Dimagi is excited to participate in the upcoming Global Digital Health Forum 2024! As a leader in innovative digital solutions for global health and development, we’re thrilled to showcase our work across eight diverse sessions. Each year, Dimagi team members share key learnings and insights, and this year is no exception. The sessions we’ll be participating in build on three important areas of Dimagi’s work:
- The Frontiers of AI in Global Health
- Improving Frontline Jobs & Advancing Service Delivery
- Technology to Scale Mental Health Care
If you’re attending in person, be sure to visit us at our booth too! Connect with our team, discover tools like Open Chat Studio for building public health chatbots, see how SureAdhere supports patient engagement for complex treatments, and explore the latest advancements in CommCare, our flagship platform for impactful frontline work.
Whether you’re here to learn, network, or innovate, we can’t wait to see you at #GDHF2024!
Find details of each session from Dimagi at the Global Digital Health Forum below:
The Frontiers of AI in Global Health
What does diversity mean for AI in digital health? – Gayatri Jayal
🗓️ 4 December, 2024 | 6:00-6:45pm EAT
Artificial Intelligence is progressing more rapidly than ever before. The emergence of Large Language Models (LLMs) like GPT-4 promises to transform how organizations operate and deliver services. The use of LLMs could greatly enhance organizational efficiency, improve knowledge management, revolutionize support for frontline workers, and provide new ways to serve clients. Our collective challenge is to ensure that the benefits of this new technology are spread equitably, that the considerable risks are addressed responsibly, and that efforts to leverage AI to benefit underserved populations are locally-driven and inclusive. There is a critical need to focus on diversity in three key areas: the creators of chatbots, the users of chatbots, and the voices and feedback from those who interact with the chatbots. 1. Diversity in Who Makes Chatbots: The development of AI chatbots should not be confined to developers or those in HICs. It is essential to include voices from diverse backgrounds, particularly from lower and middle-income countries (LMICs). These regions often have unique challenges and perspectives that can significantly influence how chatbots are designed and implemented. By ensuring diverse representation in the creation process, we can develop AI solutions that are more culturally sensitive and effective in addressing the specific needs of different communities. 2. Diversity in Who Uses Chatbots: AI chatbots must be accessible to a wide range of users, including frontline workers and other individuals in underserved populations. Training and open-source solutions need to be co-designed with and tailored to the needs of these local users. For instance, in Malawi, Dimagi is developing methods to improve GPT-4’s ability to operate in Malawian Chichewa. These efforts ensure that AI technologies are not only accessible but also usable by people who speak lower-resourced languages. 3. Diversity in What People Who Use the Chatbots Say: The inputs and interactions of users should reflect a broad spectrum of experiences and knowledge. This includes mothers, fathers, and frontline workers in LMICs who will be using these chatbots for health-related information and support. Locally-led organizations are best positioned to capture the diverse needs and feedback from their communities. Dimagi will show initial examples of results from pilot testing of chatbots built on our Open Chat Studio platform in various LMIC contexts, highlighting how diverse user feedback has been instrumental in refining chatbot functionality and effectiveness. By embracing diversity in the development, deployment, and utilization of AI chatbots, we can create more inclusive and effective digital health solutions. This approach not only addresses the equity gap but also ensures that the benefits of AI technology are maximized for all, especially the underserved populations in LMICs.
AI and Storytelling: Chatbots for Youth Sexual Health in Senegal and Kenya – Lilianna Bagnoli and Brian DeRenzi
🗓️ 4 December, 2024 | 6:00-8:00pm EAT
In this session, Dimagi will demonstrate two large language model (LLM)-powered chatbots that were co-designed with young adults and experts in Kenya and Senegal to improve knowledge and self-efficacy around sexual and reproductive health (SRH). Both of these chatbots were developed on Open Chat Studio (OCS), an open-source platform for rapidly designing and deploying LLM-powered chatbot experiments.The novelty of these chatbots’ designs is the use of LLMs to emulate existing characters from beloved comic books and TV series to communicate with young adults about SRH. In order to achieve realistic-sounding conversations, Dimagi partnered with established multimedia behavior-change organizations Shujaaz (Kenya) and RAES (Senegal) to design and test relatable chatbot interactions for young users. For example, in Senegal, adolescents will be able to chat with a character named Ange from RAES’ series ‘C’est La Vie’, an educational entertainment series that has existed since 2015 and revolves around a health center in Dakar. Similarly, in Kenya, young adults will be able to interact with a chatbot taking on the identity of DJ B, a central character of Shujaaz’s comic book series of the same name. The chatbots are designed to engage adolescents in multiple interaction types, including Q&A, role-playing, quizzes, and general information sharing all while rapport building. Chatbot conversations can begin with open-ended questions about a young person’s current concerns or questions and eventually move to SRH-related conversations, all while maintaining the voice and tone of the chatbot’s character persona.As part of the solution demonstration, Dimagi will show how the chatbots were developed on OCS using prompt engineering and source materials to support each prompt, enhancing the reliability of the information provided. The presentation will also cover the extensive chatbot evaluations and usability testing conducted during the design of the chatbots, showcasing the effectiveness and user-friendliness of the solution. Lastly, Dimagi will speak about the challenges and limitations of LLMs within the context of live human interaction.
Transforming Child Nutrition Monitoring: A Data-Driven Predictive ML Model for Proactive Interventions – Anna Dixon and Sunaina Walia
🗓️ 4 December, 2024 | 6:00-6:45pm EAT
In this session, we aim to present the development and evaluation of a predictive machine learning (ML) model designed to predict the nutrition status of children monitored by a CommCare project in India. Dimagi, the American India Foundation (AIF), and IDinsight are collaborating on this solution and intervention, which will be implemented in select blocks of Madhya Pradesh and Odisha in India. A supplemental comparative study will also be presented comparing the child nutrition predictive model efficacy to health worker intuition. This work aligns with global commitments to the Sustainable Development Goals, particularly Goal 2: Zero Hunger, addressing child malnutrition through innovative technological solutions. The ML model forecasts if healthy children aged 6 months to 5 years might become malnourished within 60 days including a unified model to denote transitions to stunting, wasting, and underweight states. This model also evaluates the impact of interventions aimed at reducing child malnutrition and follows up with at-risk children’s caregivers to provide focused counselling home visits by Frontline Health Workers (FLW) and to reinforce nutritional importance and identify barriers to better outcomes. Unlike traditional methods that often react to evident malnutrition or previous predictive models that report on the efficacy of predicting malnutrition of all children (inclusive of those already malnourished), our predictive approach allows for proactive intervention, setting a new standard in child malnutrition predictive modelling and preventive healthcare.The model is a random forest classifier trained and evaluated with time-series cross-validation. Feature reduction and hyperparameter fine-tuning were implemented to reduce overfitting and make the model as generalizable as possible, making it potentially applicable to other case studies and geographies. This work sets itself apart by focusing on predicting the transition from healthy to malnourished, a more challenging model objective and potentially more impactful for the program because it enables earlier intervention. This project builds upon the CommCare platform for data collection and the AIF program, showing the potential for scaling up targeted digital health interventions based on reliable, real-world data. The ML model was trained on data collected in the CommCare application SNEH (Skilling, Nutrition Education, and Health) for more than 22,000 children over 18 months. Our analysis showed that the most important model features included historical height and weight measurements, location and family socioeconomic demographics such as caste and parent’s education level. The model’s precision at 0.5 recall is 0.63, which given our class imbalance of .16/.84 (bad malnutrition transition/no transition) amounts to a 3.9x better precision than random selection. An RCT is planned starting August 2024 to evaluate the impact of the ML-driven targeted nutritional counselling home visits on the average child’s likelihood of retaining their nutrition status. The RCT studies 300 total villages (150 villages each in the treatment and control arms) in Madhya Pradesh and Odisha, India. The top 4 children most likely to transition to a worse nutritional status as identified by our model will be selected for intervention for a total of 600 children.
Improving Frontline Jobs & Advancing Service Delivery
Upskilling FLWs on new, impactful interventions through the CommCare Connect platform’s digital learning model – Mercy Simiyu and Themba Nyirenda
🗓️ 5 December, 2024 | 12:25-1:25pm EAT
Training frontline workers (FLWs) is essential for global development. While conventional training methods have achieved a lot, they have limitations. The standard “training of trainers” cascading model is often seen as expensive, variable in quality, and disruptive to FLW service delivery. To address these issues, Dimagi has developed the CommCare Connect (CCC) platform, a digitally-supported model of training with numerous advantages over existing methods. We provide FLWs with self-paced digital courses to learn new interventions, and pay them to deliver these interventions in their communities, with feedback from supervisors organized by locally-led organizations. This session will describe Dimagi’s learnings from building and testing a digitized FLW training model, in an effort to enhance cost-effective and quality service delivery in low-resource settings. For the last 20 years, Dimagi has digitally enabled FLWs in low- and middle-income countries (LMICs) through the open-source CommCare platform. CCC offers a novel mechanism for Locally-led Organizations (LLOs) and FLWs to opt into additional paid, purposeful work, and aligns with global digital health principles set by the WHO and Digital Development Principles by promoting accessibility, efficiency, and quality in healthcare. CCC’s digital training for FLWs adheres to the WHO’s digital health framework and emphasizes user-centered design, data-driven decision-making, and collaboration with local organizations. With a vision to improve jobs to improve outcomes, CCC provides a novel and digitally-supported way for LLOs and FLWs to: Learn how to deliver new high-impact services Deliver new, expanded services and reach more people Verify delivery of high-quality services Pay for verified services and seek new opportunities for growth and higher pay. The CCC Learn-Deliver-Verify-Pay (LDVP) model hinges on our ability to effectively digitize the ‘Learn’ process to upskill FLWs to deliver new interventions. Our core thesis is that most people learn skills through the process of practicing them. Instead of relying on knowledge tests to assess competence, we have developed visit observation forms that supervisors use to rate FLWs on their performance. We measure success by how well FLWs in our program perform based on these observation forms. Dimagi developed the training model through a series of UAT experiments in Malawi to fine-tune the digital learning content and approach. These were followed by iterative, short pilots to deploy the full LDVP model and measure FLW uptake and quality of service delivery. In each cohort, FLWs are invited to a job opportunity to deliver a new intervention in their community—such as a single-visit infant vaccine check or a multi-visit ECD intervention. Of the 263 FLWs offered a job opportunity through our initial 5 cohorts, 90% accepted the job, 88% started the digital learning process, 83% completed learning, and 79% delivered verified visits. To validate the effectiveness of training, we developed a ‘gold standard’ observation checklist that supervisors use to evaluate the services provided by FLWs, after completing the paid practice phase and beginning household visits. We selected ~50% of the FLWs in the cohort for observation, and found 82% met the criteria for quality visits in observation 1, and 88% by observation 2.
Leveraging CommCare Connect to Enhance Child Health: A case study on providing Vitamin A, Deworming, and MUAC Screening in LMICs – Saumyadeb Dasgupta and Sagar Atre
🗓️ 5 December, 2024 | 10:00-11:00pm EAT
Dimagi is developing a new initiative, CommCare Connect, to provide opportunities for Frontline Workers to learn, deliver, verify and be paid for high-impact interventions with an objective to digitally orchestrate millions of highly-cost effective interventions per year in Low and Middle Income Countries (LMICs). To this extent Dimagi has developed a Child Health Campaign within CommCare Connect (CCC) to scale door-to-door delivery of proven, impactful, and cost-effective interventions which includes delivering Vitamin A and deworming tablets, screening and referral for malnutrition, and options to include other impactful and localized interventions such as Oral Rehydration salts, zinc supplementation and chatbots with useful information (e.g., nutrition advice and information).
The learnings from the first 100000 service deliveries in Kenya, Zambia, Tanzania, Nigeria and India are discussed here. Frontline Workers (FLWs) and organizations enroll in the delivery of child health interventions through an application provided by Dimagi and are compensated based on the verified delivery of these interventions to their respective communities with the total cost amounting to $2 per child for this campaign. The impact of digital verification methods, GPS capture, timestamps and pictures at the point-of-delivery are highlighted to note their scalability. Digital collection of phone numbers also enabled follow-up messages or surveys, facilitating bundled interventions such as vaccine promotion and malnutrition screening. The CCC framework operates on a “Learn, Deliver, Verify, Pay” model to ensure effective and accountable health service delivery. Learning and verification systems ensure proper training for health workers, reliable service delivery, and accurate outcome measurement, leading to justified payments. This method enhances service quality, builds community trust, and reinforces existing health systems.
Insights from the field regarding operational logistics, challenges faced by frontline workers, and strategies employed to mitigate these challenges are reviewed in this paper. The study provides empirical evidence and practical lessons for policymakers, public health practitioners, and technologists interested in scaling digital interventions to address service delivery in low-resource settings.
Explore how CommCare Connect is transforming child health programs through Vitamin A, Deworming, and MUAC screenings in LMICs. Don’t miss this inspiring session!
Technology to Scale Mental Health Care
Cultural Adaptation in Resilience Building Interventions: Enhancing Mental Health Interventions for Frontline Health Workers – Tanya Kapoor
🗓️ 3 December, 2024 | 6:00-6:45pm EAT
In LMICs, frontline health workers often face challenges such as inadequate workplace support, heavy workloads, and limited career growth opportunities. These difficult conditions lead to burnout, physical, emotional exhaustion, job dissatisfaction, and decreased productivity. By 2030, there could be a predicted shortfall of 10 million health workers globally. This shortage makes it more critical than ever to support and retain the existing workforce. One strategy to address this issue is building resilience among frontline health workers. Fortunately, resilience is a skill that can belearned, making it possible to guard FLWS against burnout. At Dimagi, we leverage digital tools to offer frontline workers resilience training, equipping them with the skills and resources to combat stress and burnout. Effectively implementing resilience-building interventions requires adapting them to local cultures and contexts. This ensures that these interventions are relevant and meaningful, particularly in diverse environments. Key factors for the success of these interventions include acceptance by the target population, alignment with local customs, relevance to the community, and respect for diversity. Research shows that culturally adapted interventions yield better outcomes especially in resource-constrained environments. This systematic approach involves gathering local insights, modifying interventions to fit cultural nuances, testing efficacy, and consulting cultural experts to ensure the intervention is grounded in contextual and cultural reality. In this lightning talk, I share insights from my experience being the Content Lead for WellMe, a Resilience & Wellbeing application for Frontline workers. WellMe is built upon Johnson & Johnson’s framework of resilience-building behaviors and includes a comprehensive library of new content designed in collaboration with a resilience expert. This content has been tested with over 800 FLWs across India, Uganda, Nigeria, Ethiopia, and Malawi. For example, during our formative fieldwork in Katni, Madhya Pradesh, we learned that to build resilience, we must first be able to talk about it. Given the diverse languages and contexts, communicating about resilience relies heavily on localized feedback and insights. For instance, in Hindi, commonly spoken in many Indian states, there is no direct translation for ‘resilience,’ ‘stress,’ or ‘mindfulness.’ We explained mindfulness through familiar practices like yoga and meditation and used ‘tension’ as a proxy for stress. These examples highlight the importance of understanding and adapting content to meet the specific needs of FLWs. During the field work in Delhi this year, we learnt that FLWs often experience long working hours, leading to burnout and exhaustion. In response, in the application, we introduced a WellMe challenge that encourages FLWs to drink water regularly and take short breaks throughout the day. This challenge is broken down into smaller steps, starting with recognizing signs of stress and burnout, and includes setting SMART goals for hydration, with the app providing regular check-ins and reminders to help them progress towards their goals.Our work with WellMe aims to provide practical, culturally relevant tools to help FLWs manage stress and build resilience. In conclusion, adapting resilience building interventions to local cultures is not merely beneficial—it is essential for enhancing their efficacy and relevance, especially in settings with limited resources.
Mental Health Toolkit: Digital Solutions for Mental Healthcare Service Delivery – Lilianna Bagnoli
🗓️ 4 December, 2024 | 6:00-8:00pm EAT
This solution demonstration provides an overview of Dimagi’s mental health toolkit, designed to expand access to mental health care in low- and middle-income countries (LMICs). This session will highlight how Dimagi’s tools align with the World Health Organization’s (WHO) optimal mix of services for mental health. According to the WHO, there are a mix of services required to address mental health needs, which can be visualized as a pyramid having five distinct levels ranging from self-care (the base of the pyramid) to long-stay facilities and specialized services (the top of the pyramid). As you move from the base of the pyramid to the top, the cost of providing services increases while the quantity of services required decreases. In alignment with Dimagi’s focus on services that can be delivered at scale by Frontline Workers in low-resource settings, Dimagi’s mental health toolkit addresses the bottom three levels of the mental health services pyramid: self-care, informal community care, and primary healthcare. Collectively, these levels address the majority of mental health needs and can be delivered at relatively low cost by non-specialist providers. At the self-care level, attendees will see firsthand how Dimagi’s WellMe mobile application supports the wellbeing of Frontline Workers, promoting evidence-based skills that build resilience in this crucial population. We will also showcase PracticePal, a chatbot designed to help patients between therapy sessions, reinforcing therapeutic gains and providing ongoing support. At the informal community care level, we will demonstrate Dimagi’s tools for group therapy, which can be used by trained facilitators to deliver evidence-based psychological interventions in community settings. These tools aim to strengthen social support networks and improve mental health outcomes through peer-led interactions. Finally, at the primary healthcare level, we will highlight Dimagi’s tools for screening, diagnosis, treatment, and referral of common mental disorders. These tools are designed to be used by non-specialist providers, enabling them to identify and manage mental health concerns within their communities. Crucially, all of Dimagi’s solutions are built on open-source platforms, ensuring transparency, adaptability, and sustainability. Our tools are available in local languages and can be easily contextualized and adapted for use across diverse geographic, cultural, and linguistic settings, making them a valuable resource for global mental health initiatives. Throughout the demonstration, we will emphasize the toolkit’s user-friendly interface, adaptability to diverse cultural contexts, and potential to integrate with existing healthcare systems to bridge the mental health care gap in LMICs.
Replicable Models: How Technology Facilitates the Scaling of Effective Mental Health Interventions – with Lauren Magoun and Christie Civetta
🗓️ 5 December, 2024 | 9:00-10:00pm EAT
Mental health is a global concern, particularly in low- and middle-income countries where resources and specialized providers are often limited. To bridge this gap, task sharing models are being employed, where non-specialist providers (NSPs) are trained to deliver evidence-based mental health interventions. Digital tools can play a crucial role in equipping these NSPs with the resources and support they need to effectively deliver care. This panel session highlights how StrongMinds and Finemind are using digital tools to support their existing programs and NSPs. Participants will gain valuable insights into how the use of manual-based psychological interventions, combined with digital tools like CommCare, allows for rapid adaptation and deployment of these programs across different contexts and settings.
StrongMinds specializes in Group Interpersonal Therapy (IPT-G) for depression and is leveraging the CommCare platform to digitize their group sessions, which are based on an evidence-based WHO intervention approach that has been adapted and deployed across various settings. Finemind, a provider of stepped-care mental health services including Interpersonal Counseling (IPC) and Problem Management Plus (PM+), is utilizing CommCare to coordinate their multi-tiered care system. A single point of registration is used to screen and direct patients to an appropriate care pathway based on the level of treatment required. Together, the experiences of both partners offer a promising roadmap for the integration of digital tools to address the substantial mental health burden in the region and beyond.
Dimagi, a global social enterprise that powers impactful frontline work through scalable digital solutions and services, is the technical partner for these efforts. Leveraging its open-source platform, CommCare, Dimagi provides digital support to StrongMinds and Finemind, assisting them with solution design and development to scale their mental health programs effectively. The digital tools developed through this partnership are based on Dimagi’s toolkit of adaptable templates for mental health, and specifically templates for IPT-G and PM+. This approach allows both organizations to rapidly tailor and deploy the applications within their existing mental health interventions, ensuring efficient integration and minimizing disruptions to established workflows.
The panel discussion will delve into the experiences of StrongMinds and Finemind as they scale their programs, including their focus on ensuring fidelity to evidence-based interventions, managing patient data, and providing ongoing support to NSPs in remote areas. It will also highlight how digital tools can help to address these challenges, including automating data collection and analysis, facilitating remote supervision and training, and enabling stepped-care intervention pathways. Additionally, the panel will explore areas where digital tools still have limitations and discuss opportunities for future innovation to further enhance mental health care delivery in low-resource settings.
The digital tools being developed by Dimagi in partnership with StrongMinds and Finemind are available to other mental health partners seeking to leverage technology for similar impact. By making these tools accessible, we hope to empower organizations worldwide to replicate and scale evidence-based mental health interventions, ultimately expanding access to high-quality care.
