On Wednesday, February 28, 2024, the Bay Area Global Health Alliance hosted AI Demystified: Practical Insights and Real-World Applications, the first virtual convening in its AI (artificial intelligence) and global health discussion series. The discussion explored how Alliance members PATH, Pendulum, and Viamo overcame challenges and leveraged AI’s transformative power to increase their global health impact.
Kicking off the discussion, moderator Natasha Sunderji, Alliance board member and Global Health and Nutrition Lead at Accenture Development Partnerships engaged in a fireside chat with William Weeks, Director of AI for Health at Microsoft, regarding the growth of AI powered solutions in the healthcare sector.
__________________________________________________________________________________________________
Key takeaways:
1. AI’s Potential in Workforce Shortages and Patient Decision Support (Microsoft):
- Large language models are seen as playing a key role in patient decision support, providing valuable information for healthcare decisions, especially in low-resource settings.
- AI has the potential to address healthcare workforce shortages, especially in low-resource settings.
- AI can enhance diagnostics, such as AI-assisted diagnostics for eardrums and retinopathy.
- Improvements in clinician encounters, back-office operations, and diagnostics, including voice analysis for mental health, can be achieved with AI.
2. Data-centric Approach and Challenges in Global Health Datasets (Pendulum):
- Pendulum focuses on supply chain optimization and patient risk prediction using AI.
- Applied AI solutions in Côte d’Ivoire for facility-level forecasting, leading to integration into health facility ordering systems.
- Despite challenges with imperfect global health datasets, Pendulum demonstrated success in developing accurate forecasts, even with 80% missing values.
3. Challenges for Communities without Internet Access (Viamo):
- Viamo serves as a platform accessible through basic mobile phones, bringing AI solutions to communities without internet access.
- Successfully used generative AI (ChatGPT) in a pilot called “Ask Viamo Anything” to enable users to ask questions in their own voice, contributing to higher engagement, especially from women aged 18 to 24.
4. Language Technology and Bridging the Gender Digital Divide (Viamo):
- Emphasis on the importance of language technology and the need for multiple language support to make AI more accessible in low-resource settings.
- The introduction of AI, especially in multiple languages, can potentially bridge the gender digital divide in low-resource communities.
5. Open Access Data and Responsible AI Use (PATH):
- PATH highlighted challenges in developing quality and accessible health tools, particularly in low-income countries, due to the lack of data.
- Open access data is essential for de-risking and facilitating research and development (R&D) for health technologies.
- Stressed responsible and ethical data use, drawing on experiences from biospecimen repositories.
- Shared the TIMCI data repository project, focusing on advancing multimodal pulse oximeters and the importance of data repositories in facilitating early R&D.
6. Community Involvement and Responsible AI Practices:
- Emphasized the importance of involving local communities in AI development.
- Responsible AI practices, as emphasized by Bill Weeks, play a vital role in the successful implementation of AI solutions in healthcare.
- Collaboration with local communities is essential for successful AI implementation.
Focused on the importance of data security and compliance with standards like NIST, as well as regulations around protecting personal identifiable information in the context of mobile phone technology.
7. Financial Sustainability and Funding Models:
- Discussed the needs for financial sustainability of innovations in the global health sector.
- Companies like Viamo and Pendulum rely on a combination of grants and venture funding, emphasizing the need for a robust and sustainable financial approach in healthcare technology.
8. Importance of Partnerships and Collaboration:
- Collaboration with various groups and projects is crucial for addressing underserved areas in health technology.
- Partnerships and broad collaborations are emphasized for advancing innovations and overcoming challenges in global health.
__________________________________________________________________________________________________
In the opening discussion, Sunderji and Weeks explored the impact of AI on global health. Weeks emphasized AI’s potential in addressing healthcare workforce shortages, especially in low-resource settings. He cites examples like AI-assisted diagnostics for eardrums and retinopathy. Bill discusses AI’s role in improving clinician encounters, back-office operations, and diagnostics, including voice analysis for mental health.
Weeks described AI as “an opportunity to massively expand a resource constrained workforce and bring higher level quality of care in low resource settings.” Weeks shared that he considers the advances in AI as “a tremendous opportunity to improve diagnostics;” yet, guardrails are necessary to ensure that appropriate solutions are being developed. Weeks also stressed the importance of involving local communities in AI development and highlighted the need for responsible AI practices.
“What are some of the innovative applications of AI that you anticipate will shape the future of global health?” Sunderji asked Weeks before concluding the discussion. “In low- and middle-income countries, I anticipate that large language models (LLMs) will help improve patient decision support, particularly when patients are deciding whether to pursue further care,” responded Weeks.
The remainder of the discussion allowed the four presenting Alliance member panelists the opportunity to elaborate on the respective challenges and successes they’ve experienced integrating AI into global health systems.
Use Case Challenge #1: Working with imperfect “real world” global health data sets to still develop accurate predictions and recommendations
Brittany Hume Charm, Head of Growth and Global Health at Pendulum (formerly Macro-Eyes) touched on some of the challenges Pendulum has experienced when applying AI to “real world” global health datasets to develop accurate predictions and recommendations.
“Data isn’t perfect,” said Hume Charm, while elaborating on the processes deployed to ensure data accuracy. In the past “we’ve seen three different datasets say wildly different things,” added Hume Charm. “We use our data analytics to clean and correct data, and then apply AI to predict what’s going on in those datasets,” explained Hume Charm.
Specifically, they faced datasets with large discrepancies, different amounts of data, and 80% of the values missing. Despite facing challenges where 80% of the values were missing in the data, Pendulum demonstrated the capability to develop forecasts that were 38% more accurate than the three-month rolling average. This is unexpected, as missing data often poses a significant hurdle in the development of accurate and reliable predictive models. Pendulum’s success in handling such incomplete datasets showcases the effectiveness of their data-centric approach and the power of AI and machine learning in overcoming real-world challenges in global health.
A case study from Côte d’Ivoire demonstrated how Pendulum successfully applied AI solutions to strengthen facility-level forecasting of family planning products in a challenging context, where data was incomplete and had significant gaps.The success led to the integration of forecasts into health facility ordering systems, with plans to scale and expand into malaria forecasting. This showcases the potential of AI to provide accurate predictions and recommendations even when faced with substantial data limitations.
Use Case Challenge #2: Is it possible to bring generative AI solutions to communities without the internet?
To follow, Sulakshana Gupta, Viamo’s Vice President of Partnerships in East and Southern Africa and Priscilla Obeng, Senior Program Manager of Special Programs at Viamo illustrated how Viamo brings generative AI solutions to communities without access to the internet. They shared their success in using generative AI to enable users to ask questions in their own voice.
“At Viamo, we think of ourselves as the Google for the billions that still have just a basic mobile phone,”said Gupta, Viamo’s Vice President of Partnerships in East and Southern Africa. “The Viamo Platform uses AI technology to provide information on beneficial health products and services available to those in low- and middle-income countries.”
The Viamo platform initially allowed users to access pre-recorded audio information through a phone call. However, with the introduction of generative AI (specifically, ChatGPT) in a pilot called “Ask Viamo Anything,” users can now ask questions in their own voice and receive detailed AI-generated answers. The pilot has seen significant engagement, particularly from women aged 18 to 24.
“We might not only be making such technology available but also bridging the gender digital divide, which seems to be prevalent in such communities,” said Obeng.
While the Viamo AI Platform has provided critical information to “over 3 billion users without access to the internet,” the technology still experiences language limitations, noted Gupta and Obeng. “The biggest learning from our experiments is that we need to improve or develop language technology so that we can also leverage on existing tools,” said Obeng.
They called for the development of existing tools in multiple languages to make AI more accessible and equitable in low-resource settings. They emphasized the need for localizing responses to make them more actionable and useful.
Use Case Challenge #3: Need for Open Access Data to Derisk R&D for Needed Health Technologies
Helen Storey, Public Health Scientist at PATH, discussed PATH’s role in using open data to de-risk research and development for needed health technologies. Storey noted that “almost 80% of the population in low- and middle-income countries lacks access to appropriate diagnostics.”
“In primary care settings, access to objective health information is critical to guide scarce resources to where they are needed,” added Story. “PATH’s focus here is to produce data that would enable AI developers to build models that measure basic health measurements for those in low- and middle-income countries,” explained Storey. “As we talk about AI tools, it’s no different than biospecimens, but now it’s data that is really required for that early R&D and that early validation of new technologies” and often that data is lacking or inaccessible.”
Storey presented insights from the Tools for Integrated Management of Childhood Illnesses (Tim C) project, focusing on advancing multimodal pulse oximeters. Helen outlined the evolution of medical devices with AI-enabled functionality and emphasized the significance of data repositories in facilitating early R&D. The Tim C project is a data repository for advancing innovation in children’s health and low-resource settings, emphasizing responsible and ethical use of data.
“Data safety, data privacy, and ethical and responsible use of AI is central to the work that we’re doing,” concluded Storey, drawing on experiences from biospecimen repositories.
Stay tuned for the second convening in the discussion series. Please provide your input on our short survey, which will help us shape future AI convenings to make the series as valuable as possible for participants.
About the Alliance’s AI and Global Health Series:
Following the Alliance’s panel on AI during its 2023 Annual Meeting, and the Alliance’s participation in the White House Office of Science and Technology Policy (OSTP) Roundtable on AI and Global Health, members requested further convenings to understand opportunities and implications for their work.
The quarterly series will feature experts, practitioners, and stakeholders from diverse sectors to discuss applications of AI in areas such as healthcare delivery, disease surveillance, diagnostics, and treatment. This interactive series will showcase members who have started integrating AI into their programs and explore how best to ensure the responsible and ethical implementation of AI in their initiatives. It hopes to provide paths forward for those new to AI and those who are well-versed in the emerging technologies.
Leveraging AI technologies offers innovative solutions that can positively impact healthcare systems worldwide. In the context of global health, shifts in healthcare within developed markets offer the promise of significant improvements for underserved communities in global health. The impact spans from empowering community-health workers to enhance patient care in remote regions to assisting low and middle-income countries (LMICs) in proactively averting deadly disease outbreaks. There is a growing acknowledgment of the vast potential of AI tools to disrupt traditional tradeoffs in health access, quality, and cost. Health systems in LMICs grapple with significant challenges such as severe shortages of personnel, medical equipment, and other resources, demanding strategic and innovative solutions. AI tools present an exciting opportunity not only to optimize existing resources and address workforce shortages but also to significantly enhance healthcare delivery and outcomes in low-resource settings.
About Our Speakers for AI Demystified:
Natasha Sunderji (moderator) serves on the Alliance board of directors and is the Global Health and Nutrition Lead for Accenture’s Accenture Development Partnerships. She also co-leads Accenture’s Health Equity Center of Excellence. By engaging Accenture’s global workforce of over 700,000 employees, she works to address the world’s social, economic and environmental issues. Sunderji is a visionary healthcare leader with over 18 years of experience advising multinational companies, nonprofits, foundations, and multilateral agencies on growth strategy, innovative business models, digital health, and cross-sector partnerships. She is a vocal advocate for vulnerable and underserved communities. Sunderji strives to drive care delivery transformation for healthcare organizations through financially sustainable, consumer centric, and data driven business models. She has worked with leading digital health platforms to design patient-centric solutions, supported over 30 inclusive business models across low and middle-income countries, and advised policymakers on the regulations and investments needed to create robust digital health ecosystems. She was named a Top 50 Women Leader in San Francisco in 2023, and a Top 50 Innovator in 2020 by the World Summit AI community.
William Weeks is Director of AI for Health at Microsoft. In concert with not-for-profit organizations and academic centers, he conducts research designed to improve health and health outcomes, applying AI in three areas: imaging, public health, and large language modeling. Weeks also serves as the Medical Director for Bing, Microsoft’s search engine. Weeks has published extensively on the economic and business aspects of healthcare delivery science, physicians’ return on educational investment, social determinants of health, population health, and healthcare value. Weeks has received multiple national awards for his research efforts.
Brittany Hume Charm is Head of Growth, Global Health at Pendulum. Brittany works with mission-driven entrepreneurial teams to scale promising innovations in health access and equity. Prior to Pendulum, Hume Charm led global health growth and partnerships at Zipline. Hume Charm has also worked at the intersection of health delivery innovation, business, and entrepreneurship with the Skoll Foundation, Johnson & Johnson, and McKinsey & Company. Hume Charm serves on the board of Medic, a nonprofit tech startup that builds digital health tools for community health workers, and was a member of the inaugural cohort of WomenLift Health, a leadership initiative for women in global health seeded by the Gates Foundation. She has a BA in public policy from Princeton and an MBA from UC Berkeley-Haas.
Sulakshana Gupta is Viamo’s Vice President for Partnerships in East and Southern Africa, with over 14 years of experience with nonprofits and private sector organizations. Prior to Viamo. Sulakshana’s journalism and media career saw her work with BBC Media Action in Sierra Leone and South Sudan and write for publications such as TIME magazine and the BBC Focus on Africa magazine
Priscilla Obeng is Viamo’s Senior Program Manager of Special Programs, specializing in unconventional programs. With a focus on transformative mobile technology solutions, she leads the design and execution of initiatives that empower communities in low-resource settings. Priscilla holds a PMP certification, bringing over six years of experience, and holds a BSc in Agribusiness Management from the Kwame Nkrumah University of Science and Technology in Ghana.
Helen Storey is a Public Health Scientist at PATH, where for the last 10 years she has advanced the development and introduction of intentionally designed health technologies for low-resource settings through product-driven human-centered research focused on the needs of end users and health system integration. Storey works on all phases of the product development lifecycle, supporting early understanding of user and product specifications, iterative product evaluations, clinical validation studies, and operational research on fit and feasibility. Using her epidemiologic and microbiology methods training, she has advanced new tools for many diseases, including soil-transmitted helminths (STHs), onchocerciasis, lymphatic filariasis, typhoid, diabetes, HIV, TB, childhood pneumonia, and Covid19. Collaboration is integral to the success of this work, and one of the best parts of her job are working on multidisciplinary, global teams with partners ranging from academic researchers, for-profit companies, nonprofit organizations, as well as national and regional health programs. Storey earned a BA degree in molecular and cell biology from UC Berkeley, an MS in microbiology from George Washington University, and a PhD in epidemiology from the University of Washington.