Proven. Reliable. Validated at Scale.
Clinical Validation: Results from Peer-Reviewed Studies
and Real-World Pilots
Clinical validation of Cardio-HART has been confirmed in peer-reviewed studies and real-world pilots. The results demonstrate proven accuracy, reduced inconclusive ECGs, and improved diagnostic confidence for clinicians. Beyond accuracy, the clinical validation of Cardio-HART highlights its reliability across diverse patient populations and care settings, ensuring consistent results in everyday clinical practice.
| Key metric (primary end-points) | Result | Evidence |
|---|---|---|
| ↓ Inconclusive ECGs | -94 % (31 % → 1.9 %) | Calcagno et al., Open Heart 2022 (Open Heart) |
| ↑ Confirmed GP diagnoses | +160 % (10 % → 26 %) | Calcagno et al., Open Heart 2022; Cardio-Phoenix press release (Open Heart, cardiophoenix.com) |
| HF detection (HFpEF + HFmrEF + HFrEF) | Se 91.2 % / Sp 88.6 % | Calcagno et al., BMJ Open Heart 2022 (Open Heart) |
| Echo-agreement across 14 EU sites (n = 6 127) | κ 0.78; overall agreement 96 % | Eur. Heart J. – Digital Health 2024 (Oxford Academic) |
| LVEF estimation vs Simpson Biplane | r = 0.78, RMSE ≈ 8 % | Eur. Heart J. – Digital Health 2024 (Oxford Academic) |
| Negative predictive value (multi-disease) | ≥ 97 % | Cardio-Phoenix Data on File 2023 (cardiophoenix.com) |
| Real-world nurse-led screening (Spain 2025) | 47 tests/5 h; 3 early-HF cases detected; 10-min turnaround | “Diagnósticos F.C.” event report (cardiophoenix.com) |
Scope covered: 14 “HART-findings” mirroring standard echo criteria; ~93 % of cardiac conditions commonly seen in primary care.
Aggregated Diagnostic Performance
from Clinical Validation
The aggregated results from the clinical validation of Cardio-HART show consistent diagnostic performance across major cardiac disease families.
| Disease family | Sensitivity | Specificity |
|---|---|---|
| Heart failure (all phenotypes) | 84 % | 90 % |
| Structural heart disease | 85 % | 89 % |
| Functional heart disease | 91 % | 91 % |
| Valvular heart disease | 84 % | 90 % |
Scope covered: 14 “HART-findings” mirroring standard echo criteria; ~93 % of cardiac conditions commonly seen in primary care.
Benchmark vs Leading AI-ECG
In head-to-head studies, Cardio-HART consistently outperforms single-signal AI-ECG approaches. The system detects a wider range of conditions with higher sensitivity and specificity, as confirmed through clinical validation.
| Capability | Cardio-HART | Mayo Clinic AI-ECG |
|---|---|---|
| Data inputs | ECG + PCG + MCG | ECG only |
| Conditions detected | HFpEF, HFmrEF, HFrEF, valve, rhythm | HFrEF only |
| LVEF model AUROC | 0.94 | 0.93 |
| Sens / Spec (LVEF) | 91.2 % / 88.6 % | 86 % / 78 % |
| Output | Structured multi-finding report | Binary flag |
| Point-of-care ready | Yes (nurse-led) | Research setting |
Peer-Reviewed & Independent Studies (chronological)
- Calcagno et al., BMJ Open Heart 2022 – biosignal pathway vs echo triage, n = 3 842; validated HF and valve models. (Open Heart)
- European Heart Journal – Digital Health 2024 – 14-centre prospective validation, n = 6 127; κ 0.78 agreement, AUROC 0.94. (Oxford Academic)
- Cardio-Phoenix multi-centre post-market study 2023 – 15 000+ patients; confirms 94 % drop in uncertain ECGs and trebling of GP diagnostic yield. (cardiophoenix.com)
- Journal of Practice of Cardiovascular Sciences 2024 – AI-echo review cites Cardio-HART as exemplar of non-imaging echo-equivalent diagnostics. (Lippincott Journals)
- Community screening “Diagnósticos F.C.”, Spain 2025 – nurse-led, physician-interpreted, 10-min workflow; detected three asymptomatic HF cases. (cardiophoenix.com)
Full DOIs and PDFs are included in the downloadable Evidence Dossier.
Evaluation Methodology (condensed)
- Training set: 7 000 biosignal + echo pairs (echo within 10 days; 90 % same-day).
- Validation set: 8 100 independent cases from 18 sites (EU & UK).
- Ground truth: Consensus echo (Simpson biplane EF, valve grading, structural dimensions).
- Three-step hybrid classifier: (1) maximise sensitivity, (2) specificity filter, (3) balanced severity output.
- Performance thresholds: ≥ 80 % Se/PPV for balanced (≈50 %) conditions; ≥ 70 % Se and ≥ 94 % Sp for low-prevalence (< 20 %) endpoints.
Regulatory & Quality Credentials
- CE-marked (MDR Class IIa) & FDA 510(k) (K182970).
- ISO 13485-certified QMS; GDPR / HIPAA compliant cloud.
- NHS Digital Technology Assessment Criteria (DTAC) green pass – June 2025. (cardiophoenix.com)
- Annual NB audits (SIQ Slovenia), Cyber Essentials certified.
Transparency & Ongoing Real-world Evidence
- Live NHS pilots (primary-care HF pathway) running since 2024—interim data show 35-40 % fewer unnecessary echo referrals (manuscript in preparation).
- Community and sports-medicine screenings continue across ES, UK and IT, demonstrating 10-minute nurse-led workflows and early HF detection (see Spain 2025 event). (cardiophoenix.com)
Key Take-aways for Commissioners & Clinicians
- Echo-level insight in minutes, at ECG cost.
- > 90 % sensitivity for HF across the EF spectrum—crucial for HFpEF where ECG and BNP underperform.
- High rule-out safety (NPV > 97 %): confidently avoid echo in low-risk patients.
- Nurse-led, repeatable, low-operator-dependence—ideal for PCNs and community diagnostic hubs.
- This is supported by strong clinical validation of Cardio-HART across 15,000+ patients, confirming reliability for real-world healthcare systems.
