Framingham高血压发病风险预测模型对新疆哈萨克族牧民发生高血压的预测价值研究

Predictive Value Analysis of Framingham Hypertension Risk Prediction Model in Xinjiang Kazakh Herdsmen

背景新疆哈萨克族牧民高血压患病率高,寻找有效的方法对哈萨克族牧民高血压的发病风险进行评估,并进行合理的干预,预防高血压事件的发生十分重要。目的评价Framingham高血压发病风险模型(以下简称Framingham模型)预测新疆哈萨克族牧民发生高血压的准确性和适用性,为建立适用于新疆哈萨克族牧民高血压发病风险预测模型提供依据。方法2008年1月,采用分层整群随机抽样法,以新疆乌鲁木齐县南山牧区的5327例哈萨克族牧民为研究对象,建立南山动态队列。收集其基线资料(包括问卷调查表和体格检查表),并进行2年1次的随访调查,共3次。随访截至2018年11月,结局事件为发生高血压。将研究对象随机分为建模队列(60%哈萨克族牧民,3196例)和验证队列(40%哈萨克族牧民,2131例),按照Framingham模型相同的方法及预测因素,采用多因素weibull回归分析、依据建模队列调整Framingham模型,即为依据建模队列调整后的Framingham模型。采用区分能力和标定能力验证上述依据本研究对象数据调整后的Framingham模型对验证队列高血压发病风险进行预测。结果截至2018年11月,1985例哈萨克族牧民发生高血压。累计共随访16897人年,高血压发病率为11.75/100人年[95%CI(11.27/100人年,12.24/100人年)]。本研究验证队列人群经2年、4年随访,分别有269例、562例发生高血压。依据建模队列调整后的Framingham模型与原Framingham模型预测随访2年验证队列人群发生高血压的受试者工作特征曲线下面积(AUC)分别为:AUC依据建模队列调整后=0.647[95%C(0.I 624,0.670)]和AUC原Framingham模型=0.594[95%C(0.I 571,0.617)];依据建模队列调整后的2Framingham模型与原Framingham模型预测随访2年验证队列人群发生高血压的AUC比较,差异有统计学意义(χ=5.085,P<0.05);依据建模队列调整后的Framingham模型与原Framingham模型预测随访4年验证队列人群发生高血压的AUC分别为:AUC=0.609[95%CI(0.590,0.628)]和依据建模队列调整后AUC原Framingham模型=0.588[95%CI(0.569,0.607)];依据建模队列调整后的Framingham模型与原Framingham模型预测随访4年验证队列人群发生高血压的AUC比较,差异有统计学意义(χ2=3.448,P<0.001)。依据建模队列调整后的Framingham模型与原Framingham模型预测随访2年验证队列人群高血压发病率与实际发病率比较Hosmer-Lemeshowχ2(H-Lχ2)检验值分别为697.68(P<0.05)、802.40(P<0.05);预测随访4年验证队列人群高血压发病率与实际发病率比较,H-Lχ2检验值分别为682.61(P<0.05)、832.82(P<0.05)。结论Framingham模型在预测新疆哈萨克族牧民高血压的发病风险中区分能力和标定能力均较差,不能很好地预测新疆哈萨克族牧民高血压的发病风险,需要构建更适合该人群的高血压发病风险预测模型。

Background The incidence of hypertension among Kazakh herdsmen in Xinjiang is high,so it is important to find effective methods to assess the risk of hypertension among Kazakh herdsmen and to intervene reasonably to prevent the occurrence of hypertension.Objective To evaluate the accuracy and applicability of Framingham hypertension risk prediction model(hereinafter referred to as Framingham model)in predicting hypertension among Kazakh herdsmen in Xinjiang,in order to provide a reference for the establishment of risk prediction model for hypertension among Kazakh herdsmen in Xinjiang.Methods In January 2008,5327 Kazakh herdsmen in the Nanshan Pastoral Area of Urumqi County in Xinjiang were selected as research objects by stratified cluster random sampling method to establish the Nanshan dynamic cohort.Baseline data(including questionnaire and physical examination form)were collected,and research objects were followed up every 2 years for 3 times until November 2018,and the outcome event was the development of hypertension.The research objects were randomly divided into model queue(60%Kazakh herdsmen,3196 cases)and validation queue(40%Kazakh herdsmen,2131 cases).According to the same method and predictive factors of the Framingham model,the model queue was adjusted and analyzed by the multi-factor Weibull regression analysis,that is,the adjusted model queue was the Framingham model in this study.The discriminatory power and ability of calibration were used to verify the predictive ability of the above Framingham model adjusted according to the data of this study to verify the risk of hypertension in the validation queue.Results As of November 2018,there were 1,985 cases of hypertension among Kazakh herdsmen.A total of 16897 person-years were followed up,and the incidence of hypertension was 11.75 cases per 100 person-years[95%CI(11.27/100 person-years,12.24/100 person-years)].In validation queue,269 and 562 patients developed hypertension after two and four years of follow-up,respectively.The AUC of the adjusted Framingham model and the original Framingham model for verifying whether hypertension occurred in the validation queue after two years of follow-up were AUC adjusted Framingham model=0.647[95%CI(0.624,0.670)]and AUC original Framingham model=0.594[95%CI(0.571,0.617)],and the difference was statistically significant(χ2=5.085,P<0.05).The AUC of the adjusted Framingham model and the original Framingham model for verifying whether hypertension occurred after four years of follow-up in the validation queue were AUC adjusted Framingham model=0.609[95%CI(0.590,0.628)]and AUC original Framingham model=0.588[95%CI(0.569,0.607)],and the difference was statistically significant(χ2=3.448,P<0.001).In the adjusted Framingham model and the original Framingham model,the incidence of hypertension in the validation queue was compared with the actual incidence after two years of follow-up,and the Hosmer-lemeshowχ2(H-Lχ2)test values were 697.68(P<0.05)and 802.40(P<0.05),respectively.In the adjusted Framingham model and the original Framingham model,the incidence of hypertension in the validation queue was compared with the actual incidence after four years of follow-up and the H-Lχ2 test values were 682.61(P<0.05)and 832.82(P<0.05),respectively.Conclusion The Framingham model has poor ability to distinguish and calibrate the risk of hypertension among Kazak herdsmen in Xinjiang,and it can not predict the risk of hypertension well.It is necessary to build a more suitable hypertension risk prediction model for this population.