Use of AI-enabled stethoscopes in primary care examined

By Melanie Hinze

The findings of two recent studies have indicated that artificial intelligence (AI)-stethoscope implementation in primary care may have the potential to improve detection of cardiovascular disease.

Published in The Lancet, the first study was a cluster randomised trial involving 205 primary care practices in the UK, which services 1.55 million adults. The researchers randomly assigned practices to either the intervention arm (training in use and implementation of AI stethoscopes in routine care) or the control arm (routine care). Ninety-six practices with 701,933 registered patients (average age 44.5 years) were assigned to the intervention arm and 109 practices serving 851,242 registered patients (average age, 43.5 years) to the control arm.

Intention-to-treat analysis found that large-scale implementation of an AI stethoscope in primary care did not significantly increase the detection of heart failure, atrial fibrillation or valvular heart disease (VHD), nor did it significantly shift detection of heart failure towards the community at 12 months. The researchers also found that use declined over time, with clinicians citing workflow barriers to sustained use.

Commenting on the study, Dr Dannii Dougherty, Head of Clinical Evidence at the National Heart Foundation of Australia, explained that these effects were not necessarily because the technology failed, but because many practices used the device infrequently or stopped using it over time.

‘In fact, among individual patients who had at least one AI-stethoscope examination and who were matched to similar control patients, detection of all three conditions was significantly higher,’ she said.

‘The time to diagnosis was also shorter in patients examined with the AI stethoscope,’ she added, suggesting ‘the AI stethoscope and algorithm can support diagnosis in everyday general practice, but only if the device is easy to embed into workflows, integrate into existing systems and used consistently by clinicians.’

To have meaningful impact, Dr Dougherty said AI stethoscopes would need to be easy to use and not increase workload. ‘This could be by integrating [them] into electronic medical record systems and using [them] in a targeted way for higher risk patients.’

The second study, published in European Heart Journal - Digital Health, found that the use of an AI-enabled digital stethoscope more than doubled the identification of moderate-to-severe VHD during routine clinical examinations, compared with a traditional stethoscope.

In this single-arm, single-blinded, prospective study, 357 patients aged 50 years or over were examined with both a traditional stethoscope and an AI-enabled digital stethoscope. Participants (median age, 70 years) were recruited from three primary care settings.

The AI stethoscope showed significantly higher sensitivity in detecting the heart sound patterns indicating VHD, with 92.3% sensitivity compared with 46.2% for the traditional stethoscope.

The researchers noted that the AI-enabled digital stethoscope led to a minor reduction in specificity (86.9% vs 95.6% for the traditional stethoscope), which could potentially increase false positives, but they suggested this risk should be balanced against the value of earlier detection.

Dr Dougherty said this study provided promising evidence that an AI-enabled digital stethoscope could improve detection of moderate-to-severe VHD in older, at-risk adults in primary care. However, she said there were several important caveats: the number of patients with audible VHD was small (13 cases), so the sensitivity and specificity estimates were quite uncertain, and the GPs did not use the AI-enabled digital stethoscope themselves (trained study co-ordinators did), so it was unclear how well it would perform in routine GP consultations.

Dr Dougherty also noted that several of the authors were current or former employees of the company that developed the stethoscope, so independent research was needed.

Lancet 2026; 407: 704-715, and Eur Heart J Digit Health 2026; 7(2): ztag003.