Hypertension Tele-monitoring Study
July 28, 2025
CVD GUIDELINES DISSEMINATION, HYPERTENSION AND DIABETES DATA MANAGEMENT- WORLD HEART FEDERATION (WHF) FUNDED PROJECT
July 31, 2025Project Description
Detection of LVSD during the early and asymptomatic stages remains a major challenge. Whereas echocardiography can confirm the diagnosis, this imaging modality is not readily available in many resource-limited settings. Use of more readily available modalities to identify patients at high risk for LVSD can help in fast-tracking its confirmation with timely initiation of appropriate treatment.
Electrocardiography (ECG) based Artificial Intelligence (AI) algorithms have been validated to identify patients at high risk for LVSD in many populations tested. No such validation has ever been conducted.
Objective of this study was to evaluate the effectiveness of an ECG-based AI algorithm in predicting Heart Failure with Reduced Ejection Fraction (HFrEF) in a Kenyan population mainly in public hospitals.
The study aimed to determine the feasibility of using ECG-based AI for early identification of patients at high risk for LVSD, potentially improving diagnosis and treatment in resource-limited settings.
The primary Results of the study were:
1. Diagnostic Performance
- a) High Accuracy:
- -> AI algorithm demonstrated sensitivity of 88% (95% CI: 82–93%) and specificity of 91% (95% CI: 87–94%) for detecting LVSD (LVEF ≤40%) compared to echocardiography (gold standard).
- -> AUC of 0.94 (excellent discriminative ability).
- -> Superior to Traditional ECG: Outperformed manual ECG interpretation
b) Clinical Utility
- -> Early Detection: Identified 72% of asymptomatic LVSD cases missed by routine clinical evaluation.
- -> Resource Optimization: Reduced unnecessary echocardiogram referrals by 35% by prioritizing high-risk patients.
2. Population-Specific Findings
- a) Challenges with Comorbidities:
- -> Performance slightly lower in patients with hypertensive heart disease (sensitivity 83%) due to overlapping ECG patterns.
- -> Maintained robustness in HIV-positive cohorts (sensitivity 85%), a key subgroup in Kenya.
- b) False Positives: Most occurred in patients with right ventricular hypertrophy (common in pulmonary hypertension endemic regions).
3. Implementation Outcomes
- a) Feasibility:
- -> AI interpretation time: <30 seconds vs. 5–7 days for echocardiography wait times in public hospitals.
- -> Nurses/CHWs successfully used the tool with 2-hour training.
