基于CatBoost模型与SHAP分析研究地理环境因子对健康人血尿酸参考值的影响  

The influence of geographical environment factors on reference value of serum uric acid in healthy people based on CatBoost model and SHAP analysis

作　　者：梁向荣 葛淼 王聪霞[2] 何进伟[3] LIANG Xiangrong;GE Miao;WANG Congxia;HE Jinwei(School of Geography and Tourism,Shaanxi Normal University,Xi’an 710119;Department of Cardiology,The Second Affiliated Hospital of Xi’an Jiaotong University,Xi’an 710004;Yan’an University Medical College,Yan’an 716099,China)

机构地区：[1]陕西师范大学地理科学与旅游学院健康地理研究所,陕西西安710119 [2]西安交通大学第二附属医院心内科,陕西西安710004 [3]延安大学医学院,陕西延安716099

出　　处：《西安交通大学学报（医学版）》2023年第4期601-607,共7页Journal of Xi’an Jiaotong University（Medical Sciences）

基　　金：国家自然科学基金资助项目(No.41761100)。

摘　　要：目的探究可能对健康人血尿酸(uric acid,UA)产生影响的地理环境因子并探究全国尺度下UA参考值的变化趋势。方法收集全国565个位点的607905例健康人的UA参考值,运用相关分析法分析25项地理环境因素与UA参考值的相关性,构建CatBoost模型并应用SHAP值解释模型,预测全国各县市级的健康人UA参考值,并采用普通克里金绘制全国健康人的UA参考值地理分布图。结果纬度、海拔高度、年平均气温、年平均相对湿度、年降水量、气温年较差、年平均风速、表土粉粒百分率、表土容重、表土石砾含量、表土有机质含量、表土pH、表土(粘土)阳离子交换量、表土(粉土)阳离子交换量、表土盐基饱和度、表土总可交换量、T-CaCO 3、T-CaSO 4、表土碱度、表土盐分这20项指标与全国健康人UA参考值呈现相关。全国健康人UA参考值的空间分布呈现差异性,表现为高海拔地区较高,沿海地区在相近海拔高度下高于内陆地区,中东部低、西南部高的变化趋势。结论本研究为后续近一步研究不同影响因子对UA参考值的作用机制奠定基础。建立CatBoost模型在不同地区使用UA参考值作为高尿酸血症及相关慢性疾病预后因子制定参考标准时提供依据。Objective To explore the geographical environment factors that may affect serum uric acid(UA)of healthy people and explore the change trend of UA reference value at the national scale.Methods The UA reference values of 607905 healthy people from 565 loci in China were collected,and the correlation between 25 geographical environment factors and UA reference values was analyzed by correlation analysis.CatBoost model was constructed and SHAP value interpretation model was applied to predict the UA reference values of healthy people in counties and cities in China,and the geographical distribution map of UA reference values of healthy people in China was drawn by using ordinary Kriging.Results A total of 20 indicators,namely,latitude,altitude,annual average temperature,annual average relative humidity,annual precipitation,air temperature annual range,annual average wind speed,percentage of surface soil silt,surface soil bulk density,surface soil gravel content,surface soil organic matter content,surface soil PH,surface soil(clay)cation exchange capacity,surface soil(silt)cation exchange capacity,surface soil base saturation,total surface soil exchange capacity,T-CaCO3,T-CaSO4,surface soil alkalinity,and surface soil salt showed their correlation with UA reference value of healthy people nationwide.The spatial distribution of UA reference values of healthy people across the country differed,manifested as the changing trend of higher in high altitude regions,higher in coastal regions than in inland regions,lower in the mid-eastern region,and higher in Southwest China at similar altitudes.Conclusion This study lays a foundation for further studies on the mechanism of different influencing factors on UA reference value.CatBoost model was established to provide the basis for establishing reference standards using UA reference values as prognostic factors for hyperuricemia and related chronic diseases in different regions.

关 键 词：高尿酸血症 尿酸(UA) 地理环境 CatBoost SHAP模型 克里金 

分 类 号：R188[医药卫生—流行病学]