Article Impact Level: HIGH Data Quality: STRONG Summary of BMJ Open, 15(5), e098030. https://doi.org/10.1136/bmjopen-2024-098030 Dr. Mihir A Kelshiker et al.
Points
- The TRICORDER trial tested an AI-enabled stethoscope across 200 UK primary care practices to assess its ability to detect several serious yet often hidden heart conditions.
- Patients examined with the device in a primary care setting were 2.33 times more likely to be diagnosed with heart failure, facilitating earlier intervention before emergency hospitalisation occurs.
- The technology also significantly improved detection rates for other conditions, making patients 3.45 times more likely to be diagnosed with atrial fibrillation and 1.92 times more likely with valvular disease.
- Despite its diagnostic potential, the study revealed significant implementation challenges, with 70% of participating practices discontinuing or infrequently using the device after a period of twelve months.
- While promising, two-thirds of patients flagged for suspected heart failure by the AI did not have the condition confirmed, highlighting the need for confirmatory downstream diagnostic testing.
Summary
The TRICORDER study was a pragmatic, two-arm, multi-centre, cluster-randomised controlled trial investigating the clinical effectiveness of an artificial intelligence-enabled stethoscope in primary care. Up to 200 primary care practices in the UK were randomised to either usual care or to have the AI-enabled device available for discretionary use by clinicians. The study’s co-primary endpoints were the difference in coded incidence of heart failure and the ratio of heart failure diagnoses made via hospital admission versus community pathways. The trial included over 1.5 million patients, with 12,725 examined using the intervention device across 96 general practice surgeries.
Results from the trial, presented at the European Society of Cardiology’s annual congress, demonstrated a significant increase in the detection of key cardiovascular conditions over 12 months. Patients examined with the AI-enabled stethoscope were 2.33 times more likely to be diagnosed with heart failure compared to the usual care cohort. The technology also increased the likelihood of diagnosing atrial fibrillation by 3.45 times and valvular heart disease by 1.92 times. The device combines a 15-second ECG with sound analysis, sending data to a cloud-based AI for interpretation.
Despite the positive diagnostic yield, the study highlighted significant implementation challenges. After 12 months, 70% of the GP surgeries in the intervention arm had either stopped using the device or were using it infrequently, suggesting that integration into existing clinical workflows requires further attention. Furthermore, the screening tool produced a notable number of false positives for heart failure; two-thirds of patients flagged by the AI did not have the condition confirmed upon subsequent BNP blood testing or cardiac imaging. Researchers recommend its use for symptomatic patients rather than for routine screening.
Link to the article: https://bmjopen.bmj.com/content/15/5/e098030
References Kelshiker, M. A., Bächtiger, P., Mansell, J., Kramer, D. B., Nakhare, S., Almonte, M. T., Alrumayh, A., Petri, C. F., Peters, A., Costelloe, C., Falaschetti, E., Barton, C., Al-Lamee, R., Majeed, A., Plymen, C. M., & Peters, N. S. (2025). Triple cardiovascular disease detection with an artificial intelligence-enabled stethoscope (Tricorder): Design and rationale for a decentralised, real-world cluster-randomised controlled trial and implementation study. BMJ Open, 15(5), e098030. https://doi.org/10.1136/bmjopen-2024-098030
