A blueprint for an AI-powered health operating system in urban slums.

Building the backbone of new community-to-system patient pathway for urban slums. Solving delayed diagnosis in primary care with an AI-guided platform enabling community health workers to capture clinically reliable data in the field and triage patients into the best clinical workflow. 

Dr. Elina Naydenova; Biomedical & AI engineer; Ph.D. Machine Learning for Healthcare Innovation Oxford University; CEO & Co-founder Feebris and Global Health Innovation at WHO & Skoll Centre for Entrepreneurship. 

We have a global health workforce crisis – a structural shortage of skilled health professionals to meet the health needs of our populations. This is estimated to reach c.18 million by 2030 and threatens to derail our ambition to achieve universal health coverage as part of the UN SDGs. This shortage has been radically compounded by the pandemic resulting in widespread delayed diagnosis of disease, unnecessary deaths, and, massive costs. 

One of the most promising solutions to support Covid recovery and sustainable provision of universal, high-quality healthcare is in the use of AI solutions that upskill more people to provide health care in the community, more accurately. While the pandemic has acted as an initial accelerant for the application of these tools, there are clear evidence and technology and barriers preventing scale at pace and optimisation of patient and system outcomes.

There is specifically scarce evidence of its application to support the world’s largest and most at-risk group during the epidemic – those living in urban slums. Longer-term, there is a real risk of widening the global health and well-being gap if transformative AI technologies are not validated and generalisable for these groups. 

This solution seeks to improve the capacity and capability of LMIC health systems to respond and recover from Covid 19 and other health threats by establishing a blueprint for the delivery of AI-powered health tools in slum environments.  

This solution will first focus on expanding Feebris' delivery across urban India – home to the world’s largest urban slums and the third-highest number of confirmed cases and deaths from covid-19. Our solution focuses on the most clinically vulnerable patients - young children and elderly people with complex co-morbidities and with increased risk of severe Covid-19 (and other infectious diseases). The second wave will replicate this model in an East African cluster (Kenya, Uganda) and across South Africa. 

We will support these patients by providing AI-powered technology, training, and change management to front-line community health workers, health centres, and hospitals.  Feebris has a clear methodology to engage patients, users, and relevant health stakeholders including (1) solution co-creation - substantial patient/user consultation supporting product design and (2) frequent user/clinical testing – for each product release with technology calibrated and validated against diverse patient/user groups. 

Feebris is committed to working with local partners to grow local capacity for innovation. We want to build the standards of how to effectively use technology to build a backbone between community and health system. We want this evidence to be freely available to all healthcare providers, health ministers, and relevant health stakeholders to inform their plans for primary care transformation with technology. Through a series of white papers and peer-reviewed publications, we will generate unique knowledge on the most appropriate ways of evaluating cross-geography generalisability of AI solutions and tools for quantifying and minimising bias. 

We have a strong commitment to ensuring any insights we generate feed into channels for continuous development of the care and healthcare workforce. Specifically, we develop training tools that provide medical students free access to data we have captured for the purposes of accelerating exposure to complex clinical cases and practical learning. We will scale these efforts, ensuring more health workers, clinicians, and medical students have access to tools for continuous upskilling, such as an open-source training suite for respiratory rate capture in global health. 

We are democratising access to high-quality primary care diagnosis for all. Our deployment in India has been evaluated against a health-economic framework evidencing improved care quality, workforce development, and cost-savings ready for scaling to clinically vulnerable groups: 

Inputs: User training, technology optimisation, establish diagnostic and referral pathways.  

• Indicators: Inter-user variability in data quality <2% after training, variety of health outcomes (clinically confirmed), gold-standard diagnostic data available for >90% cases, hospital/community clinic referral pathway 100% cases 

Outputs: Consistency & effectiveness of examination, Accuracy of triage, Access to diagnostic tests & hospital treatment. 

• Indicators: Data acquisition <10min, missing data <1%, 95% data acquisition completion, >95% triage accuracy (acute), 100% 'at risk' cases reviewed by doctor & referred as required, 100% cases requiring treatment referred to hospital. 

Outcomes:  Early disease identification, accelerated treatment access, hospital trips reduced (non-severe), hospital access increased (severe) 

• Indicators: >95% acute cases identified while non-severe & treated locally, treatment in <24 hours (severe) & <48 hours (non-severe),40% reduction hospital trips (non-severe), referral within 12 hours (severe) 

Impact:  Fewer deaths from acute disease, fewer severe cases of respiratory infection, direct savings 

• Indicators: 60% reduction in childhood mortality, 85% reduction in severe respiratory disease, 10% monthly household savings 

Our vision is a world where no one suffers from treatable diseases simply because they can’t access a doctor.  

Globally, 15m lives and $1.7tr are lost due to poor or delayed clinical decision-making in primary care, simply because there aren’t enough clinicians. We build technology that helps health systems scale, shift early detection of disease to a community workforce and focus clinical time on severe disease. We have a particular focus on the most vulnerable patients, who struggle to access healthcare but for whom early detection is critical to avoid complications.  

Over the next year, we expect to reach more than 100,000 patients, and over the next three years project growing this to 2 million. Through our current work, we have evidenced 8x ROI, reducing the severity of disease and saving healthcare expenditure. As we continue to build up more sophisticated AI-powered functionality, we expect to grow our ROI to 10x in the next year and 25x in the next 3 years. Throughout this period, we will work with public health organisations, recovering from the burden of the pandemic to deliver more effective models of care delivery. 

In line with our health-economic framework evidencing the improved quality of care, workforce development, and cost-savings across Indian slums, our scaled model will measure progress against:

Quality of care - enabling early detection of deterioration in the community, reducing complications and hospitalisations, enabling more personalised and proactive management of conditions, reducing risk of misdiagnosis/delayed diagnosis. 

• Pilot Performance: treatment time reduction (15 - 4 days), reduction in the severity of Pneumonia (15% - 2% of all cases)

Workforce development - upskilling workforce, raising profile of community workforce and improving retention, reducing stress in primary care. 

• Pilot Performance: 100% of users (high-school educated, no medical training) were able to deliver doctor-quality measurements with the help of decision-support after 1 hour of training. 

Utilisation of clinical resource & savings - enabling doctors to prioritise most critical care; reduce avoidable conveyances through early detection/resolutions. 

• Pilot Performance: 80% reduction in hospital trips, 50% reduction in hospitalisations and cost of care reduced by 730% over a period of 6 months, including eliminating out-of-pocket healthcare expenditure equivalent to 10% of household income. 

We have a global vision to support healthcare systems to enable a non-clinical workforce to capture clinically reliable data in the field and triage patients into the best clinical workflow. One key barrier is achieving commercial sustainability and scalability in LMICs. This is primarily a challenge of healthcare system architecture, requiring transformation in budgets and operating models. We plan to overcome this barrier with defensible evidence and health economics that strengthen the ROI case for scale. We have already demonstrated a positive impact in a 6000-patient programme in Mumbai and now we will evidence that this model generalises to multiple settings. 

We also face a significantt regulatory barrier. First in establishing systems that meet stringent regulatory standards (medical device, data protection, information security), and culturally, hiring software engineers that can work in very different tech start-up environments. We are overcoming these regulatory barriers by working with specialist consultants who provide both engineers and the guidance we need to build effective and scalable systems. Our medical device is currently CE marked and over the next 12 months, we aim to achieve FDA clearance; meeting the two most internationally recognised medical device regulatory standards will streamline entry into other markets. 

NHS - NHSX, NHS Innovation Accelerator, local Clinical Commissioning Groups

UK Department for International Trade 

Innovate UK 

Google for Start-ups 

Children’s Prize 

Digital Health, London 

Social Tech Trust 

Apnalaya 

Malaria Consortium 

Feebris share the ambition to improve global health and well-being through innovation. We want to see a world where no one suffers from treatable conditions simply because they cannot access a doctor and have a global vision to support healthcare systems to enable a non-clinical workforce to capture clinically reliable data in the field and triage patients into the best clinical workflow. A well-evidence barrier, however, is achieving commercial sustainability and scalability in a LMIC context. This is primarily a challenge of healthcare system architecture, requiring transformation in budgets and operating models.  

A partnership with the Trinity Challenge and Member organisations provides us with the initial expansion funding to generate robust evidence (clinical and economic) to inform a sustainable delivery model with local health providers – this objective will be significantly supported by the access, visibility and amplification the challenge and member organisations can provide to accelerate our product-market fit and further leveraged funding. 

Trinity Challenge Members 

• JPAL – Monitoring & Evaluation partner (MIT off-shoot) to evidence the impact of AI-powered urban slum health interventions via a robust RCT.
• LSE – current partnership in development focused on publishing findings from India
• Google - current partnership in the UK, looking to expand support to India 

Other implementing Partners 

We want to partner with organisations that are deeply embedded and can drive local operations, with us acting as the technology backbone. These include:

• Ilara Health
• Save the Children
• Partners in Health
• Last Mile Health
• Local device manufacturers.