Alliance News

Empowering Frontline Health Workers with AI: Navigating Pitfalls and Possibilities

Jun 28, 2024

2024 Discussion Series on AI and Global Health Convening #3

 

WATCH THE RECORDING

 

Can AI tools redefine how we care for the world’s most vulnerable? Scarce resources and significant challenges impede the ability of frontline healthcare workers (HCWs) to provide healthcare to those in their communities and yet they are critical to reaching those most in need. 

On June 25, the Bay Area Global Health Alliance hosted its third virtual convening on AI and Global Health, bringing together leaders from diverse sectors to explore how artificial intelligence has the potential to support these essential workers and facilitate care at the last mile. 

“Imagine how exciting it would be to be able to put in the hands of frontline workers, the full knowledge base of the health system behind them, and to empower those folks who are closest to the patient with the right information, the right algorithms to support clinical decision making,” said moderator and Alliance board chair Colin Boyle, setting the stage for the conversation. 

The discussions delved into various aspects of AI integration for HCWs, recognizing the nascent nature of these interventions. One recurring theme was the critical importance of human-centered design in developing AI tools that are both effective and acceptable to frontline workers.

Elina Urli Hodges, Assistant Director of Programs at the Duke Global Health Innovation Center and Innovations in Healthcare, presented their research funded by the Bayer Foundation, on AI applications for community health workers (CHWs) in low- and middle-income countries (LMICs) in which they identified six enablers for more effective implementation of AI with CHWs. The findings echoed learnings from the application of broader digital health and intervention tools and underscored the necessity of involving CHWs in the design process to ensure that AI tools are user-friendly and contextually relevant. 

Noting that community health workers want to be professionalized, skilled, trained and ready to act for their community, Hodges emphasized, “It’s just really important that we involve the community health workers and those frontline health workers into the design of these applications, making sure that the applications fit their workflows and are making their jobs easier.” 

Andrew Ddembe, CEO and founder of MobiKlinic in Uganda, and Raghav Minocha, Head of Partnerships at Simprints, presented their collaboration on using AI-powered facial biometrics to improve patient identification and streamline immunization processes in Uganda. It showcased the tangible benefits of AI in enhancing healthcare delivery.

“There are almost a billion people on this planet who do not have any form of formal ID,” Minocha noted. “Without those IDs, especially in rural and remote areas in LMICs, there’s no way for them to be able to access lifesaving health and aid supplies.” Ddembe added that the integration of facial biometrics reduced the time required for patient enrollment and identification, significantly increasing the efficiency of HCWs and allowing them to serve more people in less time. 

Enric Jané, Chief Strategy Officer at Causal Foundry, provided insights into how AI can be scaled sustainably in healthcare. He introduced Causal Foundry’s platform designed to personalize interventions and enhance engagement with digital health tools. “Our goal is to lower the barrier to access all these technologies and basically what we have is a platform that will allow any digital tool to integrate with and leverage a set of AI tools,” Jané explained.

Jané also highlighted the importance of providing AI tools that personalize user experiences, which can be crucial for both health workers and patients. Jané stressed the importance of localizing these tools to fit the specific needs and contexts of different communities.

The convening also touched on the critical issues of data privacy and safety in AI deployment.  Echoing takeaways from the 2023 White House Office of Science and Technology convening and Alliance member input, panelists stressed the importance of adhering to any local data regulations and policies, citing the European General Data Protection Regulation (GDPR) as a good model where local laws are not yet in place. Both Ddembe and Minocha stressed the rigorous measures taken to protect personal data. “All of the data we work with is going to be encrypted in transit and storage..we’re going to follow European GDPR which is the…strongest, most stringent guidelines right now.” With regards to informed consent, Ddembe added, “Consent is very key,” noting that the majority of patients in their project willingly participated after understanding the benefits and security measures in place. 

Boyle noted the “black box” nature of some algorithms, where the decision-making processes are not transparent. This lack of clarity can be particularly concerning when dealing with vulnerable populations. The speakers unanimously agreed on the necessity of maintaining robust data privacy standards and obtaining informed consent from the communities they serve. 

Special thanks to Pfizer, sponsor of the 2024 Bay Area Global Health Alliance AI and Global Health Discussion Series. Stay tuned for our final convening of this series in the fall.

 

________________________________________________________________________________________________________________

Key takeaways:

 

  • Nascent Field: Frontline HCWs are often the last cadre of health workers to benefit from digital tools. Examples of AI-powered applications for this community remain in the research and pilot stage, with some evidence of the impact of these interventions but to date, limited if any, research on tracking health outcomes.
  • Human-Centered Design: Involve community health workers in the design of AI applications, ensuring tools fit their workflows and make their jobs easier. This includes considering appropriate language and integrating the tools within their day-to-day activities.
  • Local Context Adaptation: Design AI platforms for the local environment, considering factors like battery life, offline functionality, and limited storage in low-resource settings – emphasizing the importance of training AI on high-quality local data and conducting usability testing in the field. 
  • Interoperability and Local Partnerships: Develop AI solutions that are interoperable with existing tools and infrastructure. Stressed the value of strong local partnerships, particularly with organizations that have established relationships with community health workers and government authorities.  
  • Data Privacy and Security:  Ensure data privacy and security, including encrypting data, educating community health workers on securing their devices, adhering to local data regulations and policies and following European GDPR where local laws are not yet in place. Take into consideration that CHWs may be learning how to use these platforms for the first time, on smartphones that they may not be used to using. 
  • Personalization of Digital Health Tools:  Use AI to personalize digital health interventions for both CHWs and patients. This personalization seeks to improve user engagement and effectiveness of health interventions. 
  • Efficiency Gains through Biometric Identification Solution:  Consider AI-powered facial biometrics as one potential application to solve the challenge of patient identification in communities where many lack formal IDs. This technology was seen to enable better continuity of care and record-keeping for community health workers, significantly reducing initial patient enrollment from 13 minutes to @6.5 minutes, while subsequent identification time was reduced from over 4 minutes to just 18 seconds on average. (See Use Case #1)
  • Experimentation Platform: Pay attention to platform’s like that of Causal Foundry’s which can support the scaling of AI technologies and are designed to allow for easy integration and testing of AI-driven strategies in digital health tools. The platform enables organizations an efficient means with which to conduct A/B testing, implement interventions, and scale successful strategies, with four exciting areas where AI models can be applied in digital health: time-to-event predictions, recommendations (e.g., for e-learning modules), adaptive interventions (like precision nudging), and resource allocation optimization. It was noted that operational and triage-focused applications are likely to see wider adoption in the near term, while more complex clinical AI applications may take longer to develop and implement due to regulatory and practical constraints​​. (See Use Case #2) 

________________________________________________________________________________________________________________

Featured Use Case #1: MobiKlinic & Simprints

 

Nearly one billion people lack formal identification (ID), resulting in barriers to accessing essential services like healthcare and aid in remote and rural areas. “Our community health workers were working in communities on different patients, children, mothers, and the elderly, but there was a big challenge with identification and continuity of service. It became extremely challenging for the community health workers to keep paper records of these different members in the community,” explained Andrew Ddembe, CEO of MobiKlinic​​.

Raghav Minocha, Senior Manager of Partnerships at Simprints, and Andrew Ddembe shared their collaboration to address the challenge of patient identification in Uganda’s healthcare system. Their partnership integrated Simprints’ AI-powered facial biometrics technology (digital IDs) into MobiKlinic’s existing digital health platform, enabling community health workers to efficiently identify and track patients, thus improving continuity of care and record-keeping in communities where many lack formal identification​​.

The integration of facial biometrics significantly reduced the time required for patient enrollment and identification. On average, the time to enroll a new patient was reduced from about 13 minutes to approximately 6.5 minutes, and the time to identify returning patients was reduced from over 4 minutes to about 18 seconds​​. This improvement allowed community health workers to see more clients in a day, enhancing their productivity in delivering healthcare services​​.

In response to questions about data safety and security of this type of patient identification technology, Minocha emphasized Simprints’ commitment to data privacy and security through practices such as data siloing, encryption, and adherence to stringent data protection guidelines like General Data Protection Regulation (GDPR). Simprints practices data siloing by storing biometric templates separately from personal information, a method that won them the World Bank Group’s Mission Billion Innovation Challenge prize for their approach to protected templates​​. Ddembe highlighted MobiKlinic’s focus on data protection through anonymization and compliance with local data protection regulations in Uganda. He stressed the importance of community engagement, transparency, and obtaining informed consent, noting that when benefits are clearly explained, about 90% of people choose to opt-in for biometric identification​​. Both stressed the importance of ethical practices, transparency, and giving users control over their data.

________________________________________________________________________________________________________________

Featured Use Case #2: Causal Foundry

 

Enric Jané, Chief Strategy Officer at Causal Foundry, highlighted the potential of AI to personalize digital health interventions, addressing a critical gap in current approaches where one-size-fits-all solutions often fall short. He introduced Causal Foundry’s platform, designed to democratize access to AI tools in digital health, enabling organizations to easily integrate and test AI-driven strategies. “Our goal is to lower the barrier to access all these technologies and basically what we have is a platform that will allow any digital tool to integrate with and leverage a set of AI tools,” described Jané​​.

The platform, as Jané explained, focuses on four key areas: time-to-event predictions, personalized recommendations (such as targeted reminders and training modules), adaptive interventions, and resource allocation optimization. By lowering the barriers to AI implementation, Jané argued that healthcare organizations can more easily experiment with and scale effective strategies, emphasizing features like A/B testing for interventions and scalability based on experimental outcomes​​. He emphasized the importance of continuous testing and refinement, advocating for an “experimentation engine” approach that allows for rapid iteration and improvement of AI-driven interventions in real-world healthcare settings​​. Jané also stressed the importance of forming strategic partnerships with organizations capable of addressing clear operational pain points, navigating challenges around local data storage and infrastructure for effective AI deployment in global health contexts​​.

In response to questions about the future of platforms like that offered by Causal Foundry, Jané emphasized that the immediate focus and greatest potential lie in operational improvements and screening/triage applications, rather than direct clinical decision-making. Jané highlighted the challenges of implementing clinical decision support tools, particularly in gaining government approval. He stressed that significant value can be derived from enhancing community health worker performance, improving screening processes, and optimizing resource allocation. Jané suggested that these operational and triage-focused applications are likely to see wider adoption in the near term, while more complex clinical AI applications may take longer to develop and implement due to regulatory and practical constraints​​.

________________________________________________________________________________________________________________

About our Speakers for Empowering Frontline Health Workers with AI:

 

Colin Boyle, Lecturer, UC Berkeley Haas School of Business and Alliance board chair (moderator). Colin is a member of the professional faculty at the Haas School of Business at UC Berkeley where he teaches graduate courses on leadership and strategy for social enterprises and nonprofit organizations. Recently, he was Deputy Director of UCSF’s Institute for Global Health Sciences, dedicated to improving health and reducing inequities worldwide. Colin joined UC in 2012, after 15 years with the Boston Consulting Group (BCG), where he was a partner and managing director leading many of the firm’s social impact projects, helping industry and non-profit innovators develop new products to combat disease and bring them to market for health impact. At UCSF, Colin focused on analyzing the case for investments in health, playing a supporting role in the Lancet Commission on Investing in Health and contributing to other investment cases for specific conditions. He also has supported or led efforts at UCSF related to malaria, TB, neglected infectious diseases, maternal health, health systems strengthening, and regulatory sciences. He also serves on the board of the Oakland Museum of California.

Andrew Ddembe, CEO & Founder, MobiKlinic. Andrew is a health equity advocate, health technology innovator, and CEO and founder of MobiKlinic, a health technology company that strives for improved last mile health delivery and equitable vaccine access. Andrew founded MobiKlinic in 2018. His company has unlocked access to health care and vaccines to hundreds of thousands of people in Buikew region, Eastern Uganda and in Busia, Western Kenya. Andrew has worked as a young expert in the Africa union and European Union youth cooperation hub. He currently represents digital health innovations on the WHO Civil Society and Youth Commission. Andrew has been featured among Forbes Africa top 30 under 30. He is tapped to receive the Global Citizen Prize for healthcare in 2024. Andrew’s vision is to scale MobiKlinic across Africa and enable health equity in Africa.

Elina Urli Hodges, Assistant Director, Programs, Duke Global Health Innovation Center. As an Assistant Director of Programs for the Global Health Innovation Center and Innovations in Healthcare, Elina oversees a portfolio of programs that aim to increase access to health in low- and middle-income countries. She recently co-developed an impact measurement framework for pharmaceutical industry-led access to medicine programs. Elina also directs a project, supported by the Bill & Melinda Gates Foundation, that aims to identify key drivers behind uptake of life-saving global health interventions. Her project experience has taken her to Kazakhstan, China, India, and Kenya. Prior to joining Duke University, Elina worked for more than a decade on non-communicable disease prevention, population health, and global workplace health promotion projects with U.S. government institutions (CDC and Department of Defense), multinational corporations, and health systems. She began her career in health working in the national advocacy office of the American Heart Association in Washington, DC. Elina received a Bachelor of Arts degree in Foreign Affairs and French from the University of Virginia and a Master of Science in Public Health with a concentration in Health Policy and Management from the University of North Carolina at Chapel Hill.

Enric Jané, Chief Strategy Officer, Causal Foundry. As the Chief Strategy Officer at Causal Foundry, Enric leverages his 18+ years of experience in global health, information technology, and data use to provide AI solutions that drive impact in health. He holds a Ph.D. in Physics, an M.D., and an M.S. in Biomedical Engineering, which gave him a strong foundation in business, technical, and research aspects of health initiatives. Enric’s core competencies include strategy development and implementation, stakeholder relations, digital health, and international development. He has led and supported multiple projects with diverse and influential organizations, such as Gates Ventures, the Bill & Melinda Gates Foundation, Vital Wave, Social Impact, and Terre des hommes. His mission is to address complex challenges, support institutional development and change, and create positive social impact through data-driven solutions.

Raghav Minocha, Senior Manager, Partnerships, Simprints. Raghav has spent almost 10 years working on health and humanitarian development projects across the world while living in Africa, Asia, North and Latin America, and Europe. Prior to Simprints, he worked for Egypt’s Ministry of Social Solidarity enabling community health worker-led interventions under the country’s flagship Hayat Karima project. He has also worked in Honduras and India on rural frontline mobile health clinic implementations and technology deployment. At Simprints, Raghav manages international partnerships with governments, fellow iNGOs, corporate partners, and technology companies. Raghav holds a Bachelor of Science from University of California, San Diego.

Daphne Ngunjiri, CEO, Access Afya. Daphne is the CEO of Access Afya, where she spearheads the mission to delivery affordable, high-quality healthcare to underserved communities in Kenya. A seasoned healthcare leader with over 15 years of combined medical and management experience, she holds both an MBChB from the University of Nairobi and an MBA in Healthcare Management and Policy from Strathmore Business School. Daphne believes passionately in the power of healthcare systems, data, and analytics to revolutionize operational efficiency and clinical outcomes.