Development and implementation of a pediatric AI multi-modal digital stethoscope and respiratory surveillance system in South Africa

About This Grant

PROJECT SUMMARY: This is a submission for a National Institutes of Health R21/R33 award aiming to develop and validate an innovative, inexpensive multi-modal digital stethoscope capable of simultaneous automated lung sound and respiratory rate analysis, and mobile phone interoperability with an electronic health (eHealth) DHIS2 platform for real-time LRI and antibiotic surveillance in LMICs (R21 period), and then evaluation of the multi-modal digital stethoscope and eHealth system using an individually randomized controlled trial (R33 period). The lower respiratory infection (LRI) pneumonia is the leading infectious cause of death among children globally and in sub-Saharan Africa, but appropriate, accurate and user-friendly LRI diagnostics suitable for low-income and middle-income countries (LMICs) are severely lacking. This results in misdiagnosis and death, as well as inappropriate antibiotic use that contribute to overall rising antibiotic resistance rates globally. We aim to develop, validate, and rigorously evaluate the implementation of an innovative, inexpensive multi- modal digital stethoscope capable of simultaneous automated lung sound and respiratory rate analysis, and mobile phone interoperability with an electronic health (eHealth) DHIS2 platform for real-time LRI and antibiotic surveillance in LMICs. We hypothesize that a multi-modal digital stethoscope will meet pre-specified performance benchmarks and its implementation within an eHealth platform will address key LRI diagnosis implementation outcomes. To accomplish our overall project goal we will build upon nearly a decade of our multi-disciplinary digital stethoscope research and development across seven LMICs, our previous successful LMIC pulse oximeter device development using a human centered design (HCD) strategy, and evidence from our recent randomized controlled trial of antibiotic treatment in Malawi that suggested the vast majority of non-severe LRIs did not require antibiotics at all. During the two-year R21 period we will apply the human centered design approach to develop a multi-modal digital stethoscope and eHealth system (Aim 1). The device will be validated in a sound laboratory and field tested in various clinical settings with a range of practitioners in Cape Town, South Africa. We will then implement the multimodal digital stethoscope and DHIS2 system at a variety of facilities (Khayelitsha Hospital, and two primary care clinics) that serve a large township in Cape Town, South Africa and conduct an individual randomized controlled trial to evaluate `correct LRI diagnosis' as the primary outcome (Aim 2), as well as secondary implementation outcomes. Successful completion of the research described in this application could lead to broader scale-up of an affordable, sustainable interoperable LRI and antibiotic use eHealth system with advanced diagnostic technology contextually appropriate for LMICs, where most pediatric deaths from LRIs occur.