	Tol	VIF
Gender	0.861	1.162
Course	0.802	1.246
Occipital	0.923	1.083
Severe intensity	0.871	1.148
Nausea/vomiting	0.792	1.263
Photophobia/phonophobia	0.853	1.172
Change of headache after activities	0.903	1.107
Related to menstruation	0.861	1.161
Alleviative way	0.801	1.249