Incremental prognostic value of ultrasound-based assessment of residual congestion at discharge after acute heart failure hospitalization

Background

Patients hospitalized for acute heart failure (HF) remain at high risk of early adverse outcomes after discharge. Clinical predictive models are available, but the incremental prognostic value of adding ultrasound assessment of residual congestion at discharge using lung (LUS) and venous excess ultrasound (VExUS) to these models has not been fully established.

Purpose

To evaluate the incremental prognostic value of adding ultrasound (LUS and VExUS) assessment of residual congestion at discharge to a clinical–laboratory predictive model for 3-month adverse outcomes after acute HF hospitalization.

Methods

This prospective, single-centre study included consecutive patients hospitalized for acute HF. Clinical and laboratory variables were collected, and residual congestion at discharge was assessed using ultrasound (LUS and VExUS). Two logistic regression predictive models were developed for a 3-month composite endpoint including all-cause mortality, HF hospitalization, and urgent HF visits requiring intravenous diuretics (cEP3). The clinical model included age, sex, discharge NT-proBNP, prior HF hospitalization, and discharge estimated glomerular filtration rate (eGFR), selected based on clinical relevance and a backward stepwise approach. The second model additionally included ultrasound-detected residual congestion. Model performance was compared in terms of discrimination, calibration, reclassification, and overall model fit.

Results

A total of 213 patients were included (median age 71 years, 66.2% men, median discharge NT-proBNP 1,398 pg/mL, 31.9% with prior HF hospitalization, median discharge eGFR 55 mL/min, and median left ventricular ejection fraction 39%). Residual congestion at discharge assessed by ultrasound was present in 45.1% of patients. At 3 months, the composite endpoint (cEP3) occurred in 23.9% of patients (n=51).

The clinical model showed good discrimination (AUC 0.80, 95% CI 0.74–0.87). The addition of ultrasound-detected residual congestion significantly improved model discrimination (AUC 0.85, 95% CI 0.79–0.91; DeLong p=0.025) (Figure 1). Both models were well calibrated (Hosmer–Lemeshow p=0.50 and p=0.78, respectively). The clinical + ultrasound-based congestion model provided significant improvement in risk reclassification, with a categorical NRI of 0.27 (95% CI 0.05–0.48; p=0.021), continuous NRI of 0.73 (95% CI 0.44–1.01; p<0.00001), and IDI of 0.08 (95% CI 0.04–0.13; p=0.001). Model fit was superior for the clinical + ultrasound-based congestion model, as reflected by lower AIC and BIC values and a significant likelihood ratio test (Δdeviance 17.1; p=0.00004) (Figure 2).

Conclusion

In patients hospitalized for acute HF, the addition of ultrasound (LUS and VExUS) assessment of residual congestion at discharge to a clinical predictive model significantly improves 3-month risk prediction of adverse events and enhances discrimination, risk reclassification, and overall model fit.Figure 1For image description, please refer to the figure legend and surrounding text.Figure 2For image description, please refer to the figure legend and surrounding text.