Heart failure & cardiomyopathies
Personalized evaluation of NT-proBNP and clinical characteristics in suspected heart failure: the PRECISE-HF study
Authors' conclusion
The PRECISE-HF model, integrating multiple clinical characteristics with NT-proBNP, demonstrated superior diagnostic performance compared with universal rule-out and age-adjusted rule-in thresholds in community-based patients with suspected HF.
Study context
- Population
- Community-based patients with NT-proBNP measured for suspected heart failure (HF), no known HF; UK primary care (Clinical Practice Research Datalink (CPRD)), 2018–2025internal validation n=160,674
- Participants
- 535,583 patients
- Comparison
- PRECISE-HF model vs European Society of Cardiology (ESC) thresholds
- Outcome
- HF diagnosis (primary care diagnosis or HF hospitalisation)
- Follow-up
- Within 12 months of NT-proBNP measurement
Methodological detailscohort study, machine learning (XGBoost), internal and external validation
- Cohort study
- Machine learning (XGBoost)
- Internal and external validation
Performance
As reported in the abstract
Diagnostic performance, internal validation
| Test or model | AUROC | Brier score | Rule-out sensitivity | Rule-out NPV | Ruled out | Ruled in | Rule-in specificity | Rule-in PPV |
|---|---|---|---|---|---|---|---|---|
| PRECISE-HF | 0.896 | 0.061 | 90.1% | 98.5% | 64.8% | 16.1% | 90.0% | 44.1% |
PRECISE-HF
AUROC0.896
Brier score0.061
Rule-out sensitivity90.1%
Rule-out NPV98.5%
Ruled out64.8%
Ruled in16.1%
Rule-in specificity90.0%
Rule-in PPV44.1%
- vs ESC thresholds: additional patients ruled out per 1000
- 144
- vs ESC thresholds: missed diagnoses per 1000
- 3
- vs ESC thresholds: fewer false positives per 1000 at rule-in
- 73
Further findings
External validation (REVOLUTION-HF Swedish cohort)
- Model retained good discrimination (AUROC)
- 0.757
- Calibration (Brier score)
- 0.163
Original publication
Personalized evaluation of NT-proBNP and clinical characteristics in suspected heart failure: the PRECISE-HF study. Eur Heart J. Published 28 August 2026. doi:10.
Summarised from the published abstract; every number is checked against it