Objective To systematically integrate echocardiographic parameters （including right heart structural and functional indicators） and clinical features （biomarkers, heart function classification, and underlying diseases） in heart failure （HF） patients, and to construct a multimodal predictive model for pulmonary hypertension （PH） and validate its diagnostic efficacy.

Methods A total of 146 HF patients were included in this study. Among them, 105 patients were diagnosed with HF combined with PH by right heart catheterization （RHC）, including 18 cases of isolated pre-capillary PH （pre-capillary PH）, 56 cases of isolated post-capillary PH （Ipc-PH）, and 31 cases of combined pre-capillary and post-capillary PH （Cpc-PH）. The accuracy of echocardiography （ECHO） in diagnosing PH was evaluated with RHC as the gold standard. Based on this, a multimodal predictive model was developed by integrating clinical features, biomarkers, and ECHO parameters. The model's predicted values were assessed for correlation with major clinical adverse events （including HF rehospitalization, respiratory failure, pulmonary encephalopathy, and death） using the Cox proportional hazards model. In the validation cohort, 50 HF patients （36 diagnosed with PH） were assessed using the receiver operating characteristic （ROC） curve to evaluate the model's diagnostic performance.

Results In this study, 71.92% （105/146） patients with heart failure were diagnosed with PH according to the new guideline criteria （mean pulmonary artery pressure, mPAP＞20 mmHg）. The efficacy of echocardiography assessed pulmonary arterial systolic pressure （PASP） was limited for diagnosing PH, with a sensitivity of 53.3% at the optimal cutoff of 46.5 mmHg. Based on the results of the binary logistic regression analysis, we constructed a multimodal predictive model with the following scoring formula: PH score = 0.185 × I （brain natriuretic peptide＞358 ng/L） + 1.243 × I （left ventricular ejection fraction＜54.5%） + 3.580 × I （end-systolic right atrial area＞20.5 cm²） + 4.180 × I （right ventricular end-diastolic diameter＞3.26 cm） + 2.883 × I （tricuspid annular plane systolic excursion/PASP ratio＜0.325） + 1.965 × I （atrial fibrillation） −6.865; where I（·） is the indicator function, taking a value of 1 when the condition is satisfied, otherwise 0. The corresponding probability for this score is: P = e^{PH score} / （1 + e^{PH score}）. The predictive performance of this model was significantly superior to that of any single ECHO indicators (AUC = 0.955 vs. 0.792 the highest AUC for a single indicator, achieved by tricuspid annular plane systolic excursion/PASP ratio, ΔAUC = 0.163, 95%CI: 0.078 to 0.242, P＜0.001). The model's predicted values were also significantly correlated with major clinical adverse events （HR = 14.985）. In the validation cohort, AUC = 0.960 （95%CI: 0.912 to 0.999）, with a sensitivity of 77.78% and specificity of 92.86%.

Conclusions This study successfully integrated core echocardiographic parameters and clinical feature variables to construct a multimodal predictive model for PH in HF patients. The model can accurately screen for the presence of pulmonary hypertension in heart failure patients non-invasively and its predicted values are associated with adverse prognosis.