Heart failure with preserved ejection fraction (HFpEF) is under-recognized in clinical practice. Although a previously developed risk score, termed H₂FPEF, can be used to estimate HFpEF probability, this score requires imaging data, which is often unavailable. Here we sought to develop an HFpEF screening model that is based exclusively on clinical variables and that can guide the need for echocardiography and further testing. In a derivation cohort (n = 414, 249 women), a clinical model using age, body mass index and history of atrial fibrillation (termed the HFpEF-ABA score) showed good discrimination (area under the curve (AUC) = 0.839 (95% confidence interval (CI) = 0.800–0.877), P < 0.0001). The performance of the model was validated in an international, multicenter cohort (n = 736, 443 women; AUC = 0.813 (95% CI = 0.779–0.847), P < 0.0001) and further validated in two additional cohorts: a cohort including patients with unexplained dyspnea (n = 228, 136 women; AUC = 0.840 (95% CI = 0.782–0.900), P < 0.0001) and a cohort for which HF hospitalization was used instead of hemodynamics to establish an HFpEF diagnosis (n = 456, 272 women; AUC = 0.929 (95% CI = 0.909–0.948), P < 0.0001). Model-based probabilities were also associated with increased risk of HF hospitalization or death among patients from the Mayo Clinic (n = 790) and a US national cohort across the Veteran Affairs health system (n = 3076, 110 women). Using the HFpEF-ABA score, rapid and efficient screening for risk of undiagnosed HFpEF can be performed in patients with dyspnea using only age, body mass index and history of atrial fibrillation.

射血分数保留的心力衰竭（HFpEF）在临床实践中尚未得到充分认识。尽管先前开发的风险评分（称为 H ₂ FPEF）可用于估计 HFpEF 概率，但该评分需要影像数据，而影像数据通常无法获得。在这里，我们试图开发一种 HFpEF 筛查模型，该模型完全基于临床变量，可以指导超声心动图和进一步测试的需要。在衍生队列（ n = 414，249 名女性）中，使用年龄、体重指数和心房颤动病史（称为 HFpEF-ABA 评分）的临床模型显示出良好的区分度（曲线下面积 (AUC) = 0.839 (95 % 置信区间 (CI) = 0.800–0.877)， P < 0.0001)。该模型的性能在国际多中心队列中得到验证（ n = 736，443 名女性；AUC = 0.813（95% CI = 0.779–0.847）， P < 0.0001），并在另外两个队列中得到进一步验证：一个包括患者的队列原因不明的呼吸困难（ n = 228，136 名女性；AUC = 0.840（95% CI = 0.782–0.900）， P < 0.0001）以及使用 HF 住院治疗而不是血流动力学来建立 HFpEF 诊断的队列（ n = 456， 272 名女性；AUC = 0.929（95% CI = 0.909–0.948）， P < 0.0001）。基于模型的概率还与来自梅奥诊所 ( n = 790) 和整个退伍军人事务卫生系统的美国全国队列 ( n = 3076，110 名女性) 的患者心衰住院或死亡风险增加相关。使用 HFpEF-ABA 评分，仅根据年龄、体重指数和房颤病史即可对呼吸困难患者进行快速有效的未确诊 HFpEF 风险筛查。