目的:使用来源人群以外的风险量表或预测模型,有必要通过外部验证来评估预测指标。目的是验证FAsscore,最初是在巴伦西亚地区初级保健的高血压患者中建造的,在巴斯克地区初级保健中患有高血压的外部队列中。
方法:一项回顾性队列研究旨在对毕尔巴鄂市26个健康中心附属患者的FAscore应用程序进行外部验证。ROC曲线下面积和预测指标用其95%置信区间计算。
结果:纳入三万六千八百九十九例患者:53.6%(n=19,719)为女性,平均年龄为75.1岁,41.8%(n=15,381)。在四年的随访期内,1420例患者被诊断为房颤(累积发生率3.9%)。FAscore估计的中位风险为4.5%,第五,25日,75,95百分位数是1.0%,2.5%,6.1%,和14.8%,分别。通过FAscore估计的风险和观察到的房颤病例的ROC曲线为AUC0.715(95%CI0.703-0.727)。5%的风险临界值提供了70.8%的敏感度,特异性为61.0%,阳性预测值为6.8%,阴性预测值为98.1%,正似然比和负似然比分别为1.82和0.48。
结论:本研究报告了高血压患者房颤风险量表的外部验证,这显示了可接受的预测能力。表现最好的风险界限,提供良好的预测指标,可以设定在5%。
OBJECTIVE: To use a risk scale or predictive model outside the population of origin, it is necessary to evaluate the predictive indicators through external validation. The aim was to validate the FAscore, originally constructed in hypertensive patients in primary care in the Valencian Region, in an external cohort with hypertension in primary care in the Basque Country.
METHODS: A retrospective cohort study was designed to perform an external validation of the FAscore app in patients affiliated with 26 health centers in the municipality of Bilbao. The area under the ROC curve and predictive indicators were calculated with their 95% confidence intervals.
RESULTS: Thirty-six thousand eight hundred nine patients were included: 53.6% (n=19,719) were women, the mean age was 75.1 years, 41.8% (n=15,381). Over the four-year follow-up period, 1420 patients were diagnosed with AF (cumulative incidence 3.9%). The median risk estimated by FAscore was 4.5%, and the 5th, 25th, 75th, and 95th percentiles were 1.0%, 2.5%, 6.1%, and 14.8%, respectively. The ROC curve for the risk estimated by FAscore and the cases of atrial fibrillation observed was AUC 0.715 (95% CI 0.703-0.727). The 5% risk cutoff provides a sensitivity of 70.8%, specificity of 61.0%, positive predictive value of 6.8%, negative predictive value of 98.1%, and positive and negative likelihood ratios of 1.82 and 0.48, respectively.
CONCLUSIONS: This study reports on the external validation of the atrial fibrillation risk scale in hypertensive patients, which shows an acceptable predictive capacity. The best-performing risk cutoff, providing good predictive indicators, can be set at 5%.