Warning Indicators for Cardiogenic Death during the Vulnerable Phase in Chronic Heart Failure with Concurrent Essential Hypertension
Objective To investigate the warning indicators for cardiogenic death during the vulnerable phase in patients with coexisting essential hypertension and chronic heart failure,comparing Chinese and Western medicine approaches based on multiple support vector machine-recursive feature elimination(MSVM-RFE)and Cox regression analysis.Methods Patients with chronic heart failure and essential hypertension who met the study criteria from 7 Chinese medicine(CM)hospitals in Shandong Province were included in this prospective cohort study.A comprehensive collection of 78 clinical variables were obtained,including patient demographics,comorbidities,cardiac ultrasound results,laboratory test outcomes,CM syndrome information,and medication records.Endpoint events were counted at follow-up 3 months after discharge.Cox univariate regression analysis was employed to identify variables that exhibited a significant association with cardiogenic death during the vulnerable phase.These variables were subsequently incorporated as potential warning indicators in the MSVM-RFE model,and their relative importance was determined based on average ranking scores.The efficiency estimation was used to evaluate the model error rate and accuracy index in order to select the optimal subset of warning indicators.Additionally,the predictive property of the warning indicators were assessed through Cox multivariate regression analysis.Results Totally 1 721 patients were included in analyses,among whom 272 experienced cardiogenic death during the vulnerable phase.Subsequently,the Cox univariate regression analysis yielded 64 potential warning indicators,which were then prioritized based on their significance and successively incorporated into the MSVM-RFE model for efficiency estimation.The inclusion of the top 13 potential warning indicators resulted in the model attaining a minimum error rate of 0.136 and a maximum accuracy of 0.864,suggesting that this particular subset of warning indicators was optimal.Furthermore,Cox multivariate regression analysis demonstrated significant associations between the following factors and cardiogenic death during the vulnerable phase:plasma N-terminal pro-brain natriuretic peptide(NT-ProBNP)levels exceeding 1 000 ng/L(HR=3.65),the use of angiotensin-converting enzyme inhibitors/angiotensin receptor blockers(ACEI/ARB,HR=0.66),the presence of qi-yin deficiency and blood stasis syndrome(HR=1.98),the manifestation of dark complexion and lips(HR=2.19),the administration of diuretic medication(HR=0.73),heart rate falling below 60 or exceeding 100 beats per minute(HR=1.30,HR=1.33),age surpassing 60 years(HR=3.64),New York Heart Association(NYHA)functional classification(Ⅲ vs Ⅱ:HR=2.50;Ⅳ vs Ⅱ:HR=3.58),a history of smoking(HR=1.78),and systolic blood pressure not exceeding 140 mmHg(HR=2.56),blood-activating Chinese herbal injections(HR=0.64),presence of a slippery pulse(HR=1.32),and the presence of scanty or cracked coating on the tongue(HR=1.45).Conclusions The study identified several detrimental warning indicators,including NT-ProBNP levels exceeding 1 000 ng/L,qi-yin deficiency and blood stasis syndrome,dark complexion and lips,heart rate falling below 60 or exceeding 100 per minute,age surpassing 60 years,NYHA functional classification reaching or exceeding Ⅲ,smoking history,systolic blood pressure not exceeding 140 mmHg,slippery pulse,and scanty or cracked coating on the tongue.On the other hand,the use of ACEI/ARB,diuretic medication,and blood-activating Chinese herbal injections were identified as protective warning indicators.The combined MSVM-RFE and Cox regression model proves to be a suitable approach for analyzing warning indicators.
chronic heart failureessential hypertensionvulnerable phase of heart failureintegrative medicinewarning indicatorsmultiple support vector machine-recursive feature eliminationCox regression analysis
国家科技期刊平台
NSTL
万方数据
